DEVELOPMEHT OF A GENERAL PURPOSE, Sh’éULATSON MODEL F08 EVAWATION AND SELECTEON 0F QPTiMAL DEMAND RESPONSWE BUS SYSTEMS Dissertation for the Degree of Ph. D. MiCHIGAN STATE UNIVERSiTY TAPAN KUMAR DATTA 1973 ABSTRACT DEVELOPMENT OF A GENERAL PURPOSE SIMULATION MODEL FOR EVALUATION AND SELECTION OF OPTIMAL DEMAND RESPONSIVE BUS SYSTEMS BY Tapan Kumar Datta As the number of applications of demand actuated public transit systems increases, it is essential that careful consideration be given to the evaluation and selection of Operating policies. There exists a broad spectrum of policy variables within demand actuated systems, and without a generalized evaluation procedure, consider- able inefficiencies can result from poor systems and operating strategies. As a part of this research, a general purpose simulation model has been developed which can replicate the operations of various types of Demand Responsive Bus Systems on a single data base. This model produces and aggregates both user (wait time, ride time, travel time) and operator (bus utilization, vehicle productivity, vehicle hours of operation, driver hours required) statistics in a format compatible with techniques for the evaluation of alternative system. The research also explored the effect of several variables (demand density, bus pooling policy, street network configuration, service area) on the economic and service characteristics of demand responsive bus systems. Comparative tables and figures are deve10ped and a systematic evaluation and selection process is demonstrated. Tapan Kumar Datta The selection criteria include user statistics such as ride time and waiting time and operator statistics such as total capital cost, Operating hours and vehicle productivity. Since the selec- tion of a system will necessitate a trade-off between service and operating costs, a technique for formalizing these decisions are presented along with results of an example application of the selection technique. The simulation.model developed in this research can replicate operations of various types of demand responsive bus systems under a many to one or one to many environments. The basic model is set up such that it can be changed to a many to few and/or many to many environment with minimal programming effort. The data produced by the model for a case study indicates that it can be used for any formal evaluation technique which quantifies community goals and policies reflected in user and operator characteristics. DEVELOPMENT OF A GENERAL PURPOSE SIMULATION MODEL FOR EVALUATION AND SELECTION OF OPTIMAL DEMAND RESPONSIVE BUS SYSTEMS By Tapan Kumar Datta A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the Degree of DOCTOR OF PHILOSOPHY Department of Civil Engineering 1973 J.“‘”‘ a/ This Thesis is Dedicated To My Beloved Wife Mira ii ACKNOWLEDGEMENTS The author is deeply indebted to his major academic advisor, Dr. William.C. Taylor, Professor and Chairman of the Department of Civil Engineering, for his continuous guidance and assistance during the course of this research and preparation of the thesis. The author also wishes to thank the other members of his Ph.D. Committee, Dr. John B. Keer, Dr. G. Bloomquist,Dr. D. Anderson, for their timely and invaluable guidance in the conduct of the research and prepara— tion of the thesis. Sincere appreciation is extended to the Transportation Research and Planning Office of Ford Motor Company for supporting the initial model deve10pment part of this research. The author also eXpresses appreciation for the partial support provided by Goodall, Grivas and Associates, Inc. during the conduct of the research. The author wishes to thank Mr. David M. Litvin, Mr. Ken 8. Opiela for helping him in the conduct and preparation of this thesis. Special appreciation is extended to Mrs. Pat G. Wagner and Mrs. Pat A. Schneider for their help in the typing of the thesis. The author also wishes to thank his wife Mira for her continued encouragement and help during the entire period of this research. iii LIST OF LIST OF SECTION SECTION B01. B02. 3.3. 3.4. SECTION RESEARCH SECTION D.1. D.2. D.2. SECTION E.l. E.2. E.3. E.4. E.5. SECTION APPENDIX I APPENDIX II TABLE OF CONTENTS FIGURES TABLES — -_ A. INTRODUCTION ___ B. STATE OF THE ART—— _ Demand Actuated Bus Systems- as defined by researchers===~ Published Papers and Project Reports on DAB Systems Experiences of Implementation Studies on DAB Systems-=—— Conclusions of the State of the Art ---------- C. PROBLEM STATEMENT & OBJECTIVE OF THE D. METHODOLOGY System Definition -— a. Simulation b. The Test System- — E. RESULTS System Performance Characteristics - System Evaluation and Selection Pooling Policy Effect of Various Network Configurations ------- Effect of Size of Service Area F. CONCLUSION iv Page v vii 10 10 14 39 48 52 55 55 59 7O 77 77 90 106 117 125 134 140 148 Figure A.1 A.2 B.2.l B.2.2 B0203 B0204 B.2.5 3.2.6 B0207 B.2.8 B.2.9 3.2.10 3.2.11 B.2.12 B.2.13 C.l D.2.b.l D.2.b.2 D020b03 LIST OF FIGURES Transportation Trends -- Transportation Trends Basic Diversion to D-J Based on Fare -------- Profit and Loss Versus Ridership, Fare Box Revenue Operating Costs for Manual Versus Computer Assignment- Influence of Accuracy of Ridership Estimate on Profit P-L Versus Ridership, Two-Thirds Capital Grant P-L Versus Ridership, Effect of Wage Rate — - Effect of Demand on Vehicle Productivity Effect of Area Size on Vehicle Productivity--- -__ Relationship Between Number of Vehicles and Level of Service Sensitivity Data of Northwestern Study—--—-- Sensitivity Data of Northwestern Study---- Results of Ford Motor Company Studies ------ Results of Ford Motor Company Studies- ----- Optimal Regions for DAB Systems Alternatives —— Basic Network Configurations---- Comparison of Simulated Productivity with Operating Systems Comparison of Simulated Results and M.I.T. Study #nw--- Page 29 29 3O 30 31 31 32 33 34 35 36 37 38 54 71 75 76 Figure Page E.1.l Systems Performance Functions - Wait Time -------------- 84 E.1.2 Systems Performance Functions - Ride Time —— 85 E.1.3 Systems Performance Functions - Bus Requirements- — -------- 86 E.1.4 Systems Performance Functions - Vehicle Productivity — 87 E.1.5 Systems Performance Functions - Driver Hours Required—— 88 E.1.6 Systems Performance Functions - Total Vehicle Hours of Operation 89 E.2.1 Utility Cost Functions — 95 E.2.2 Utility Cost Functions --- 96 E.2.3 Utility Cost Functions — 97 E.2.4 Utility Cost Functions 98 E.2.5 Utility Cost Functions - 99 E.2.6 Utility Cost Functions 100 E.2.7 Typical Distribution of Rider Demand--------- 102 E.3.1 Pooling Policy Data - Waiting Time—— 111 E.3.2 Pooling Policy Data - Mean Ride Time- --------- 112 E.3.3 Pooling Policy Data — Driver Hour Requirements ———— -—— 113 E.3.4 Pooling Policy Data - Vehicle Hour' Requirements — 114 E.3.S Utility Cost Functions for Pooling Policy ----- 115 E.4.1 Effect of Network Configuration on Cost------— 124 E.5.1 Utility Cost Function for Variable Demand Levels 133 vi Table 8.3.1 B0302 D.1.1 D.2.b.l E.1.l E.1.2 E.1.3 E.1.4 E.2.1 B.2.2 E.3.1 E.3.2 E.3.3 E.3.4 E.4.1 E.4.2 E.4.3 E0404 E.5.1 E.5.2 LIST OF TABLES Demand Actuated Bus Implementation Information Demand Productivity Relationship of 7 Dial-A-Ride Projects-— - Performance Criteria Levels in the Hierarchy of DAB Systems Simulated Data for Variable Route- Variable Headway Systems — System Performance Data — Fixed Route Fixed Headway - System Performance Data - Fixed Route Variable Headway — — System Performance Data - Variable Route Fixed Headway-~- System Performance Data - Variable Route Variable Headway Utility Costs of Alternative Systems ----------- Costs of Various Operating Policies-- — System Performance Data - Pooling Policy -------- System Performance Data - Pooling Policy—------- System Performance Data - Pooling Policy-----—-- Utility Cost Data-- System Performance Data for Network Configuration System Performance Data for Network Configuration System Performance Data for Network Configuration Utility Cost Data — System Performance Data System Performance Data vi 4 Page 46 47 56 74 78 79 80 81 103 104 107 108 109 110 120 121 122 123 127 128 Table Page E.5.3 System Performance Data — 129 E.5.4 System Performance Data — — 130 E.5.5 System Performance Data— -- 131 E.5.6 Utility Cost Data for Service Area-------- 132 viii A. INTRODUCTION Mass transit was the most important mode of transportation until the early twentieth century when automobiles started influencing the life style of Americans. The personal mobility offered by the auto- mobile increased sharply with the availability of personalized trans- port container and this brought about a revolution in the trip making characteristics of the people. The life style of the people changed as did the physical dimension of our cities. The auto-highway system has evolved to the point where it now accounts for about 90 percent of the nations intercity travel. In many ways, it has shaped its own environment. However, with an increasing population in metropolitan areas, this environment is showing signs of trouble. Traffic congestion, air and noise pollution, accidents, displace- ment of pe0p1e, the unfilled needs of those people who do not drive, unfilled partially because of the decline in alternative forms of urban transportation--a11 of these are causing concern in the industry, in government, and among the people. In addition, the nation's transportation "system" was not designed for integrated use by the various transport forms. The resulting in- terface problems are well-known today in terms of congestion, delay and inconvenience. Existence of mass transportation system dates back to the horse and buggy era. Although some form of public transportation system has been present since this time, they are not well planned to form a balanced transportation system. In early years patronage of mass transportation was due-to economic and technological reasons. The availability of a personalized transport container was limited by high cost and lack of technology. The increasing automobile popularity in the twentieth century, brought competition to mass transportation systems. Public transportation in urban areas consisted of steel tracked street cars, trains, rubber wheeled trolley buses, and some motor buses in the late forties and early fifties. However, as auto volumes in- creased fixed track systems in the same right of way caused increasing problems. Provision of separate right of way was economically infea- sible in most urban areas due to declining ridership. The result of this dilemma is evident today. Most metropolitan environments have eliminated fixed track carriers (street cars) and instead have multi- passenger automobiles or what we call buses. This mode of public transportation (Bus) has more flexibility in terms of reaching to and from potential origins and destinations than the fixed tracked mass transit systems like the street cars. Buses can travel anywhere within an urban environment without additional ex- penses for facilities and right of way. As bus systems replaced trolley lines, analysts and operators started to lay out the bus routes with the goal of providing as much coverage of the urban areas as possible. During the first few years, most systems were not only self-supporting but also profitable. How- ever, the increasing usage of the automobile started to affect the bus ridership and most private enterprises faced regular losses in their Operation. This decline in.mass transportation patronage has been recorded over the last decade. As a result, public transportation usage gener- ally does not offset the Operating costs of such systems. As the sys- tems experience Operating deficits, many privately owned systems either cease to operate entirely or sell their depleted Operating systems to public agencies. Intermediate measures, such as service cuts, in- creased fares and public subsidies have all proved to be ineffective in either making public transportation systems Operate at cost or to increase patronage. Concern for these problems and their relation to other urban ills has resulted in a nationwide movement toward major changes and improve- ments in the transport of people and goods. The Federal government has defined an integrated and balanced transportation system as a national goal. Its achievement is a top priority item of the current Administration, and a wide range of efforts is moving forward. Widest in scope and concept are system analyses attempting to define national and regional transport needs ranging up to fifty years into the future. These integrated planning efforts can be attributed to the Urban Mass Transportation Act of 1964. This act was amended in 1966 to include a new subsection which reads as follows: "The Secretary shall, in consultation with the Secretary of Commerce, undertake a project to study and prepare a program of research, development, and demonstration of new systems of urban transportation that will carry people and goods within metropolitan areas speedily, safely, without polluting the air, and in a manner that will contribute to sound city planning. The pro- gram shall (1) concern itself with all aspects of new systems of urban transportation for metropolitan areas of various sizes, including technological, financial, economic, government and social aspects; (2) take into account the most advanced available technologies and materials; and (3) provide national leadership to efforts of States, localities, private industry, universities, and foundations. The Secretary shall report his findings and recommendations to the President, for submission to the Congress, as rapidly as possible and in any event not later than eighteen.months after the effective date of this submission. In addition to the systems planning studies, a second level of effort is aimed at develOpment of specific concepts, hardware and components. At the lower end of the spectrum are attempts to improve existing sub- systems through innovation until long-term solutions can be provided. As a result of this legislation, considerable attention has been focused on development and testing of innovative mass-transportation systems. Some of these systems consist of completely new hardware as well as software technology. Others consider innovation in system technology which utilizes basic existing "facilities" and modified "carrier". Demand-scheduled bus system (DSB), Dial-A-Bus (DAB), Dial-A-Ride (DAR), Demand Jitney (DJ), Computer aided routing system (CARS), etc., are some variety of methods being proposed by various agencies which utilizes basic highway facilities and consists of some modified version of existing mass-transportation vehicles. The continuing decline in ridership can be attributed to several interdependent causes. Decentralization and affluence contributes to the ridership decline, as does the changing life style of Americans. Whatever may be the cause, it is evident from existing studies and research that public transportation services must be re-oriented towards todays transportation needs if this trend is to be reversed. It has become evident that a public transportation system will attract poten- tial riders only when the service is such that it can supplement cer- tain auto trips without causing major inconvenience or discomfort to the public. The trip making behavior has been following a trend as shown in figure A-l, at least for the past decade. The number of person trips by mass transportation system has been decreasing slightly, but mass transit trips as a percent of total trips is decreasing rapidly as person trips by automobiles have increased (figure A-l). It may not be possible to reverse the trend of overall auto trips by introducing innovative systems. However, the rate of increase of auto trips can surely be slowed by using viable mass transportation system. Prior to theediscussion of the potential impact of demand respon- sive bus system, it is pertinent to define the potential market. A Person Trips DAB Riders Diverted From Auto to Bus I / a’ / / 1’ a’ By Auto fo M888 TIW t 1 tn tntx (today) Time Figure A.l Transportation Trends person who does not have a car or does not have access to a car during a trip is defined as a captive rider. A person who uses an automobile, but may change to a public transportation system if and when a certain level of service is provided, is called a potential rider. There are also generated riders, who would not have made a trip, but for the high level of service of a public transportation system. Those three definitions can be illustrated in figure A-Z. Say tl is the time of reference when auto travel started increasing and the total number of mass transit trips remained more or less constant as shown. Let tn is the time reference when, by introduction of a higher level of in mass transportation service riders are attracted to mass transit. 't 'is any point on the time scale occurring at a later date n+x than 'tn'. M represents the captive rider. 'Al' indicates the auto riders who are diverted from auto travel (the potential riders) and M2 is equal to A1. M3 represents the new transit trips made who would not have made the trip unless the mass transit system provided a higher level of service (the generated rider). General Concept of Demand Actuated Bus System It was recognized that there are several ways of improving the service of a bus system. Studies were conducted to accomplish the ob- jective of improving the level of service as early as the 1950's. The first concept used to provide better service was the express bus system. The next major improvement came when premium service was tested in demonstration programs in Peoria and Decatur, Illinois. The Person Trips Auto Travel ’/Mass Transit Travel / 2 n ntx (today) Time Figure A.2 Transportation Trends concept of Operating bus service without having fixed routes and fixed headway but responding to calls for door to door service came next. The flexibility in routing and scheduling of buses on demand is the basic underlying philosophy of these systems. However, under this acronym there are various alternative Operational strategies available. Depending on the operational alternatives and service Options, it is possible to define a number of distinctly separable bus systems. These can be arranged in the order of level of serviceto form a hier- archy of bus systems. Since passenger demand over space and time will determine route and schedule of these bus systems, the entire family of bus systems falling within this spectrum are classified as demand actuated. It is important that we understand the relationship between the operational variables used to define this system, their respective service characteristics, which will determine their impact on potential and generated riders, and their cost. Only with this thorough under- standing can.we select the Optimal system for any given environment. The purpose of this research was to develop a generalized simula- tion model, which can replicate various operating strategies of Bus- Transit systems for various service Options and provide necessary sys- tem performance data for system evaluation and selection. 8.1. DEMAND ACTUATED BUS SYSTEMS - AS DEFINED BY RESEARCHERS AND ANALYSTS Considerable research activity has been pursued both funded by public and private agencies to investigate and determine the guide lines and feasibility of implementation of such demand responsive systems. 'Ford Motor Company is currently pursuing a program of company- funded research in urban transportation. This work is being carried on by the Transportation Research and Planning Office that is staffed by a multi-disciplinary team of researchers and engineers. The pro- grams cover a wide variety of critical problem areas and place strong emphasis on the proper role of public transportation as well as the use of the personal vehicle. To make certain that these problems are relevant and useful, we work closely with transportation system opera— tors and specific communities. Typical of this approach is the work on dynamically dispatched public transportation conducted by Ford Motor Company'.1 Researchers at General Mbtors Tech Center were involved in de— fining and testing the same basic concept. "It was the primary purpose of this study to provide information to aid in decisions regarding the merit of demand reaponsive trans- portation systems and the need for vehicles specifically designed for D-J transportation systems." 1Incremental Implementation of Dial-A-Ride systems by Karl Guenther .R. B. Special Report 124. Case Study of a Demand Responsive Transportation System by H. J. Bauer, H.R. B. Special Report 124. 10 ll Researchers at M.I.T. were forerunners in the field of DAB and were involved in utilizing simulation technique for evaluating the feasibility of such a system. '"Dial-ArBus", a public transportation system offering the desir- able characteristics of automobile and taxi travel at a cost only slightly higher than conventional transit, provides door-to-door ser- vice with maximum waiting and travel time guarantees. The principal objectives here will be to review the results of research conducted at the M.I.T. for the U.M.T.A. and related to the feasibility of dial-a- bus systems.‘1 The State of the Art of DAB indicates that researchers have one thing in common, the utilization of simulation techniques for the basic evaluation of systems they defined. It is noteworthy to look at the various definitions that have come from various researchers from the basic concept of demand-responsive bus systems. 1. "Before exploring the essential modeling questions, it is important to briefly describe the CARS system. CARS is an attempt to provide high quality mass-transportation service to complement the automobile as a major urban transportation mode. It offers door-to- door service using small multi- passenger vehicles. Someone using the system would phone his request into a control center consisting of a digital computer with peripheral communication equipment. The digi- tal computer assigns vehicles to serve passengers and, at the appropri- ate time, causes a message to be sent to the vehicle informing the 1"Dial-A-Bus System Feasibility", by Daniel Roos, H.R.B. Special Report 124. 12 driver of his next stOp. The vehicle travels the existing street network, generally carrying more than one passenger, and collecting and delivering passengers as instructed by the computer."1 2. "Here's how Dial-A-Ride works. The customer calls the dis- patcher on the telephone and tells him where he wants to go and when he wants to be picked up. The dispatcher plans the most effective route for the requested time period, considering all other requests. He gives the call to the vehicle driver on the two-way radio. The driver, following a route sequence specified by the dispatcher, picks up each customer wherever he is and delivers him to his destination."2 3. "TELEBUS" is a hybrid form of public transit, incorporating the demand responsive characteristics of taxi systems, while accommo- dating multiple requests for service at one time. By being demand re- sponsive, TELEBUS is able to reduce or eliminate completely the delay times associated with walking to and waiting for conventional public transit services. The heart of this system lies in its ability to provide the con- venience of limousine service on demand, to and from the user's home. This depends, in turn, on an effective dispatching system. In initial, smaller scale experiments which have been carried out to date, manual ledeling the.CARS System by J.D. Kennedy & N.H.M. Wilson - Paper pre- sented to Ohio Transportation Research Forum - May 1969. gDialeA-Ride is here by Stephen D. Brumble - Paper presented in 1972 National Planning Conference of the American Society of Planning Officials. 13 dispatching has been used, similar to taxi dispatching. In larger scale, TELEBUS operations dispatching would be handled by a central computer.1 An overview of the above few examples of definitions indicate that though the terminology used by various researchers and analyst vary, the essential features of all such systems belong to demand actuated bus systems. The implementation studies of various demon- stration projects reveal that even with varying levels of service in different projects the transit analysts tend to call them generally Dial-A-Bus or Dial-A-Ride. The research activities performed to date for analysis and eval- uation of applicability of various systems to different environments generally utilized simulation techniques. JSummary Report - Regina TELEBUS Study. B.2. PUBLISHED PAPERS AND PROJECT REPORTS ON DAB SYSTEMS It is important to realize the various interesting phenomena observed and experienced by the researchers, analyst, and Operators of demand actuated bus systems. This section summarizes the observations of the theoretical studies and papers relevant to the proposed research in the field of DAB. Specific attention has been directed towards the work at Northwestern University, Westinghouse Brake Co., MIT, Ford Motor Company and GMRL. A critical review of the model assumptions, operations and reported results is made highlighting their differ- ences, similarities and possible limitations. Specific attention have been directed towards the following in reviewing the relevant state of the art.1 1. The proposed levels of service. 2. The selection of a vehicle for pick up and delivery of passengers. 3. The ability to handle groups of passengers with common origins and with or without common destinations. 4. The method of contact and the frequency of communication with individual vehicles. 1Howson, L.L. & Heathington, R.W., "Algorithms for Routing and Scheduling in Demand Responsive Transportation Systems". H.R.B. Record 318. The model assumptions and operations are mostly adopted from this paper. 14 15 5. The method of describing the street network character- istics of the area to be served. 6. The dispatching policies. 7. The cost elements of DAB systems. Level of Service The Northwestern simulation package had 3 items included in the level of service-dminimum and maximum pick up times and a maximum travel time. The minimum pick up time was automatically set at 1 minute. The maximum pick up time guaranteed the passenger that he would be picked up before a specified time limit has elapsed; the maximum pick up time was arbitrarily chosen to be 6 minutes. The maximum travel time was a linear function: kn, for n'i n' Max. Travel Time = T + en, for n > n' where n - the number of links between origin and destination (a link is 1 block long) n = a control parameter constant set equal to 10 links k a a control parameter constant that equals 1 minute per link T - a control parameter constant that equals 5 minutes (D I a control parameter constant that equals one-half minutes per link The level of service as defined by these three constraints were guaranteed to each individual demand for service. In the analyses reported in this study, 100 percent of the demand was 16 assured this level of service, and the operating and dispatching policies were thus constrained. The Northwestern study1 reported the sensitivity of wait, ride and total travel time for various levels of Service. In this case, level of service have been used as "maximum pick up time" and "link travel time". The WABCO guaranteed level of service used a single measure of time that includes the waiting time. This time varied from 2 to 6 times the automobile driving time. The automobile driving time was assumed to be based on 20 mph average speed plus a fixed value of 2 minutes. The WABCO simulation was not constrained so that each passenger was not guaranteed a "phone to destination" time. However, in the simulation runs, 95 percent of all passengers were served within the prescribed time. The M.I.T. simulation work used a waiting time of 15 minutes and a linear function of the distance as a travel time constraint. The M.I.T. study2 reported the sensitivity of "vehicle require— ments" for various levels (figure B.2.9). The level of service is defined as the ratio of total service time by Dial-A-Bus to total travel time by automobile. 1Bruggeman, J.M., Heathington, R.W., "Sensitivity to Various Para- meters of a Demand Scheduled Bus System Computer Simulation Model." 2Roos, Daniel, "DialeA-Bus Feasibility", H.R.B. Special Report 124. 17 The GMRL model also used guaranteed pick up and delivery time as a measure of the level of service. This study simulated bus operations for 4 systems (A, B, C, D) on the basis of the level of service. The simulation results were related to trip diversions for various levels of fare (figure B.2.l.). Ford Motor Co's3 Transportation Research group investigated the sensitivity of "level of service" as functions of: 1. Service area (ref. fig. 3.2.12) 2. Demand (ref. fig. 3.2. 12) 3. Vehicle capcity (ref. fig. B.2.13) where "level of serviCe" is defined as the ratio of total travel time (from call to delivery) via Dial-A-Ride to the direct driving time. They also recognized that this definition does not account for the delays at either end of the auto trip, it tends to over estimate the actual ratio of door to door time. All of the above mentioned studies used time as the only variable in describing level of service. Some allowed weights for total travel time and waiting time whereas others did not differentiate between them. Vehicle to Passenger Assignment Rules The NOrthwestern University simulation program specifies a methodology for vehicle to passenger assignment. As a call for 1M'ason, F.J. and Mumford, J.R., "Computer Models for Designing Dial-ArRide Systems". Automotive Engineering Congress, Detroit, Michigan, January 1972 . 18 service is received, it scans the location and usage of all the VBhicles that are currently operating in the system. The vehicle which is nearest (by distance) to the origin of the call is the first one considered for service. It is then checked to see if the new demand can be serviced without violating any of the pas- senger service criteria that have been established for passengers already in the bus or for those passengers that are scheduled to be picked up by this bus. If the nearest bus to the new call is not able to service due to violation of any commitment, then the next nearest bus is selected and is scanned to see if it can ser- vice the new demand. When all the buses in the system are found not to be able to meet the new demand, then a new bus is sent from the terminal to service the call. When a new bus is generated, it then is considered to be in the system and will be considered for any other incoming calls. Since the closest bus is always the first one to be considered as a possible server, there is really no guarantee that the vehicle to passenger assignment will be Optimal. The simulation model places a priority on the pick up of new demands at the earliest opportunity. When a passenger is assigned to a vehicle, the vehicle will pick up first before making any drops not directly on his existing route pattern. In the Wéstinghouse Air Brake Co. simulation, a given vehicle makes a pick up or a delivery to serve the caller or the passenger with the highest priority. For the caller (the passenger that is 19 to be picked up), a priority is assigned that has an arbitrary value of 110 minus his distance from the vehicle under considera- tion in grid units minus the callers origin to destination (O-D) distance in grid units. A grid unit is a function of the grid density; for example, in an area of 25 square miles and a 20 by 20 grid, the grid unit would have a value of 0.25 miles. For those users already on the bus, the priority used was 100 minus the destination distance in grid units plus one unit for every five minutes of total wait time plus one unit extra for every 10 minutes of total wait time.1 If a passenger has been waiting for 40 minutes, including the time that he spent on the vehicle, he is assigned a tap priority. The WABCO model uses one additional vehicle designated to handle the longest trips and the longest wait times. This vehicle attempts to handle those callers who cannot receive service from the other vehicles in the system. A passenger waiting to be picked up will be given a priority relating to the largest value computed by adding his O—D distance in grid units to his waiting time measured by 1 unit for 3 minutes of wait, 1 unit for additional 2 minutes, and 1 unit for additional 1 minute of wait time. If this extra vehicle cannot keep up with the task of handling the callers with longer 1Study of Evolutionary Urban Transportation. Westinghouse Air Brake Company, Volume 3, Appendix 4, Section 3.9, February 1968 20 waiting times, provision is made for other vehicles to assist by assigning a tap priority to those waiting over 50 minutes. In the M.I.T. CARS project, a vehicle is selected to serve a new demand depending on the criteria of waiting time of the new user, his travel time, his overall service time, link constraints, and the travel constraints of all current users. The link con- straints are the specified time that a vehicle is due at a parti- cular node in course of the tour of the bus. The travel con- straints of all current users are waiting time, travel time, and total service time. The new user is assigned to a vehicle that can serve him without violating the travel constraints of those already traveling and who are scheduled to be served by the parti— cular vehicle. General Mbtors Research Laboratories simulation model assigns vehicle to passengers on the basis of an weighted function of: A. Minimizing the increase in travel to those already on board. B. Attempting to minimize the waiting time and travel time to the new demand. C. Minimizing the deviation of the vehicle from a given path. D. Reducing the number of buses in the system. The weighted summation of the quantities relating to these criteria was used as the cost (in time) of providing service. A cost increase was considered if the new demand assigned to a bus 21 under consideration had a scope for minimizing cost on a system wide basis. The value used for the weights was arrived at, once the sys- tem performance levels were set, the necessary values for the weights to make the system cost the least to the user. In formulating the vehicle to passenger assignment policy, some optimized the user travel characteristics whereas others looked at the entire system and attempted to optimize the system characteristics. The Ability to Handle Groups of Passenggrs with Common Origins and With or Without Common Destinations The NOrthwestern University's simulation package does not specifically handle multiple origins or destinations. If multiple origins or destinations need to be handled, they would have to be treated as separate and individual destinations. For example, if there were three people desiring to make a trip together from a common origin, this would be treated in the simulation model as three separate demand calls for service. Also, each destination would be treated separately. Thus, there would be three destina- tions used by the simulation model. Three people, perhaps of one family, desiring to have service to the same destination might be assigned to more than one vehicle to obtain the service. The Westinghouse Air Brake Company (WABCO) report does not specify its method of handling multiple origins and destinations. 22 It would seem from their report that they also treat multiple ori— gins and destinations in the same manner as the Northwestern Uni— versity simulation package. In the M.I.T. formulation of the problem, a demand may consist of more than one person and is referred to as a passenger group. For each passenger group there is a unique origin and destination. The possibility exists that more than one bus would be used to satisfy the demand for travel from this one origin, if the destina- tions were not coincident. The GMRL model handles groups of passengers with a common ori- gin and with or without a common destination. The passengers at the same origin are picked up by the same vehicle treating them as a single group. If they have the same destination, they are treated as one step each for pick up and delivery, considering extra em- barkation and disembarkation time depending on the number of pas- sengers involved. If the passengers originating from one point have different destinations, the vehicle makes separate stops for unloading, like two different demands. The Method of Contact and the Frequency of Communication with Individual Vehicles The Northwestern University simulation package assumes con- tinuous contact with all vehicles operating in the system. There was no prOposal as to the methods of contacts as far as electronic equipment is concerned. 23 The WABCO model also assumed automatic vehicle monitoring. WABCO, having problems with the general purpose simulation program (GPSS), used a negative time feature to account for automatic vehicle monitoring. This study suggested the possibility of two way radio communications as a means of achieving this constant monitoring of the vehicle. The M.I.T. Dial-A-Bus project considered two possibilities: 1. Continuous communication 2. Discrete communication For continuous communication, the simulation model allows instant reassignment of demand. For the discrete communication system, in— termediate reassignment is not possible since the precise location of the vehicle is only known when passengers are either picked up or delivered. GMRL model allows communication on demand. New passengers can be assigned to vehicles and re-routed if vehicles could be located when desired. The GMRL approach seems most viable in terms of system costs and simulation of real world situations. Method of Describing_yarious Network Patterns The Northwestern.University package used a simple, 1 mile square grid, This grid was in turn divided into 100 basic units that represent the intersections of blocks. Equal travel times were assumed on all links of the network. The WABCO simulation also used a square grid. Their square grid ranged from a 1 mile by 1 mile square up to 5 mile by 5 mile 24 area. However, the grid unit size varied from one tenth to one- quarter mile. Like the Northwestern simulation, the WABCO origin and destination points occur at grid intersections. The link travel time appeared to be uniform over the entire area. The M.I.T. work of an earlier date was based on a rectilinear grid ranging from 1 mile by 1 mile to 3 mile by'3 mile area. The recent work discusses a network that is based on airline distances between points, using the reasoning that the specific street network under study is not pertinent to the grasp of the fundamental algo- rithm concepts. The GMRL used areal network with one way links. The size of the GMRL study area is approximately 36 square miles. Because of the desire to be able to pick up a new demand within 5 or 10 minutes after a request for service is received and an average speed on the street network of 20 mph, it becomes apparent that 1 station or terminal located in the center of the city will not be sufficient. At a speed of 20‘mph, it would require 0.21 hour or 12.6 minutes to travel to the corner of the area from the center. Therefore, there will be more than one terminal required if either the 5 or 10 minute pick up service guarantee is used. The networks used in the various studies are either very simple like grid network, or very specific system like a case study. The airline distances were considered in one study. Thus, the impor- tance of network configuration in the overall system performance 25 has not been tested. None of the studies reported the effect of variability in the network pattern. The assumption of only one way streets in a real world case study is possibly unrealistic. Dispatchingnggig In the Northwestern simulation model, when all of the vehicles on the grid cannot service a demand call, then a new vehicle is dispatched from the terminal to service this demand. This vehicle will not be dispatched until after all of the other vehicles have been rejected. The WABCO simulation designates a special vehicle that handles particular situations. When WABCO's 3 vehicles on the system can not service a given demand, then the special vehicle handles this individual demand. Each individual demand that cannot be serviced by the other vehicles is assigned a special waiting time measured by 1 unit for 3 minutes of wait, 1 unit for 2 minutes, and 1 unit for 1 minute of wait time. Once the extra vehicle is on its way to pick up a passenger and if there is a new demand located on the way to this origin, he will also be picked up. Essentially, the extra vehicle uses an.algorithm that is designed to handle the longest trips and those waiting the longest. In the CARS project a bus is dispatched when none of the other vehicles in the system.can service the demand. When a demand call is recorded on the system, the vehicles which are currently operat- ing are scanned to determine which vehicles can service the demand. 26 If none of the vehicles can serve the demand, due to violation of constraints 'of current users, then a new vehicle is dispatched from the station. The GMRL model considers dispatching a new bus from the term- inal when none of the buses currently in the area can service a new demand because of violation of user guarantees. In this case, a stOred bus will be dispatched from the nearest terminal to the demand. Cost Element of DAB Systems The cost of DAB operation is the critical issue in the feas- ibility of such a system in any environment and almost all researches and analysts attempted to address this in.various forms. The Northwestern study looked at cost per passenger mile of travel time as a function of (ref. fig. 3.2.10 and B.2.11) A. maximum pick.up time B. link travel time C. demand frequency D. ratio of sides of service area E. bus capacity The bus operating characteristics have been converted to Op- erating costs. The unit costs data used for evaluating system costs are as follows: A. vehicle cost - $5,000 ammortized over 10 year period of 62 interest and based on usage of only 4 peak hours per day (1,000 hours per year). 27 B. driver-labor cost was set at $3.00 per hour per vehicle generated and came to $3.35 per simulated period which allowed for an average of 7 extra minutes required by the up and down operation. Finally, a cost of 2.2 cents per vehicle minute of operation, which corresponded to 12 cents per vehicle mile was used to cover the cost of vehicle Operation. Other costs like terminal facilities, taxes, licenses, administration, etc. were felt relatively constant and as such not considered in the system cost analysis. The North- western study presented sensitivity of user cost elements (wait, ride and total travel time) separately. The M.I.T. study as presented by Roos (Dial-A-Bus feasibility, H.R.B. Special Report 124) did not attanpt to convert either user or operator characteristics generated by simulation results to any dollar figure. His presentation dealt with operator cost element as vehicle productivity (no. of trips/vehicle/hour) as a function of (ref. fig. B.2.7.and .8) A. demand density B. service area C. mean level of service However, the entire work was performed on the basis of a single system strategy (Dial-A-Bus). GMRL1(General MOtors Research Laboratory) also conducted a system cost study in their case study area for Demand-Jitney System. 1Bauer, H.J., "Case Study of a Demand Responsive Transportation System", H.R.B. Special Report 124. 28 Their definition of alternative systems pertain to variation of passenger wait time and ride time only. Costs were expressed as a function of both peak period demand and demand for specific hours, and were develOped from the hourly distribution of estimated rider- ship. The GMRL simulation study used fare as one of the control variables to look at profit and loss. They have used wait time and ride time of the passengers as indices of level of service and all of their cost functions are presented for fixed levels of user characteristics (refer. figures B.2.2 to B.2.6). TRPO (Transportation Research and Planning Office) of Ford Metor Company1 did not attempt to convert the system parameters to dollar values. The approach used in this study was to identify and quantify the variables that would be used in an economic analy- sis, but not to affix prices to these variables. All of the studies mentioned above treated cost (assumed dollar values) or cost elements as independent of demand parameters except the GMRL study which considered demand as a function of cost. This study also looked at the sensitivity of demand due to various fare structures. lMason, F.J. and Mumford, J.R., paper presented at Automotive Engineering Congress, Detroit, Michigan, January 1972. 29 'm- "0.!“ '0'“ 7m 0 ~ ~‘ Z ‘0 .. . ~~§‘ 2 e ~~‘~ \ \g x g l ‘L a ‘i n ~. a ‘o\ ~s“ Q r . 4 r- ..\o. ~~‘§~ — Q Q E .\.. ~. ‘a..._0uh Sm- wmnm swam ‘ ‘ " "' --- — c 15 Min. 3 at Auto w". D 25 Min. 3 I Auto 0 l l . r I l 1 0 .IO .70 LG I.” From: FARE IN DOLLARS Fm7.schoncoosbu-donm.) System", H. R. B. Special Report 124 Figure B.2.l. 0 c J H. J. Bauer, "Case Study of Demand Responsive Transportation Basic Diversion to D-J Based on Fare RIDERS-"P IN 'WOS g.' g... ,svstfiu‘fi ;.. aC/WW" 3.9 \X’ g,“ a...» «27“ \\ g‘ '1‘ ”\ \ R. — %” —-'3 — \AA‘ g..__='-'= .....f... VA Hun 10.0“)“: and loss mu rm. two-hos m.) From: H. J. Bauer, "Case Study of Demand Responsive Transportation H. R. B. Special Report 124 System", Figure B.2.2. PrOfit and Loss Versus Ridership, Fare Box Revenue 30 O M .. a /"/ 'O 2 mm enormous ’7 3 an x 444’ 8 E g. murmurs mourns no : l ‘ v : t / ', I uses on svsrm "o" 5 mm mas maxmum - as muons TRAVEL rm: - a 1: AUTO 0 o as 600 1.0m EMMY” Flows". MM'UWWWWMWM.) From: H.J. Bauer, "Case Study of Demand Responsive Transportation Systems , H.R.B. Special Report 124 Figure 3.2.3. Operating Costs for Manual Versus Computer Assignment 9. ,, u.» out ”DEBS!!!" UN man.) 0 ‘0 I3 ” I! l . ‘Numdm : useo ON SYSTEM "0" 4 \ munme (momma) - 25 muons \ TRAVEL mas - s 3: AUTO In", A \\ . ......" 5......" 5' P In l l ma mom/Loss are our m 31.000: in Iv Hum 12. (Mdmdfidwflpmhmwonmfit.) From: H.J. Bauer, "Case Study of Demand Responsive Transportation System , H.R.B. Special Report 124 Figure E.2.4. Influence of Accuracy of Ridership Estimate on Profit 31 O C RIDERWJN t0“). 3 . ” so u S ' T L E . ".3. svsveu A' - ' I 8 ' ' N No '.' "TE“" 5-1 > 3-. C t I 8.0 ...l u". .0 .. 5 Log um, 351* .:.':..‘:.'.°. was s . '- .. 1:00.... : R g 1.00- g .0: : E., on 0'90... 000..“ I. ...? ... 0L .. ' 41L mtimoazmmmmep.mapm m.) From: H.J. Bauer, "Case Study of Demand Responsive Transportation Systems", H.R.B. Special Report 124 Figure B.2.5. P-L Versus Ridership, Two-Thirds Capital Grant .3 s “DEM-IN t”! fl A It! FME 01.“ m & mos-tyros: PER DAV II Stan's O 0 WE: 54:3: “"1 ruldmummwmhadmm.) From: H.J. Bauer, "Case Study of Demand Responsive Transportation System", H.R.B. Special Report 124 Figure B.2.6. P—L Versus Ridership,Effect of Wage Rate 32 lrips/Vch1clclflour 124 10‘ Conditions: 3x3 mile area 2.5 mean level of ‘ ‘ service 2 en t I I T I 20 40 60 80 too Demands/Sq. m lelflr. ( Figaro 2. Effect of demand level on vehicle productivity.) From: Daniel Roos, "Dial-A-Bus Feasibility", H.R.B. Special Report 124 Figure B.2.7. Effect of Demand on Vehicle Productivity 33 Trips/Vehtcle/Hour l2 .1 x 40 Demnds/Sq. wile/hr. X . x to a f / 25 Demands/SQ. Ills/hr. Condirlons: 2.5 mean level of service l V— I 1 l to 20 30 40 so MI (Square Niles) (Figaro 3. Ethct of am use on vehicle productivity) From: Daniel Roos, "Dial-A-Bus Feasibility", H.R.B. Special Report 124 Figure B.2.8. Effect of Area Size on Vehicle Productivity 34 km: or mum Countess: 25 mum/sq. oil-ll- . 100- an- ‘°" mm. ‘0‘ nor».- 203 l l r I . T ‘ l.0 ' 2.0 3.0 0.0 5.0 0.0 "can level of Service (Flawed. ammmwummworm) From: Daniel Roos, "Dial—A-Bus Feasibility", H.R.B. Special Report 124 Figure B.2.9. Relationship Between Number of Vehicles and Level of Service 35 Glgurc l. Semitlvity to chooguin link travel Mme.) an m V“ . g. 0 . 0 . '3 3' " tum tow . {A 0 . o . 0. (no! 0 -vI-O M r» o 0 .7." ' 3:3: (0' on . no- 00 ”.0 m a u 0 s: n .r" Us" . '0 u us- R a... «0:0 ' . I. 0. '0 8. .0 0.0-0 w mm 0' 000. $090" 3. Smltlvity to changes in shops of area.) . 'fl‘. ti 'I i . a 0 !' I . o I 0 ' :1 ° § 00 00 00 '0 I. 00 .0 meow- Now ‘0 “0" (Flynn. 2. Sensitivity to changes in mal- mum pickup time.) .0 u" r. .0 M" /'—; v v v f 0 '00 MSG. ! * 0". “I. .00- , O.“ ‘9 color-u- mm .. fliv ;A v :v4 v i I I. .3.- ..(IIKV U (“H (ram 4. Sensitivity to changes ln am) From: Bruggenman, J .M. & Heathington, R.W., "Sensitivity to Various Parameters of a Danand Scheduled Bus System Computer Simulation Model," H.R.R. 251 Figure B.2.10 Sensitivity Data of Northwestern Study 36 0 on . a. III : . o ' 0 0 00 Mt I. o I . o - 3 t i 8 0 00 M (U. I a“. o .o. an! o 00“ I" I . .6 (I. 0‘ o 0.“ 00 O o no um: . 0 “(.00 “I!“ V. 0 . o . o 0 0 ' I. i o . 0 I no . . 3 ' ° . 0 0 Cl C, a. m ‘0‘” Q0 m (9810! figure 6. 'Smitivit'y to reduction in short trip.) '” tom not g u s 0 0 00 g team not i ' 2 U 9 8 2 u 0' .Ish .01 .0.“ II .nul hr Wcslb~ 0. o emu .- 0 II“ to ... m V. .- ‘5 l 0 0 9 0 ° 0 s e 0. 1‘ ° 4 i 030 ' o o . .- 0 o I 0 I 0 — MCM' ( Figure 5. Somitlvity to bu enmity) From: Bruggenman, J.M., Heathington, R.W., "Sensitivity to Various Parameters of a Demand Scheduled Bus System Computer Simula- tion Mbdel", H.R.R. 251 Figure B.2.ll. Sensitivity Data of Northwestern Study 37 40 , 34 - \ 3 5 L- r- to 30 - ¥ :0 A. v ' L05 2 6 _ L08 ' 20 J 2.5 I- 2" 30—/ 2-2 " \ -——/ 40 2.0 - 9 __ .0 nu "° 0.}: L00 J, L: r 50:! r noouctmrv-ss J 1 . ' 9° 11 0 1 1 1 1 1 0 1 1 1 1 I 2 3 4 5 .4 .8 l2 l6 2 O muse HI) AREA (50 HILES) (Figure 5. 3) (Figure 5.4) ”FLY CURVES M Morn SECTOR SIZE 5 1- OPTIWM SECTOR SIZE vs 30 1- nos-Jam" ‘L ormuo orusm a 2.5 b 3 :- A'EA L05 \ men. 20_ 2} L 1.5 1- up —L'° '0)?- --"L"OWN. ‘ 0 1 L l l 01 1 1 1 -1 1 ' 20 40 60 00 000 20 4O 60 00 000 I20 P Puma. P SGML (Figure 5.5) (Figure 5. 6) From: Mason, F.J. & Mumford, J .R., "Computer Models for Designing Dial-A-Ride Systans", Automotive Engineering Congress, Detroit Michigan, January 1972. Figure 3.2.12. Results of Ford Motor Company Study 38 ACHIEVABLE LEVEL OF SERV.ICE"VS VEH8CLE CAPACITY 3.0 *- . PRODUCTIVITY - 15 1 \ L'IOMIN. 2.5 .. \ p -ao {so 20 | \4 __————__————p020 1 \ ‘ 2.2 ’- \ ‘1 \ 40 L00 L05 ' “ \‘~ ————— — J ------ -J;- --------------------so l.8'- \ ‘ L4 ‘I 2 4 ' .6 8 IO 12 :4 l6 VEHICLE CAPACITY (Figure 5.7) From: Mason, F.J. 8 Mumford, J.R., "Computer Models for Designing Dial-A-Ride Systems", Automotive Engineering Congress, Detroit Michigan, January 1972. Figure 8.2.13. Results of Ford Motor Company Study 3.3. EXPERIENCES OF IMPLEMENTATION STUDIES OF DEMAND ACTUATED BUS SYSTEMS Several demand responsive systems have been implemented across the country for study and experiment during the past few‘years. An attempt is made to consolidate the most important experiences of such test programs in this chapter which are most pertinent to the proposed research. Vehicle Size and Type The vehicle size used in the simulation of various demand sched- uled bus systems varies from 30 to 20 seats. The Mprthwestern Uni- versity's simulation.mode1 considered vehicle size with 3 to 8 seat capacities. WABCO model used 5 to 20 passenger capacity vehicles for simulation. The MIT CARS project considered 6 to 10 seats per bus. The GMRL simulation package did not use any capacity constraint; rather they were attempting to study the reasonable number of seats per bus necessary to provide various levels of service. The vehicles used in the demonstration and experimental projects across the country also varied from 10 passenger vehicles to 53 pass- enger big coaches. Table 3.3.1 shows the various types of vehicles, types of service and implementing authorities associated with some of the experimental projects in the spectrum of demand actuated bus systems. Various types and sizes of vehicles have been used to date for DAB systems across the country. However, they can be summarized in three basic categories: 39 40 1. Van type units with 10 to 12 seats. 2. Mini buses with 19 to 25 seats. 3. Standard buses with 40 to 53 seats. The vehicle productivity data of the implemented systems tend to support small size buses with greater maneuverable capability rather than standard or large coaches. Type of Service The experience of DAB to date indicates that the type of service between experiments vary, though all of them can be defined as Demand Responsive Systems. The Mansfield, Columbus and Emmen projects used a route deviation system though each varied in criteria. Toronto started with many to one service and then added many to many service. Ann Arbor and Reginia basically provided many to few service. Columbia and Batavia provided many to many service with premium service during rush hours with subscription and predetermined rider demand. The variety in service over time in each system seems to be based on the subjective analysis of the Operators and the system designers in absence of any rigorous method for pre-evaluation of optimality of the type of service. Service Area The service area populations vary from 10,000 in Ann Arbor to 55,000 in Columbus. 41 The area coverage also varies from 1.34 square miles in Bay Ridges, Ontario to 6.5 square miles in Haddonfield, New Jersey. DispatchinggTechniques The dispatching activities performed to date in all implementa- tion projects are manual. The Mansfield project used the decision of the driver for dispatching, where as all other projects utilized a central control center and manual dispatchers. Most systems concluded that the manual dispatching and voice communication are satisfactory. However, one of the major limitations of such dispatching and communication systems is the area coverage and population. Nearly all the reports and studies concluded that manual dis- patching, though not Optimal, proved to be cost effective for the systems tested . Batavia, New York, recently conducted a study on a digital com- munication system. It also concluded that digital communication be- comes cost effective only when the system is sufficiently large to require more than 12 to 15 buses. Fare Structure There is considerable variability in the fare used and tested in different projects. These seem to depend on, local labor rates for drivers and dispatchers, the amount of subsidy and the basic goal of the project. The fare varies from 20¢ in Columbus, Ohio to 50¢ in Haddonfield, New’Jersey. 42 The simulation studies performed at M.I.T. and G.M. Research Tech. Center indicated that a fare of $.50 to $1.25 is required to cover all fixed and operating costs of a demand responsive system. Ridership The ridership in all the implemented projects increased during the study period. However, no quantitative estimate of future ridership for improved efficiency and service have been established to date. Some DAB systems are extension of fixed route system whereas others are completely new. As such, the rate of increase of ridership in rela- tion to improved service is hard to establish from the study reports. The G.M.R.L. simulation study of a real community predicted a significant ridership increase. The demand estimation for improved levels of service have not yet been fully explored by the researchers, though simulation provides a viable tool for ridership estimation. Vehicle Productivity Vehicle productivity has been defined by all as the number of pass- engers per hour, per vehicle and has been treated as the most important system performance measure. However, this system attribute can be limited either by supply or by demand, or by both. Since no standard level of service have been maintained by all the, operating systems, it is hard to compare the various vehicle productivity figures presented by various studies and reports. The factors which either singly or in combination affect the vehicle productivity as obtained from the studies and reports seem to include: 43 1. Type of Service: The M.I.T. researchers and the analysts at Ford Motor Company who were involved with at least six major DAR projects seem to favor route deviation service as having a greater potential for higher vehicle productivity over many to many service. 2. Service Requests: Vehicle productivity increases with the in- crease Of prebooked or standing service requests. This provides for preplanning, and results in more efficient vehicle routings. The operator has a guaranteed ridership for which they can sched— ule buses before they receive any on-line calls for service. 3. DispatchingyEfficiency: Efficient dispatching appears to increase productivity, since more passengers can be handled by the same vehicle with no or little extra time to the passengers. 4. Demand Density: Since no two systems have equal levels of ser- vice and passenger and trip characteristics, comparison of the effect of demand on bus utilization is not possible from the reports. Cost of Ride The cost per person trip is a function of the total Operating and fixed costs and vehicle productivity. Thus, the local labor rates play a very important role in the cost of a ride. The cost varies from project to project. Regina, Saskatchewan re- ported cost of $0.54 per ride and Haddonfield, New Jersey reported $1.75 per ride. The labor rates in the above two projects vary considerably, and the vehicle productivity of Regina is 3 times that of Haddonfield. 44 Cost Functions There has been considerable effort expended in the past in relating the system attributes to dollar cost both in research and implementation. A comprehensive study was done at M.I.T. on the "Dial-A-Ride" sys- tem in defining and establishing the cost functions. The cost of a typi- cal DAR system was divided into four basic components: 1. Message handling 2. Message processing 3. Passenger transport 4. System support The primary objective on the research was to "develop a set of predictive cost models for a Dial-A-Ride system."1 The following cost models were the product of the above research. 1. Message Handling: A. Customer (voice) Cost per.($_. to $2.5)(Demand/Hr.)(Operator Wage Rate) demand (Operator Productivity) B. Vehicle (voice): Cost perafi§l.8 to $2.1)(Demand/Hr.)(0utput Operator wage Rate) demand (Output Operator Productivity) 2. Message Processing: Cost per demand a function of system.sizex(2¢ to 20¢ per demand) 1Gary L. Urbanek - Cost Considerations for Dial-A-Ride. M.I.T. Systems Lab. publication, June 1971. 3. 45 Passenger Transport: Cost per hour-($1.45 to $1.55)x (Driver's wage rate) + ($1.30 to $2.71) per vehicle per hour. System Support: Cost-($.02 to $.05)x(tota1 cost of the other functions) 46 ooa>hom mvmamo muowoommmm noon Ou uooonoasomnom .omsmnoumxmmm Na huauonu=< OHHnsm OHAMHHM> Ousom Oanmfium> .mcwwom a muowaommmm OOHHosoom NH huauonus< OHHnam ovaI a: xowa wocwfimon mumwsmmmmm uoov ou uoovuoasvonom auamaumam AH auauosusa unanaa manmaum> musom masmanm> wz .uumamaouumm a powwow Hmfiooam muowammmmm mm pom: .Oadvonom uamwwsmno cameo 2o mm Asauosus< oafipsm amass .onaom woxam m.umaaonm anuuun o muowaommmm oaavonom onus: ca amaze Hmsca>aecH uaaaaaa .musom waxes Hz .Aan uuuamuu< m muowsommmm Ousom OHanum> mocwaoooum NH usonoaouusm Oum>wum -fiamm .mmzvmom Oanmwum> owno .vamfimmnmz a mom Hamam .usOeHuonxm you .VOHuom mnomaommmm vomamsu .Oaavmnom man as“: ma suauonus< uaansa waxes .mssom waxes «a .manmmoamz m snowsommmm zosow< cowuoaoum nmouaxmx sumac zo mm a me an“; auauonua< unapsm = nauseous .uauflm u muowsommmm .>wsa no“? sowuocshsoo Ousom owamcha .Onavonom mfioaHHHH nomou so me su.oo uamsmua Oum>aum voxwm .soaumauomnsm uaumooa w muuoom H OHOHSO> mo muaomamo OOH>uOm usoaauonxm .oz Oxmz a waxy nowaommmm soaumucoauaaaH mo make no sowumooa cowumahowsH soaumuaoamaaaH mam vmumsuu< panama H.m.m OHAOH 47 smsonoumxmwm o.ma mo.o~ ma.~ na.o~ coma .maammm v homuon.3mz o.~ mu.~ m.a .mu; «N mam .uaoamaouumm oano n.m um.~H n.~ m.mH mwc .mnpssaoo vsmnhumz q.q N.H 0.0 m.m on .owne=Hoo xuow 302 ~.oH wa.~ nn.¢ NH Mme .Ofi>nuum owumuno m.HH w~.a~ em.H om mos .momvam mum nuances: no.5 HH.nH cm.H NH «Ha .uonu4 =s< man you nfiamuovwm muw>auosuoum mou< nouumuoao mo vaoxoos mafia umom anuno> .um\oafizn.um\osmson Ooa>uom mason Houoa Hmowama mo Osmz muuofioum ovam1auaaan a co aaanoaoauuaom mua>auoauoua uauaun .~.m.n «Home 3.4. CONCLUSIONS OF THE STATE OF THE ART An overview of the state of the art indicates that there has been research, demonstration and implementation studies made on demand actuated bus systems throughout North America and abroad. These stud- ies range from very simple application to sOphisticated computer simu- lation studies. However, all of them are oriented towards the case study approach, and do not treat a family of Demand Actuated Bus sys- tems on a single set of parameters. Thus, it is not possible to define a set of parameter values that would determine the applicability of the various systems. Though simulation has been used, most of the models developed are for a specific level of system constraints rather than a family of systems. It is pertinent to briefly review at this stage the questions: A. What do we know about DAB systems from the State of Art? B. What are the aspects of DAB systems still unknown to date which require additional study? The following conclusions can be made from the experience of DAB demonstration and implementation studies. 1. Several factors are apparently of great importance to the success of a DAB systems; with service area, service selection and dispatching strategies the most important. 2. DAB systens have not been self supporting or money making systens. Public subsidy is necessary to support a bus systen with this higher level Of service. Ridership increase due to implementation 48 49 of DAB systems compared to fixed route fixed headway systems is significant. 3. Public subsidy is not a guarantee for system success. This requires a competent system analyst and management and staff. 4. Significant operating economies can be achieved by using small coaches which increase the occupancy-capacity ratio. 5. Increased vehicle productivity (no. of passengers per hour per vehicle) is possible with DAB systems. 6. Off peak fare reductions policy results in redistribution of ridership--i.e., peak smoothing-~but does not affect total ridership or revenue. 7. Ridership seems to have limited sensitivity to fares between 25¢ and 60¢. 8. Manual dispatching works fairly well for smaller service areas with few buses. For larger areas computer routing and dispatch- ing is required. 9. Careful analysis of service area may allow a system to be efficiently Operated under many to one or many to few environments. 10. For high density (riders) areas, fixed time fixed route sys- tems work as well as any DAB system. The following conclusions can be made from the research and development studies: 1. System simulation provides a tool for system evaluation and selection. 50 2. The concept of DAB systems is technically feasible. 3. Financially the systems cannot be self supporting. 4. Demand estimation analysis shall be a part of system study prior to implementation for selecting viable service criteria. 5. Heuristic Traveling Salesman algorithms works for optimal routing in most of the existing simulation models. 6. Computer dispatching is economical for large systems. 7. The limited sensitivity analysis presented in various studies and papers indicate sensitivity of system performance to waiting time, trip time, and link travel time for specific systems only. 8. Mbst simulation studies are for specific systems, with limited sensitivity analysis to performance as a function of levels of input parameters like, demand, link travel time, etc.. As such, no comparison can be made between various system alternatives. The following questions regarding demand actuated bus system still remain unanswered: A. Can a single simulation model replicate the Operations of variOus types of demand actuated bus systems, which will generate system performance data for use in the alternative system evaluation and selection process? B. 1. Do the variables used in demand estimation and economic feasibility tests (e.g. wait time, ride time, driver hours required for service, vehicle hours Of operation, bus utilization, etc.) vary significantly with demand for various DAB systems? and; 51 2. Can the parametric information produced by a general purpose bus system models be used for any organized system selec- tion procedure? C. Do variations in system parameters such as service area, bus pooling and street configuration cause major changes in the value of the Operating variables? C. PROBLEM STATEMENT AND OBJECTIVE OF THE RESEARCH The concept of publicly owned multi-passenger vehicles circul— ating through our residential areas, responding to telephone requests for picking up or delivering riders to and from their door step is certainly an attractive one. This enables the riders to enjoy per- sonalized service without the cares of vehicle maintenance and storage. However, the search of the literature indicates that insuffi- cient information on the cost-effectiveness of various systems exist. The simulated studies to date, generally worked with specified levels of service Options and particularly do not provide any means of eval- uating various system alternatives. A generalized study approach demonstrating a viable system selection strategy of various discrete system alternatives for a wide range of demand is needed for deci- sion makers for optimal system selection. The objective of this research was to develOp such procedure which can be used to determine a set of parameters which provide necessary information to define the cost-effective regime for demand actuated bus system alternatives. To accomplish this objective the research included the follow- 1. Development of system definitions of several discrete DAB systems. 2. Development of a generalized model, which will simulate the operation of all systems on a common data base. 52 53 3. Determination of system attribute functions for each system as a function of: a) ridership levels b) roadway network configuration c) pooling policy d) service area 4. Demonstration of a method for constructing cost functions for evaluation of alternative system strategies. The research was limited to "many to one" and "one to many" en- vironments. However, the simulation model has been designed such that with minimal change, it can be used to simulate "many to few" or "many to many" environnents. The basic objective of this research was to develop a decision making tool (generalized simulation.mode1) which can be used to estab- lish trade-Off functions to determine the bounds of demand and user characteristics for different systems. If system attributes under various demand conditions are reduced to cost functions, it is ex- pected that the viability of the various system alternatives will change with the demand. Figure 0.1 indicates a hypothetical situa- tion where cost limits the use of various system alternatives for in- creased level Of ridership demand. Evaluations of this nature re- quires a system performance data and a general purpose simulation model which can replicate various operational alternatives and service options provides such information. 54 < System "D" ( System "C" Cost "A" Optimal "B" Optimal "C" Optimal "D" Optimal 4-}. ~ A ~ ‘ ’- Demand (Riders/Hr./Sq. Mile) Figure C.1 Optimal Regions for DAB Systems Alternatives D. METHODOLOGY D.1 SYSTEMS DEFINITION To initiate the study, it was necessary to define the character- istics of the various systems to be tested and evaluated. Table D.1.1 is a list of public transportation systems which belong to the family of Demand Actuated Bus Systems. The dispatching cri- teria to meet the specified requirements of system performance para- meters are also shown in the table. A brief description of the Operating characteristics of some of the systems under investigation are as follows: 1. Fixed headway - fixed route bus system. This system is based on multiple passenger vehicles traveling on pre-determined routes at pre-scheduled headways for passenger pick up and delivery. The route and headways for such Operations are pre—determined from past demand experiences. This type of system now exists in.most urban areas. 2. Variable headway - fixed route bus system. This system is based on the concept of riders being able to register their demand by calling the central dispatcher. The dispatcher notes the location and time of such demands, and depending on the dispatching criteria, vehicles are sent out for servicing such demand. There can be defined a host of bus systems under this category on the basis of the dispatching logic. The routes of systems under this category are also fixed (pre-determined) though headways are determined from the demand register. 55 56 Table D.1.1 Performance Criteria Levels in the Hierarchy of DAB Systems System Dispatch Response Priority Priority of Route No. _ngic to Demand of Pickeup delivery Predetermined Non Constant Function of 1 (fixed) route res onsive for all position on Fixed and fixed head- p position route way (general urpose) Variable Semi Constant Function of 2 headway responsive for position on Fixed fixed to fixed each route route nodgfiii position Fixed headway Semi Function Function of Traveling 3 variable route responsive of posi— demand posi— salesman to to fixed tion rela- tion fixed nodes points tive to other details Fixed Semi Function Function of 4 headway responsive of position demand Traveling variable route to relative to position salesman to address address other calls Variable route Responsive Guaranteed Function of 5 variable headway to within demand Traveling (non-dynamic) fixed specified position salesman to figed godes points interval Variable route Fully Guaranteed Function of Travelin 6 variable headway responsive within demand g (non-dynamic) to specified position salesman door-to-door address time Variable route Fully Guaranteed Guaranteed 7 variable responsive within within Traveling headway to specified specified salemsan (dynamic) yggdress time 9392, Fully 3 Taxi responsive type to First First Minimum address path 57 3. W This system is based on the premise that a bus is dispatched to service riders in any area on a fixed schedule. An Optimal routing technique is necessary for system Operation. To increase the efficiency in computing the optimal rout- ing strategies, high speed digital computers are necessary. In this system the riders call the central dispatcher to register their demand. The locations of the demands are then analyzed for Optimal routing for a given set of buses and pre-determined headways. The task performed at the dispatch center is the determination of op- timal routes of vehicles and assignment of passengers to vehicles. 4. Variable headway - variable route (non-dynamic). This system is based on the concept that riders call the dispatching center and register the demand. The demands are recorded and analyzed according to their time of calls and spatial locations for determining the op- timal routes according to the dispatching criteria. The dispatching criteria can be functionally related to costs and level of service to the users. To accomplish this, the total demand and socio—economic characteristics of the users, must be known. Thus, the determination of headways does not pose a great problem and in most cases depends on the environment where the system is operating. However, the problem of selecting routes requires an optimal routing algorithm as in case 3. The methodology employed in this study was to develop a simula- tion model which can simulate each of these systems under various 58 environmental (socio-economic characteristics - Demand) conditions to produce system performance data necessary to provide data for selection of alternative bus systems. D.2.a. SIMULATION The basic tool for this research is a simulation model develOped as a part of the study to generate system performance and cost func- tions. The simulation model is flexible enough to replicate all the systems as defined here with.various levels of operational alternatives. Characteristics of Simulation The variability and probabilistic nature of the variables in- volved in demand actuated transportation problems has made analytical solutions complex. Past transportation system studies and evaluations have tended to approach the problems by making simplifying assumptions or by testing only one system under given conditions. The evaluation process involved in this system simulation ap- proach provides a viable alternative for determining the trade-off functions between cost and benefit parameters over a wide range of alternative policy questions. The simulation technique adapted in this study allows generation of rider demand from random numbers for any type of probability dis- tribution. The time and location of passenger demands are independent random variables and as such utilization of random generators in the simulation provides an effective and realistic means of replication of real world situations. Simulation MOdel The simulation model was developed with the capability of rep- licating the Operations of bus systems under various system strate- gies, and collecting the data required for statistical summaries of 59 60 system performance parameters (refer Appendix II for flow charts). The model is initiated by generating strings of demands (collec— tion and distribution) for each specific area of Operation. There is no limit with regard to the size of an area. For larger areas, the model can handle sectoring or splitting of the entire area in seg- ments, and can Operate the simulated bus systems concurrently in each segment. The model assumes a central pool of buses from which vehicles are dispatched to any sector on demand. For fixed route systems, the routes of bus Operation must be specified. For variable route systems, the shortest path will be formed and used by the model. The simulation program develops internally a point to point travel distance matrix from the given size of the area under investi— gation and density of demand generation points. The distance matrix is then converted to a travel time matrix for given speed. The speed of travel can either be used as an average speed or as a function of other parameters. The simulation.mode1 consists of two separate but interconnected programs. The first program handles the systems which are operated on the basis of separate distribution and collection. And the second program handles the alternative operational procedure of concurrent collection and distribution. The computer models have a block data part which specifies the values of the parameters, number of sectors to be serviced, fixed 61 headways, dispatch logic, vehicle capacity, and number of buses. This part is the same for both the programs. The next part of the model is the main program which does the initializing of the various parameters, calls all different subroutines in sequence and executes demand. The execution is event ordered and the output statistics (waiting time, riding time, total travel time, etc.) are accumulated. Subroutine The "generate" subroutine generates rider demands on the basis of input, average arrival rate and type of demand function. The door to door or node to node travel distances are formed from the given dimensions of an area of service and the criteria set in the system operation. The "BS Pool" subroutine performs the function of servicing the demands using the number of buses assigned to it by the block data. When a bus completes its tour it is returned to the pool and at this time, the subroutine "BSSTAT" is called to execute the summary statis— tics. Besides this summary, statistical output is printed for the individual bus tour at the completion of its tour. The "STATIS" subroutine (also referred to as DSTAT, CSTAT AND QUTSTS) calculates pertinent statistical information on passenger wait, ride and total travel time. The subroutine prints out the in- formation at the end of the simulation period. The functions SMEAN, HIGH, NOVER and the subroutine SORT are used by the subroutine STATIS (DSAT, CSTAT) in the calculation of passenger statistics. 62 BSSTC (BSSTAT) subroutine performs the same function as the STATIS subroutine, except it calculates pertinent statistics for each bus in the bus pool. The "RDList" subroutine is used to read into the system all in- formation to be used by the program in modeling. This includes the "Time-Travel" matrix and the demand string. The demand string is separated in two manners: First, by sector and secondly, whether the demand is for distribution or collection. For convenience, the "RDList" subroutine prints out all data by sector and by demand type before any simulation takes place. The "DList" subroutine, chooses which sector to execute next. The sector with a demand string to be executed that has the lowest absolute clock time associated with it is chosen. "SLICE" (also referred to as SLICE 2) slices the distribution demands on the basis of capacity and headways into buses and the dispatch logic of the sys- tem into executable sub lists. It also calculates the absolute clock time that bus will be required in a particular sector. "C" List" performs the same operation as "D" List" on the col- lection demands. Subroutine "DISTR" produces the best tour on route, to minimize passenger travel time of the string of executable demands by "SLICE". Should there be demands of the sub-string to be executed after the best tour has been constructed the "DISTR" subroutine, returns them to their respective lists. 63 "CLICT'I performs the same operation as "DISTR" on the collection demands. "TTIME" is used by "DISTR" and "CLICT" to construct a miniature Time-Travel matrix. Routing MOdule The algorithm is an heuristic solution of the classical "travel- ing salesman problem" under many to one or one to many environments. This module computes the optimal routes for given demand and their location. In the evaluation of policy questions regarding service options, fleet size, vehicle capacity, etc., different variations of the pro- gram can be constructed by setting pertinent input parameters. Service Options - Since there are the possible triggers in the diapatch logic, the fixed headway operation can be achieved by set- ting the demand parameter high enough to trigger a vehicle dispatch on the basis of predetermined headway. The variable headway operation can be achieved by setting the time parameter high. The concurrent or separate collection and distribution Operation can be achieved by specifying the input data. The simulation model requires the following inputs: Fleet size - The number of vehicles available in the bus pool must be specified in the input data. The simulation program is set up to use vehicles from the top of the list. Thus a high number of buses used in any system simulation can answer questions regarding optimal fleet sizes. 64 Vehicle capacity - The capacity of vehicles used in the simula- tion is also an input parameter and must be specified. This capacity is treated as a constraint in making the decision of vehicle to pas- senger service. Activity Center - The zone number for identification of activity center is also an input parameter. Thus changing the location of an activity center for various trip purposes is possible. Demand - The mean arrival rate for the area under study must be an input. The generation subroutine computes the rider demand on the basis of the pre-assigned probability distribution and random numbers. The frequency of collection and distribution demand is also computed according to the probability function and its parameters. Thus, the generation of rider demand is completely random (poisson's or any other probability distribution) and is computed for any specified length of time in the model. The demand locations are identified as points in the model, and as such can be specific bus stops or individual homes. Sector size and number of blocks - The program generates poss- ible demand points (door steps or nodes) from specified length and breadth of each sector and the number of blocks in each direction. A travel distance matrix and finally a point to point travel time matrix is generated for given speed of travel or Speed function. '.legy_- The delay at any stop for a bus is a function of the number of passengers involved and whether they are loading or 65 unloading. The model includes separate delay functions for evalua- tion of travel time. For a given bus stop the delay is computed on the basis of the number of passengers entering and leaving at that location. This function is set up as a default option in the absence of specific input. Dispatch 1ggig_- The dispatching criteria are input to the model. Dispatching logic may signify a certain type of bus system as outlined in Table D.1.1. The model requires specific inputs regarding minimum number of passenger demand to warrant dispatching a bus or maximum headway of buses if the specified demand is not registered. These, of course, have to be pre-determined in the system definition part of any problem. Service Constraints The maximum tolerable waiting time for a passenger is also an input to this model. This parameter signifies to some extent the level of service provided to the riders. It also is used to define the limit at which potential riders switch to other means of transportation. The Assumptions and Operation of the Simulation Mbdel 1. Demand is considered as call for service and can consist of either single or multiple passenger. 2. Number of passenger in a demand is generated from given pro- bability function with assigned probabilities for 1 passenger, 2 pas- sengers, etc. 66 3. The available buses in the ”Bus Pool" are serially numbered and the simulation model always searches for lowest numbered buses for use, and will not call for a new bus (though available from the system constraints) as long as there is a bus in the "Bus Pool" which have been used before. Thus, maximum utilization of each bus is accomplished for given number of buses. 4. The travel times between all possible generation points are determined from the computed distances and given average travel speed or speed function. The simulation results presented here are based on average travel speed of 20 miles per hour. However, the model can handle individual speed functions for specified links of travel. Model Operation 1. The "Generate" subroutine generates passenger demands (col- lection and distribution) with necessary identification information like origin zone no., distination zone no., number of passengers in each demand, absolute clock time of generation, etc. 2. The passenger demand is then separated by sectors and arranged in increasing order of generation time. 3. The model then looks at the demand lists of each sector at specified interval of time (5 sec., 10 sec., 15 sec., 20 sec., 30 sec., etc.) and tests with the maximum headway constraint and dis- patch logic. If none of the above two dispatch trigger is satisfied, the model moves to the next increment of time and tests again to in— vestigate whether or not dispatch criteria is met. This operation is done concurrently in all sectors with incrementing the time clock. 67 As soon a dispatch criteria is satisfied, a bus is dispatched with selected demand list into the specific sector for service. 4. The "Bus Pool" subroutine receives message for bus dispatch- ing and searches for a bus with the lowest serial number and sends it to the warranted destinations. 5. The time of bus dispatch is noted and a time clock is ad- vanced in each bus to keep record of tour time. This time clock con— siders point to point travel time from the generated travel time mat- rix and also accumulates embarkation and disembarkation times at each service point. This data is also accumulated to generate bus tour time, percent of time buses in use, number of distribution and collection passengers served. 6. Another module keeps track of individual passenger genera- tion time, time when picked up by a bus, length (time) of travel, time of reaching at destinations, etc. Accumulates waiting times, riding times, total travel times, etc. This module summarizes pas- senger service data after each bus tour and also compiles summary statistics. The summary statistics also segregates passenger service information hour by hour basis. Outout of the Simulation Model The output of the model consists of the following system per- formance parameters: 1. User Statistics - The simulation of system is performed by the model and all pertinent statistics of the riders are accumulated. 68 The waiting time at the point of demand, riding time, and total travel time are accumulated by the model and printed for each bus trip. The origin and destination of the riders, their time of gen— eration, waiting time, riding time and total travel time are printed for each bus trip and also for each segment of the service area (refer to Appendix I). The summary statistics tables consist of "wait time", "ride time" and "total travel time". wait time is the time difference between the absolute clock time of rider demand generation (time of demand request in the real world situation) and the time when the rider is picked up by a bus. If any passenger has to skip a bus due to capacity constraints, the time of wait until the next bus picks the rider up is also added to the wait time. Individual wait time of riders are accumulated to determine the distribution of waiting time. Ride time is the time difference between a rider boarding time and drop off time. This time includes bus running time and embarka- tion and disembarkation delay on route. Total travel time is the time from the demand generation to the end of the demand service. Thus, total travel time = wait time + ride time. These tables are printed for specified time segments and for all sectors (refer to Appen. I). The mean, maximum and frequency over a specified limit is accumulated and printed for all of the x _ 69 above summary statistics. This enables the analyst to take a closer look at the system attributes. 2. Bus Statistics - The bus occupancy, the trip time and the bus utilization are accumulated for the entire simulation period. The first set of output consists of a table for each vehicle indi- cating each tour, the number of passengers collected, distributed and total tour time. The summary statistics for the buses consist of one table for each time segment indicating the number of tours performed by each bus, number of passengers collected and distributed, total tour time, mean tour time and bus utilization. Bus utilization is the percent of time a bus was on the road within the entire period of simulation. Vehicle productivity is the average number of passengers serviced per vehicle per unit time. D.2.b. THE TEST SYSTEM The service area tested in this research composed of 3 sectors with a common activity center. Each sector is one mile square and contains 225 passenger generation points. The roadway pattern assumed to be grid with a radial road in each sector (figure D.Z.b.l). The variable route systems assumes access to each generation points through the grid network. The fixed route systems follow the route shown in figure D.2.b.l. It is important to mention here, that the fixed route plan used in this study provides a higher level of area coverage than what can be found in real urban environment. The buses follow this fixed pattern for all of their tours and make a total of 20 fixed stops to pick up or drop Off passengers. The gen- erated riders are assumed to walk to their nearest stops for service. The walking time is included in the waiting time of the individual passengers. The passengers are generated at each of the 225 generation points/ sector using a random generator. The distribution of passenger ar- rivals at each point is considered to be poisson, with the mean value being one of the variable tested. The probability of generating a collection demand is used as 0.5. Where the collection demand re- presents the riders who need to be picked up at their point of genera— tion and dropped Off at the activity center. The distribution demands are the passengers who are picked up at the activity center and drOpped off at the various destination points within each sector. 70 71 Sector #3 Sector #2 :1 Route (typical) Typical Node Numbe L_ 1 Typical Rider ‘////‘ Generation Point Sector #1 \----_--—-- Figure D.2.b.l Basic Network Configurations 72 Prior to an analysis of the differences in system performance characteristics produced by the models, validation of the model re- sults is essential. To test this, a comparison was made between the model results over a range of demand densities and actual field ex- periences, as shown in figure D.2.b.2. As expected, the existing systems perform better than fixed time - fixed route systems considered in the' ‘ study but not as good as an optimally routed Demand Actuated System. This figure also indicates that the advantages of demand actuation decreases with increasing demand as would be expected. The results of this study were also compared to the results of the M.I.T. studyl. There are several differences in the basic con- ditions assumed in the two studies. The service area in the M.I.T. studywas 9 square miles compared to total of 3 square miles for the study. A network expansion coefficient Of 1.8 was used to convert the service area of this study into an area of 3.24 square miles per sector. The mean level Of service of the M.I.T. study was set at 2.5. In this study, level of service is an output and thus not controllable. The resultant figure (LOS - 2.26) represents a better service. Figure D.2.b.3. illustrates the results of one of the most im- portant measures of system performance, system productivity, as a function of demand. Productivity, expressed in trips per hour per lRoos, Daniel, "Dial-ArBus Feasibility", H.R.B. Special Report 124. 73 vehicle, will dictate the economics of the operation. Since the simulation used in this study is non-dynamic, Operating expenses will be less even though the service level was higher than that resulting from the M.I.T. study. Some further observations can be made from this comparison: I. The current simulation is more efficient when measured by vehicle productivity. 2. The increase in productivity of both systems levels off at a demand of 60-80/sq. mile. 3. The non-dynamic simulation in this study produces higher productivity than the dynamic system since the bus waits until a cer- tain number of demands are registered. This eliminates Operation during those periods of low demand. Thus, the high values of the productivity from the current study is quite reasonable. However, wait time in a dynamic system is always lower than in the nonrdynamic operation. The differences in productivity for equal levels of demand can be attributed to this difference in operational strategy. The points considered are not sufficient to conclude that the current product- ivity function will remain higher at all levels of demand. 74 Table D.2.b.l Simulated Data for Variable Route-Variable Headway Systems System Attributes Demandlhr./sector 10 20 30 40 60 80 100 Mean Total Travel Time by Bus 15.59 16.62 14.9 13.94 15.46 15.24 17.33 Mean Ride Time By Bus 6.7 7.67 7.75 7.69 7.96 8.56 9.92 Delay Due To Embarkation And Disembarkation 0.5 0.67 1.05 1.19 1.06 1.26 1.92 A-p. Travel Time By Auto 6.2 7.0 6.70 6.50 6.90 7.30 8.0 Level of Service ' TTT byyBus TT by Auto 2.51 2.4 2.22 2.15 2.26 2.1 2.17 Vehicle Productivity 11.7 19.33 21.08 20.9 27.21. 27.7 26.9 75 Fixed Route Fixed Headway System Hand Fitted Productivity 30 ._ Functions-Data From Simulated Bus Operation by Computer Model Hr.) A —__ C] 25 1- ' Variable Route Variable z .A 3 [5 Headway System ‘ 20 -~ C] “a”. A .... 0 L4 5. Hand Fitted Productivity Function a For 7 Dial-A-Ride Projects Regina, Saskatchewan h m o ’ Bay Ridges, Ontario Batavia, New York Columbus, Ohio Ann Arbor, Michigan Vehicle Productivity 9 Columbia, Maryland Haddonfield, New Jersey “:::::+::+ 1o 20 3o 40 so 60 7o 80 90 100 110 Demand Density (Trips/MiZ/Hr.) Figure D.2.b.2. Comparison of Simulated Productivity with Operating Systems 28A 261 2m 22. 20 . 1: :i O :1: \ 3: 18' U -.--1 .3: m > \ 3' 16. H E: >. u ‘; 11.. :H u U 5 'U 0 a 12. a) o-G U H .1: O) > 10‘ 84 6 . p 76 Variable Route Variable Headway System Hand Fitted Curve Conditions - 3 sectors @ l x 1 mile area. Level of service: Mean = 2.26 Std. DIV. a 0.15 Conditions - 3 sectors @ 3.24 sq. mile area each. Level of service = 2.2 app. Simulated Dial—A—Bus productivity function, presented by Daniel Roos (Dial—A—Bus Feasibility-H.R.B Special Report-124) Conditions - 3 x 3 mile area. 2.5 mean level of service. Demands/Sq.Mile/Hr. Figure D.2.b.3. Comparison of Simulated Results and M.I.T. Study E. RESULTS E.l. SYSTEM PERFORMANCE CHARACTERISTICS Each of the four systems identified in the previous section were used as the basis for simulation on a fixed data set represent- ing three hours of operation at each of several levels of demand. The demand levels tested in the analysis were: 10, 20, 30, 40, 60, 80 and 100 demands per hour, per sector. Summary statistics, both user and operator, were collected and compared as a method of establishing system performance. These statistics (refer Tables E.1.l. to E.1.4.) include waiting time, riding time, number of vehicles required and vehicle productivity. In Figure E.1.l the average wait time per passenger is plotted against the demand per hour/sector for each system. The variable headway-variable route system provides the lowest wait time for a demand level below 75 per hour per sector. Beyond this level, the fixed route variable headway has the lowest waiting time, although the variable headway variable route system is quite close. At demand levels beyond 60/hr./sector the fixed route fixed headway system has almost the same wait time characteristics as that of the variable route variable headway system. The variable route fixed headway system is clearly not comparable as it results in a much higher waiting time than any other system. Figure E.1.2 illustrates the ride time characteristics of each system. The variable route variable headway system can reduce riding time in the network for all levels of demand tested. Both the fixed route systems yielded very high ride time, as one might expect. 77 78 Table E.1.l System Performance Data - Fixed Route Fixed Headway SYSTEM: OPERATION STRATEGY: Fixed Route Fixed Headway Concurrent collection and Distribution SYSTEM Demand/Hr/Sector ATTRIBUTES 10 20 3O 40 60 80 100 No. of Sectors 3 3 3 3 3 3 3 Total Area _(sq;7mile§)_ 3 3 3 3 3 3 3 Simulation Period (hrs) 3 3 3 3 3 3 3 Netw°rk C°“’ Grid Grid Grid Grid ‘ Grid Grid Grid figuration Dispatch Logic (No. of 35 4O 4O 4O 4O 4O 40 em!) Headway (pins) 15 15 15 15 10 10 10 T°tal Passe“‘* 91 198 318 374 613 827 1168 _ggrs Served Total No. of Buses Reqd. 8 9 9 9 13 14 15 Mean Waiting Time (mins) 10.98 9.84 9.86 10.34 7.6 7.42 7.77 Mean Ride Time 16.8 18.00 18.50 19.31 18.98 19.56 20.17 Mean Total Travel Time 27.78 27.84 28.36 26.95 26.58 26.98 27.94 Total Driver Hrs. Reqd. 30 32 34 34 50 52 55 Vehicle Productivity 6 11.04 16.13 18.64 19.77 25.65 34.25 Mean Bus Utilization 512, 56% 582 592 62% 622 622 Total Vehicle 5:3; °f Ope‘a‘ 15.29 17.92 19.72 20.06 31.0 32.24 34.1 *Total number of passengers for 3 hours of simulation. 79 Table E.1.2. System Performance Data - Fixed Route Variable Headway SYSTEM: FIXED ROUTE VARIABLE HEADWAY OPERATION STRATEGY: CONCURRENT COLLECTION AND DISTRIBUTION SYSTEM Demand/Hr/Sector ATTRIBUTES 10 20 3O 4O 60 80 100 No. of Sectors 3 3 3 3 3 33, .3_ Total Area (sq. miles) 3 3 3 3 .3_, i3 .3 Simulation Period (hrs) 3 3 3 3 .3 .3 .3 Network Con- figuratiOn” Grid Grid Grid Grid Grid Grid, Grid Disptach Logic (No. of 3 5 5 5 5 5 5 pax) . . Headway (mins) 15 15 15 15 10 10 10 Total Passen- * _gers Served 91 199 320 388 628 845 1079 Total No. of Buses Reqd. 9 11 12 14 19 25 30 Mean waiting Time (mins) 8.6 8.74 8.68 7.49 5.4 5.65 4.26 Mean Ride Time (mins)y 14.68 18.18 17.96 18.75 18.5 17.89 17.67 Mean Total Trad vel Time (mins]23.28 26.92 26.64 26.24 23.9 23.54 21.93 Total Driver Hrs. Reqd. 3O 36 42 49 74 91 122 Vehicle Pro- ductivity (no. 4.8 5.5 7.6 7.9 8.5 9.3 9.63 ofgpax/hr/veh) "ea“ 3““ 632 547a 5874 562 622 63% 61: Utilization Total Vehicle Hrs. of Opera— 18.98 19.44 24.36 27.44 45.88 57.33 68.35 tion *Total number of passengers for 3 hours of simulation. 80 Table E.1.3 System Performance Data - Variable Route Fixed Headway SYSTEM: Variable Route Fixed Headway OPERATION STRATEGY: Concurrent Collection & Distribution ‘ SYSTEM Demand/Hr./Sector ATTRIBUTES 10 20 3O 4O 60 80 100 No. of Sectors 3 3 3 3 3 3 3 Total Area (sq. Miles) 3 3 3 3 3 3 - 3 Simulation Period (hrs) 33 3 3 3 3 3 3 Network Con- figuration Grid Grid Grid Grid Grid Grid Grid Dispatch Logic (no. of pax)‘ 35 20 20 20 40 4O 4O Headway (mins.) 15 15 15 15 10 10 10 Total Pass- * engers Served 91 208 309 459 684 867 1101 Total No. of Buses Reqd. 4 6 6 6 9 9 10 Mean Waiting Time (mins.) 12.23 12.87 13.41 13.48 11.52 13.2 16.02 Mean Ride Time (mins.) 8.17 8.02 9.38 10.89 12.48 13.13 13.63 Mean Total Travel Time (mins.) 20.4 20.89 22.79 24.37 24.00 26.34 29.65 Total Drive Hours Reqd. 14 20 21 23 33 34 40 Vehicle Pro- ductivity (no. of pax/ hr./veh.) 14.4 25.36 31.98 37.65 38.38 43.22 45.88 Mean Bus Utilization 45% 41% 46% 53% 54% 59% 66% Total Vehicle Hours of Operation 6.31 8.2 9.66 12.19 17.82 20.06 4 24 *Total number of passengers for 3 hours of simulation. 81 Table E.1.4 System Performance Data - Variable Route Variable Headway SYSTEM: Variable Route Variable Headway OPERATION STRATEGY: Concurrent Collection 8 Distribution SYSTEM Demand/Hr./Sector ATTRIBUTES 10 20 I ' 30 40 60 80 100 No. of Sectors '3 3 3 3 3 3 3 Total Area (sq. miles) 3 3 3 3 3 3 3 Network Cone figuration Grid Grid Grid Grid Grid Grid Grid Dispatch Logic (no. of pax) 3 5 5 5 5 5 5 Headway (mins.) 15 15 15 15 10 10 10 .Total Pass- * engers Served 91 214 324 432 717 893 1115 Total No. of Buses Reqd. 5 9 9 ll 15 15 15 Mean Wait Time 6mins.) 8.89 8.95 7.15 6.25 7.5 6.68 7.41 Mean Ride Time Cmins.) 6.7 7.67 7.75 7.69 7.96 8.56 9.92 Mean Total . Travel Time 3 Total Driver Hours of Operation Reqd. 15 27 29 39 47 52 60 Vehicle Pro- ductivity NO. of pax/hr./ veh. 11.7 19.33 21.08 20.9 27.24 27.7 26.9 Mean Bus Utilization 52% 41% 53% 53% 56% 62% 69% Total Vehicle Hours of Operation 7.76 11.07 15.37 20.67 26.32 32.24 41.4 *Total number of passengers for 3 hours of simulation. 82 There is no single system that provides the Optimal value of each characteristics at all demand levels. However, if the two recorded times are added to obtain total travel time the variable route variable headway system yields the lowest total travel time at all tested levels of demand. Figure E.1.4 shows that the variable route fixed headway operation yields the highest vehicle productivity for all levels of demand. The variable route variable headway system produces a higher productivity than either fixed route system for demand levels below 80 per hour. Beyond this point the fixed route fixed headway system shows higher productivity. Figure E.1.3 shows the total number of buses required for each system investigated in this study. Since this measure is in- versely related to productivity the variable route fixed headway system has the lowest vehicle requirement for all levels of demand. A similar result is shown if we look at the "Driver Hours" require- ment for each system as shown in Figure E.1.5. . Again, the "vari— able route fixed headway" operation results in the lowest driver hour requirements for all levels of demand. The variable headway operation has a driver requirement lower than the fixed route fixed headway system1in the low demand levels, but becomes nearly coin- cident with the fixed route fixed headway system for higher demands. A final statistical measure "vehicle hours of Operation" is plotted for various demand levels in figure E.1.6. The variable 83 route fixed headway operation produces the lowest vehicle hours of operation among the systems. The variable route variable headway Operation has a lower vehicle hour requirement than fixed time fixed route system up to the demand of 65/hr./sector. These curves cross at a different demand level than in the driver hour statistic because the variable route variable headway system is a more efficient system when measured by bus utilization. Therefore, there are fewer idle hours for the bus driver while waiting for their run to start. Mean Wait Time 20 15 .....1 0 U1 1 84 Variable Route Fixed Headway C] Variable Route (:flVariable/0 Hea y ‘Eéffied/o Route .Eixed Headway D \\A _ Fixed Route Variable Headway 10 20 30 40 60 80 100 Demand/Hr./Sector Total Area of Operation - 3 Square Miles Figure E.1.l Systems Performance Functions - Wait Time Mean Ride Time 20" 15' 5. 85 Fixed Route Fixed Headway 12— “ZS-7““‘-- Variable Headway 20‘? ' C) 0 o / 15 ii /' 9’ o . 10., / A. _. _ ..A ’ ” ’ Fixed Route 5 .. 8, A Variable Headway 10 30 30 3.0 60 36 160 Demand/Hr./Sector Figure E.1.4 Systems Performance Functions - Vehicle Productivity 88 120,, / 110.. ’65 Fixed Route Variable 100.L Headway / 9o .. far 80 1” «p I, a/ 704, 1’; 6O , Variable Route d / Variable ‘ ’0’ Headway 50.. 1£§’,/’ ",§3”””’.‘TEE::::::;”’4y” O 1 ’l’ Fixed Route Fixed 40 i . A /< Headway Total Driver Hours Reqd. for Service 1” ’ Variagle Route Fixed 30'? Headway 20 do 10., 10 20 ‘ 30 £0 60 80 100 Demand/Hr . [Sec tor Total Area of Operation i 3 Sq. Mile Figure E.1.5. System Performance Functions - Driver Hours Required Total Vehicle Hours of Operation 7OTP 60.. 50.1- 401% 30. 201 104 89 K Fixed Route Variable Headway / A/ Variable Route'9/ Variable Headway El ° Variable Route Fixed Headway A I <7- r b b at f V 10 20 30 40 60 80 100 Demand/Hr./Sector Total Area of Operation - 3 Sq. Mile Figure E.1.6 Systems Performance Functions - Total Vehicle Hours of Operation E.2. SYSTEM EVALUATION AND SELECTION The primary objective of this study was to develOp a compre- hensive model which can generate data to assist in the selection of demand actuated bus system alternatives. If one were to make the very simple assumption that demand, service, and operating costs are all non-elastic, then the curves shown in the preceding section would suffice. The bus manager attempting to Operate a bus system for profit could survey the demand and select the system with the lowest pro- duct of vehicle operating time multiplied by his cost per hour and the capital cost of required buses amoritized over their life span. The cost of this bus service could then be compared with a feasible fare.and a decision made. Similarly, if it were apprOpriate to consider public transport- ation as a public service, a specified level of service as measured by waiting time and ride time statistic could be established, and the system that meets these criteria for the projected demand selec- ted. The appropriation required for the service could be deter- mined from the bus statistics, and a decision reached by the appro- priate body. . There is sufficient evidence, however, to reject this assump- tion of a non-elastic market. The structure of existing trip gen- eration and model split models used in transportation planning in- clude a dependent relationship between demand, service and cost. 90 91 Thus, the problem is not that simple, even if the data are known. From the preceding graphs, it is obvious a trade-off must be made between the user's performance characteristics (wait time, ride time, total travel time) and the transit operating charcteristics (number of buses, vehicle hours of Operation, vehicle productivity, bus utilization). The use of constant unit cost figures for each of these characteristics is unrealistic, as they vary with location and in many cases are quite subjective. The value of these system attributes depends largely on the goals and policies of the community for which the transit system is being planned. In economic analysis, where the trade-off between intangible and tangible costs need to be established, it is common practice to use a value scale for both items. This value scale varies depending on the goals of the environment and is often re— ferred to as a utility scale with system alternatives selected on the basis of their score, or utility function. Whether the simple benefit-cost analysis or the more complex utility analysis is preferred by the analyst, there are two common factors in all evaluation techniques: 1. A scaling factor representing the "value" of an incremental change in each of the significant measures of performance (Vi). 2. A quantitative representation of the change in the magni- tude (x1). The evaluation models can combine these factors in an additive, 92 multiplicative, or exponential manner. They can be linear or non- linear, independent or interdependent and time independent or time dependent. This modeling represents an entire area of study and is beyond the scope of this paper. However, for the purpose of demonstrating the applicability of these model outputs to the evaluation process, a simple, additive, linear type model is assumed. The scaling factors (Vi) are also assumed, but variations in these values are tested to demonstrate user-oriented, operator-oriented, and system- oriented assumptions. The model structure used is: Utility (Uj) - f (system performance data, policies) U - V x +1V x + V x + ... + V x j 1 1 2 2 3 3 n n Where xi = The score of a selected system performance parameter V - Value coefficient selected in accordance with the goals and policies The following variables have been selected to demonstrate the applicability of the methodology of utility cost functions for sys— tem alternatives in the selection procedure. X1 - Mean wait time/passenger X2 - Mean ride time/passenger X3 Driver hours required for service X4 Total vehicle hours of Operation Vl = Value coefficient for mean wait time 93 V2 = Value coefficient for mean ride time V3 8 Value coefficient for driver hours required for service V4 - value coefficient for vehicle hours Since these parameters all become increasingly undesirable with in- creasing scores, this model defines the disutility or utility cost (Uj - V x +>V x + V x + V x4) of each system. 1 1 2 2 3 3 4 Selection of high value coefficients for variables x1 and x2 (i.e. mean wait and ride time) will result in the selection of a service oriented operation where passenger service criteria are quite stringent and the nondmonetary benefits of mass transportation given high weights compared to the monetary costs of operation. High value coefficients for variable x3 and x4 (i.e. driver hours required and vehicle hours of operation) will lead to the selection of low cost operation with greater emphasis on tangible costs than user benefits. The methodology presented here does not suggest any specific values for these coefficients, but merely demonstrates a procedure which could be used in evaluating alternative systems. Figure E.2.1 shows the utility cost plotted against danand density for all four systems where the utility coefficients are all equal to 1.0 or there is no bias between user values and operator values. At a very low demand density (IO/hr) the variable route-variable headway Operation results in the lowest utility cost. As the demand 94 increases, variable route fixed headway operation results in the lowest cost system for demand in excess of(20/hr). The fixed route fixed headway operation which results in a high cost at low demand approaches the variable route variable headway system at approximately (lOO/hr) demand. It appears that for low demand situations characteristics of Columbia, Maryland or Haddonfield, New Jersey, the variable route variable headway system should be selected if the user and the operators characteristics are considered to be equally important. Figure B.2.2 indicates the utility cost function for value co- efficients V1 and V2 equal to 1, and V3 and V4 equal to 3.0. This might be representative of a system with a limited subsidy. In this case, some increase in user costs would be accepted in return for lower operating costs. As a result of this change in the weighting functions, the total user time increased from 15.6 to 20.4 minutes at a demand level of 10/hr, while vehicle hours of operation was reduced by 18.6% from 7.76 to 6.31 hours. We are equating this 5 minutes of extra travel time to 1.2 hours reduction in bus hours. At the other extreme, figure E.2.3 represents a case in which the user characteristics are weighted very heavily. This might be characteristics of a highly subsidized service for a model city neighborhood, in which service is the important factor to be con- sidered. The operating characteristics of the variable route var- iable headway system are far superior to any other system at all 3x3 + V4x4) +V 1 2 3 200 v - /4] 130 ‘P E31 160 it I g, 53 Fixed Route 140 Variable Headway I N 120 .. Fixed Route Fixed . 9 Headway Variable Route Variable Headway- -A Hi 8 (le1 + sz ...: o o 80 ‘b 9 Q ’ / O A// '/ §Variable Route Fixed 60 / Headway 1' / f 40 'F 4V 10 20 30 4O 6O 80 100 Demand/Hr . [Sector Figure E.2.1 Utility Cost Functions Ui - (lel + szz + V3x3 + Véx‘) 580 560 (NH 520 500 480 440 420 400 380 360 340 320 300 280 260 220 200 180 160 140 120 100 80 V V IV I110 *V .3s0 96 : C] 502.98 468.53 Fixed Variable Route Variable Headway /A Route Variable Headway Fixed Route Fixed Headway ,0. / aariable Route Fixed Headway ; 4 , ¢ : ‘fi . . 4 G 10 20 3O 4O 60 80 100 Demand/Hr./Sector Figure 8.2.2. Utilitv Cost Functions '° = .' 4- ' + '3“. + ' L1 (\1x1 sz2 \3 3 \4x4) 300 280 260 240 220 200 180 160 140 V1 = 10 V2 3 5 V3 = Va = l # 0- 10 20 Figure 5.2.3. 4}- 3O 97 Fixed Route Variable Headway Fixed Route 9 0 Fixed Headway/ A / Variable Route Fixed Headway A L l r V 40 60 Demand/Hr./Sector Utility Cost Functions Variable Route Variable Headway «P 80 w 100 Ui - (lel + sz2 + V3):3 + V4x4) 400 380 360 340 320 300 280 260 240 220 200 180 98 V = V 8 10 3 4 Fixed Route Variable Headway O 4% Fixed Route Fixed // .. E] I Headway _ 43 / , / ' I /Variabl R0 te Fix d / . Al“ Headway / C! A // G / / _ / /A / 6) Variable Route Variable Headway l I 1|- i q- + 10 20 30 40 60 80 100 Demand/Hr./Sector Figure E.2.4. Utility Cost Functions Ui = (lel + szz + V31:3 + Vaxl‘) 99 V1 = v2 - 3 ’//FK 300 "L v a 1 I I 320 «n V - 2 280 -L 260 «- Axed Route , Variable Headway 240 «T E] 220 «- I Fixed Route Fixed Headway ..0" I <9 ' E] /ariable Route . Variable Headwa 180 “l. / A 160 4. ' <9 9 / 4 Variable / Route Variable 120 " 100 up k ’K 80 d» : 4¢~ : 44: 4* : : : 10 20 30 4O 50 6O 80 100 Demand/Hr . / Sec tor Figure B.2.5. Utility Cost Functions 111 = (lel + sz2 + V3x3 + V4x4) 520.I 480. 440 . 4004 360' 320‘ 280' 240* b T 100 v1=20,v2-1 /é v3-3,v-1 . ’ Fixed Rout Variable Headway £3 E] ’ / / / «ariable Route / / Fixed Headway Variable Route Variable Headway b b p p p p p h p 1'0 20 30 4'0 50 60 7o 80 90 165 Demand/Hr./Sector Figure E.2.6. Utility Coat Functions 101 levels of demand. Comparing this operation with the normal opera- tion (fixed route-fixed headway) at a demand level of lOO/hr., savings in total travel time is 512 with a penalty of 17% in opera- ting hours. In addition to system selection policies, the analyst is also faced with the task of recommending operating policies. In this study variations in system characteristics and utility costs as they are implemented by different bus pooling policies, vehicle to sector assignment policies, sector size and network accessability factors and variations in demand over-time, were also investigated and results are presented in the subsequent sections. The demand for bus transit ridership in any area varies with time, it is extremely important that such variations in demand be considered in the system selection process. As it was shown previously, the demand at which any system becomes "better" than other systems depends on the value coefficients. Thus, different systems can be optimal at different times during a typical day's bus operation. To demonstrate the effect of changes in operating system alternatives, a hypothetical passenger demand distribution (figure B.2.7.) has been assumed. Using the same decision variables; wait time, ride time, vehicle hours of operation and driver hours required and the value coefficient as in figure E.2.2.: V1 = V2 - 1.0; 102 on 12 I - 7//// m. . _\\\\\\\\\ 9 7///////// a 7//// e. - m\\\\\\ H 7//////// 2 - §\\\\\\& 1 rW//////// u o w\\\\\\\\\\\\& , - w\\\\. .... m - 7/////////////////// .. i w\\\\\\\\\\\\\\\\\\\\\ r ., 7/////// . s W m. m. m. m m w“ m" mam 3:: 6.3.36 535a 338 Time of Day & Night Typical Distribution of Rider Danand E0207. Figure 103 Table E.2.1. Utility Costs of Alternative Systems Utility Cost of System Time of Demand Rate Variable Route Fixed Route Day (Nos,/Sq. Mile) Variable Headway Fixed Headway (nonfidynamic) 6—7 AM. 35 * 170 208 7-8 AM 90 298 * 282 8-9 AM 85 284 * 275 9‘10 AM 80 276 * 272 lObll AM 60 * 232 248 ll-12 Noon 50 * 210 233 12-1 PM 40 * 182 217 1'2 PM 35 * 170 208 2-3 PM 40 * 182 217 3'4 PM 30 * 155 198 4'5 PM 100 318 * 292 5-6 PM ' 95 308 * 282 6‘7 PM 80 376 * 272 7-8 PM 60 * 232 248 3‘9 PM 40 * 182 217 9-10 PM 35 * 170 208 10-11 PM 20 * 121 178 11-12 PM 10 * 83 163 104 V3 2 V4 = 3.0 and plotting the utility cost versus demand density, the utility costs of two alternative systems (variable route-variable headway and fixed route-fixed headway) for hourly variations in demand were constructed. Then, reading the cost functions from these curves, Table E.2.1. was develOped which shows the time, demand and utility cost of the two alternative systans. The cost entries preceeded by an asterik indicate the optimal system for that hour. With this information, 3 different alternative operating strategies were compared. 1. Variable route variable headway operation for entire period. 2. Fixed route fixed headway operation for entire period. 3. Combination of the above two systems to achieve minimum total cost. Table B.2.2. Costs of Various Operating Policies Operating Policy Total System Cost 2 Inefficient A. Single System - 3849 3 Variable Route Variable Headway B. Single System - Fixed Route 4218 12 Fixed Headway C. Mixed System - Combination of two 3754 105 This example only serves to illustrate how the results of this study might be used. It is important to point out that the differ- ence in cost saving or magnitude of system efficiency can be quite significant in some cases. This depends on the value coefficients selected for evaluation purposes (which really reflect the goals and policies of the community) and the magnitude of the variations in demand. Other policies, like bus pooling, sector selection, guaranteed pick-up time, dispatching logic and level-of-service provided will also influence these numbers. The approach used in this study pro- vides the tools necessary to assess each of these and bring a little more rationale in the system selection process. E.3. POOLINC POLICY Since the bus utilization factor for the highest numbered bus assigned in all of the previous examples is low, it is possible that serving several sectors from a single bus pool may be more efficient. To test this hypothesis, the variable route variable 'headway simulation model was executed for multiple sector assignment. Three separate policies; single sector, double sector and triple sector were tested for a range of demands to determine the effect on bus utilization as well as the other user and Operator characteristics. The results are indicated in tables E.3.1., E.3.2 and E.3.3. and shown in figures E.3.1. through E.3.4. As shown in figures E.3.1. and E.3.2., there is little dif— ference in the user statistics of mean wait time and mean ride time as expected. The dispatch logic was not altered and thus the effect on the user should be minimal. The single sector assignment policy consistently requires higher driver hours and vehicle hours of operation for all levels of demand as shown in figures E.3.3. and E.3.4.. These performance indicators, which are most directly related to operator costs, illustrate that multiple sector assignment is definitely a better pooling strategy than fixed vehicle to sector assignment. The impact of a multiple sector pooling policy can best be illustrated in the context of the utility analysis. The combination of V1 - V2 3 for this purpose and is shown in figure E.3.5. = l; V a V4 - 3 used in the previous analysis was used 106 Table E.3.1. SYSTEM: OPaRATION STRATEGY: 107 Variable Route Variable Headway Concurrent Collect & Distribution System Performance Data - Pooling Policy SYSTEM ATTRIBUTES 1 Sector Served from 1 Bus Pool Demand/Hr./Sector 20 3O 40 60 80 100 Area/Sector (sq. miles) Avg. Waiting Time per Passenger 8.53 7.6 6.54 5.92 5.12 5.07 Avg. Ride Time per Passenger 6.45 6.86 7.85 6.96 7.5 7.53 Total Travel Time (exp.) 14.98 14.46 14.39 12.88 12.62 12.6 Total Pass. * Served/Sector 84 119 147 223 295 364 Max. No. of Buses Reqd./ Sector (avg. for multi sector strategy) Avg. Driver Hours of Operation/ Sector 10 13 15 22 28 32 Avg. Vehicle Hours of Oper- ation/Sector 4.1 6.1 7.25 9.15 11.32 14.53 Vehicle Productivity 20.5 19.5 20.28 24.37 26.06 25.05 *Total number of passengers for 3 hours of simulation. 108 .COwumasaHm mo meson m How muowsmmmma mo pupae: Hmuohx Hm.¢~ H.0N. mo.mN mo.HN Nh.mH co.wH huH>Hu nonwoum mauwso> MN.¢H NH.HH mm.m HN.~ Nm.m mH.¢ Houumm\:0Humuomo mo mason .:o> .we< on mm mm m.¢H HA m.m nouuom \oOfiunuoao mo mason um>wuv .w>< w.w m.m m.¢ Ahmoumuum youoom_ «uaaa now .w>mv wou00m\vuuwaou woman mo .0: . mmm mam aNN omH moa ms wouuom\vo>uom_ «muuwwommmm Hooch Nw.NH ma.~H no.ma mm.nH mm.cH wo.~H Avouounxov mafia Ho>muu memos o.n mo.n m.n mm.n oo.h mm.“ sowsummua so mafia oven .w24 HN.m om.n mH.e «.0 Nm.u nc.a nowouumma we saw» magmas: .w> ahead: .omv noncom\uo .Dm howl DR bfl DN (‘1 wouoomxuum\vomaua Hoom mam H scum mo>uom muouoom N muusnwuuu some» souuanwuumwn van uuuHHoo unusuaonoo F38: 3:395 338 «23:5 “ESE “Moma 370 .. + m x M > + g” 330.. N > + H x ,4 > " 290-- n-i : n U U) 8 Fixed Route Fixed Headwnv % .. . 3 250 System H w! u D A = Service Area = 3.0 Sq. Mile 210-- 170-- 130+ 0 20 3'0 6'0 8'0 100 Figure E.5.1. Demand/Hr./Sector Utility Cost Function for Variable Demand Levels F. CONCLUSLONS Conclusions of the State of the Art An overview of the state of the art indicates that there has been research, demonstration and implementation studies made on demand actuated bus systems throughout North America and abroad. These studies range from very simple demonstration project to so- phisticated computer simulation studies. However, all of them are oriented towards the case study approach, and do not treat a family of Demand Actuated Bus Systems on a single set of parameters. Thus, it is not possible to define a set of parameter values that would determine the applicability of the various systems. Though simula- tion has been used, most of the models developed are for a specific level of system constraints rather than a family of systems. The following conclusions can be made from the experience of DAB demonstration and implementation studies. 1. Several factors are apparently of great importance to the success of a DAB systems; with service area, service selection and dispatching strategies the most important. 2. Public subsidy is necessary in most cases to support a DAB system. 3. Public subsidy is not a guarantee for system success. This requires a competent system analyst and management and staff. 4. Significant operating economies can be achieved by using small coaches which increase the occupancy-capacity ratio. 134 135 5. Increased vehicle productivity (no. of passengers per hour per vehicle) is possible with DAB systems. 6. Ridership increase due to implementation of DAB systems compared to fixed route-fixed headway systems is significant. 7. Off peak fare reductions policy results in redistribution of ridership--i.e. peak smoothing--but does not affect total rider- ship or revenue. 8. Ridership seems to have limited sensitivity to fares between 25¢ and 60¢. 9. Manual dispatching works fairly well for smaller service areas with few buses. For larger areas computer routing and dis- patching is required. 10. Careful analysis of service area may allow a system to be efficiently operated under many to one or many to few environments. 11. For high density (riders) areas, fixed time fixed route systems work as well as any DAB system. The following conclusions can be made from the research and develOpment studies: 1. System simulation provides a tool for system evaluation ' and selection. 2. The concept of DAB systems is technically feasible. 3. Financially the systems cannot be self supporting. 4. Demand estimation analysis shall be a part of system study prior to implementation for selecting viable service criteria. 136 5. Heuristic Traveling Salesman algorithms work for optimal routing in most of the existing simulation models. 6. Computer dispatching is economical for large systems. 7. The limited sensitivity analysis presented in various studies and papers indicate sensitivity of systen performance to waiting time, trip time, and link travel time for specific systems only. 8. The other simulation studies have tested certain variables; demand, level of service, etc. and found significant differences in the user and performance characteristics, but they have not compared alternative systems on a single base. Conclusions of this Stgdy The general purpose simulation model developed as a part of this research can replicate operations Of alternative systems within the spectrum Of demand actuated bus systems. The accumulation and production of the statistical data by the model provide an excellent base for the system evaluation and selection process. The analysis regarding the effect Of variable demand condition on system performance parameters (both user and operator) indicates that significant gains in system attributes can be attained by changing Operating policies for varying demand levels. This study also reaffirms the hypothesis that the selection Of a system without a comparative evaluation of alternatives may result in significant inefficiencies. 137 The utility cost analysis as presented in this study pro- vides a means for evaluating system alternatives for selection of optimal Operating policies. This method can use both tangible and intangible cost parameters in the perspective of specific goals of the community. The analysis of the effect of fixed vehicle to sector assign~ ment, as Opposed to multi—sector service from a single bus pool, indicated significant differences in some cost parameters. It can be concluded that multi-sector Operation by a central bus pool provides cost savings in most cases. The system cost parameters indicated significant sensitivity to network patterns. Thus, it appears that a method Of classifying street patterns must be developed before general conclusions re— garding system selection can be made. The attempt to determine the effect Of service area on the system attributes was inconclusive. BIBLIOGRAPHY 10. ll. 12. BIBLIOGRAPHY Stevens, Robert D. and Smith, Richard L., "Demand Bus for a New Town", H.R.B. Special Report 124. Bauer, H.J., "Case Study of a Demand Responsive Transportation System", H.R.B. Special Report 124. Roos, Daniel, "Dial-A-Bus System Feasibility", H.R.B. Special Report 124. Guenther, Karl, "Incremental Implementation Of Dial-A-Ride Sys- tems", H.R.B. Special Report 124. Report Of Seminars on "Demand Modeling and Estimation of Demand", "System.Attributes and Performance", "System Simulation", "Pricing and Economic Evaluation" and "System Evaluation", H.R.B. Special Report 124. Gwynn, David W., "Dial-A-Ride Demonstration Project in Haddon- field", Traffic Engineerigg, Volume 42, NO. 12, September 1972. Ayres, Robert 0., Mckenna, Richard and walker, M.L., "Evaluation of New Urban Transportation Systems", H.R.B. Record 367. Gustafson, R.L., Curd, H.N. and Golob, T.F., "User Preferences for a Demand-Responsive Transportation System: A Case Study Report", H.R.B. Record 367. Gustafson, R.L. and Golob, T.F., "Economic Analysis of a Demand Responsive Public Transportation System", H.R.B. Record 367. Archer, E. and Shortreed, J.H., "Potential Demands for Demand Scheduled Bus Services", H.R.B. Record 367. Howson, L.L. and Heathington, R.W., "Algorithms for Routing and Scheduling in Demand Responsive Transportation Systems", H.R.B. Record 318. Vitt, J.R., Bauer, H.J., Canty, E.T., Golob, T.F. & Heathington, K;W., "Determining the Importance Of User Related Attributes for a Demand-Responsive Transportation System", H.R.B. Record 318. 138 13. 14. 15. l6. 17. 18. 19. 20. 21. 22. 23. 24. 139 Wilson, N.H.M., Sussman, J.M., Higonnet, B.T. and Goodman, L.A., "Simulation of a Computer Aided Routing System (CARS), H.R.B. Record 318. Wilson, N.H.M. and Kennedy, John Darrel, "Modeling the CARS System", Paper presented to Ohio Transportation Research Forum, May 1969, Regina Telebus Study, Summary Repgrg. "Columbus, Ohio, Model Cities Dial-A-Ride System-Design and Implementation", Final Reporp, Prepared berransportation Research and Planninngffic LFord Motor Company. "Dial-A-Ride for Middletown, Ohio", Prepared by T.R.P.O. Ford Motor Company. Mason, F.J. and Mumford, J.R., "Computer Models for Designing Dial-A-Ride Systems", Presented in Automotive Engineering Congress, Detroip, Michgiany January 1972, Society of Automotive Epgineers. Guenther, K.w., "Public Transportation Service Innovations: What are the lessons for Dial-AIBus?" T.R.P.O. Ford Motor Co. Publication. Guenther, K.W., "Implementation of Dial-A-Ride in Your Community: A Practical Checklist", T.R.P.O. Ford Motor Company_Publication. Heathington, K.W., Miller, J., Knox, R.R., Hoff, 0.0. and Bruggenman, J.M., "Computer Simulation of a Demand-Scheduled Bus System Offering Door to Door Service", Paper presented at the H.R.B. 47th Annual Meeting, 1968. Blurton, M., Final Report on the Peoria-Decatur Demonstration Projects. Bureau of Economic and Business Research, University of Illinois; 1969. Guenther, K.W. and Givens, W.E., "The Courier: A Prototype Vehicle for Dial-A—Ride Service", SAE Congress, Paper 700186, January 1970. Bruggeman, J.M., Heathington, K.W., "Sensitivity to Various Parameters of 8 Demand Scheduled Bus System Computer Simulation Model", H.R.B. Record 367. Guenther, Karl, Ford Motor Company's Role in Dial-A-Ride Development-1972 and Beyond. H.R.B. Special Report 136. APPENDIX I 140 IIIIIIIIIaassaaaaaaa*aaaaaasuaanaIssuiaaaaaauaauaaasaaa***«I4*uasanaauaaufiusisauu IIIpause.iauauaaaussauaaauaaaauaa*Iuaaauisuaaaauauususuauauaa*Inausasuaeiaeusasua usaaauas}448***************§*****auauuaauraaa844444..«aas»******«******I****§saus suausaasa{IavsaauauswseueI*§***§**I«*****ssusaasassessusua*euausssuuaaasaaaaauaeu IIIIIIIIIIIIIassauuassiasaaaa8*I********II**I***444*I»***§***4*4*aaaaaaaauauuauas sow u tmmdd w ummezmu ~BH>HHU¢ cm H Mummcm mo mmmnsz mm H N9H0¢m¢u .sz .mr u rdao¢mm m n UHUON muafimMHa .§****§***{***********iifiii***§***********ii**§******l*§§t*i9!***¥§*i*§§**§***§*ifl Q§§§§§§§§*****§*i******§*§*§****i**§*§**§******************Q****************§***. §*§§**¥*§*§****§****i*******§***i********i*§*iiii*§*****§****§******************* ***§********§************§***************§**fiiiifiiiifi*****#*********************§ *******¥******§*fl**fii**ffi§i**********************§ifi*ii*i*§******§********§***§* 141 N n eacH §§**§*****§***§**§*§***§*****§******l*§**§*****§#§**i*****§**§**i*********** w ooo.om u mnz¢£ma ZOHBUMAAOU Hszmmm mmaom o.n u Bmaam mo mznp 0.0N 0 mp0: mm» mazmmm _ euagou . mmamue . _ -Hquas _ mace . mace . .mmeu . .mmam . .mwam _ muses _ mam _ 2am: _ gases . gases _ mumzaz _ mumsez _ amazez _ a mam .ZH: mm.mpp OB .ZHsoo.oc sump ”Ema. N mmmrnz Hzmuomm '5'----------'-'---------------------"--------------'--"--"--'---'-'-"--------'--'-'--. n _ 0.0 _ 0.0 ._ 0 _ 0.0 . . 0.0 _. 0 _ 0.0 . 0.0 .— .0 _ . _ _. . — ._ _ _ .. a .op.on .ms.o _. m , .ee.ep .mo.m . ..1 n _m~.=~ _m=.h .. m . . __ _ _ .. . . .- 0 .mn.0~ _~0.MP ._ Np _0m.pm _nm.nw _- o _nm.m~ .om.m _. N . . .. . _ .. _ . .- ~p .Np.on _p=.sp .. 0 .0—.h~ .mo.mp .. h .N0.nm .hm.0p .. P . . .- - _ .. . _ .- I we: 5. _ 2.2:. a. .. ...z: T. _ : mmso . .xen _ zen: ._ «use _ .xem _ sew: .. mmso . .24” _ zame ... .Omm . n. museum ._ ~. magnum _. . .. meaoum .. mesa-.zme zo sumac mmaazun om ~ma>m mzos mm uz~azmruem .monamneeam emcu~-euau= emcemmmam 145 n . 0.0 . 0.0 _. 0 _ 0.0 . 0.0 _. 0 _ 0.0 . 0.0 ._ s . . . .. . . .. . . .. s .m—.o— .mo.o _. ~ .mm.o~ ....a ._ m .6..e. _on.e _. n . . . .. . . . .. _ . .. a .o~.~— .oo.o .. m .me.~p .mo.op _. n .h..m— .sm.m _. N . . _. . . .. . . / ... m _-..~ .um.op .. m .Nm.~— . .mm.o. —. o. _ws.om .mm.~. _. p . . _. . . , .. . . .. .zuz m. _ . ...zun m. _ . ...zHe m. . . __ muse . .uau _ mama ._ muse _ .xa: . same __ muse _ .xec . zamn ... .umm n. museum . ._ N. museum ._ P. monumm _. . mnH81.zmo zo ammm N200 mH wzaazmcwmm smUHHmHH¢Hm smnuflumaHm: mmUzmmm~.sm .om.mm .. up .mm.m~ .ms.m. _. m . . __ _ _ ._ . . .. mu ..m.em ..m.:w __ me _o~.mm .oo.s~ _. mm _ee.sm .No.~m __ F . . _. . . _. . _ ._ .zaz m. _ . ...zH: m. _ _ __.2Hs me . _ __ «use . .xa: _ mama __ mmso _ .xa: . zauu _. mm>o _ .xam . zamx ... .omm _ mu 0090mm __ m0 mohumm ._ Pu monumm —_ wzH81.sz zo ammam mmBDZHn ow Nmm>m mzoa mH UZHHzmnwmm .mUHHmuhahm :mcHHI.>¢men.H an mmozmmmdm 147 00.o~ . p pw.—N :0.0~ MU‘N p:.—~ 00.0w P mN.o~ am.0p :N.av mama M009 ~4HOF ow: "5..“ HUuuuU .mmdm . mmmuzmmmam hm M \ on N o .meamHu .mmum . N p . manta: mach mmaaan mm.mww H mm: 2H usua A¢HOH sausquzH mmsoa m masseuse P xmmnau mam APPENDIX II (Flow Charts) 148 INPUT Geographic Distribution of Area Time Distribution of Demand Spatial Distribution Demand System Constraints Dispatch Logic Max. Hdwy . Veh. Capacity * NO. of Buses in Bus Pool Read Input Data Prob. Distribution Of Collection and Distribution Demand Call Generate Prob. Distribution Generates Demand by of NO. of Passengers Space and Time forms in Demand Calls Point to Point Travel ' ' Time Matrix l Call "RD List" Screens unordered List and Splits in Respective Sectors F“! '1 ‘ 0.- ..IJ.’ ‘Ilufii -..-U ,- v... ..e... 11. I... «m. o:— NO ~ 149 ; Call "Slice" Prepares List for Pick Up and Delivery Of Passengers for all Sectors Call "D List" Call "us Pool" Assigns Bus to Demand Execute Service and Collect Statistics a) by passengers b) by bus Print Tour Statistics Are There Any Mbre Demand For Service? N0 Call "STATIS" End 150 SUBROUTINE "GENERATE" Entry "GENERATE" V Read: Type of Rider Arrival Function, Mean Arrival Rate, Time Length of Simulation enerate Arrivals of Riders at Specified Points from Random Numbers and Cumulative Density Function Passenger Arrival List (Unordered) Prepared Passenger Node to Node Travel Time Matrix From Given Dimensions. of Space, Density of Passenger Generation Points & Travel Time anction Return 151 Subroutine "RD List" Enter "RD List" (Sect.) l” All Read Unordered List One By One Return with all Allocate Each Dmnd. lst Time Slices Pt. in Appropriate Sector List , ‘ ‘ Go To Next Sector Have All Demand Points Been Read Have lst Time Slice For All the Sectors Made Yes Print Out All List Select Appropriate [_ Call Slice List 152 Subroutine "Slice" Enter "Slice" Read Dmnd. Pts. One By One Is Bus Occ. Less Than Bus Capacity? No Yes Is Bus Occ. Less Than J Disptach Logic? NO Yes Is ABS. Time of Generation ess Than Headway NO Constrain Yes Return 6 I Save This List For "D List" 153 I I Subroutine D List Entry D List - Are Rgtgrnaszt :00 No There Any 0 em n 5 Demand To Be Exec? Yes 18 Yes This lst Time Slice? No ‘re a NO here Any Next BUS Time 0C rI Demands In this Secto Yes Are ¥EXt EWgoguiaSt Bus There Dist. me nutes Demands? Yes Select Proper List 1 Call "Slice 2" i Note Next Bus Time Sector I.D. J i ,{ Call "Select" L Return With Call Routing Routine Ordered List 71‘ "Distr" Subroutine "BS Pool' 154 Entry "BS Pool Test If Bus is Needed Record Information If a Bus Is Returning L Return Bus to Available List NO Return It Is End of Simulation Test If It Is Initializing Stage i Yes Yes Test If a Bus Is Available Yes Calculate Delay Until Next Bus is Available i Print All Pertinent Statistics (Bus Utilization, Etc.) Initialize all Para- meters & Make Buses Available Return Return — Wait 'R' Minutes F‘l Record Which Bus is Available - Note Bus I.D. NO. 1 The Bus is Ready T Execute Demand. Strike That Bus Of Available List Return 155 Subroutine "Select" Enter Sub-Routine "Select" Abs. Next Bus Time . Othhe':}1§edtLiStS Read Abs. Time of or 9C ors Next Bus for all the Lists Does lst Have Least Next Bus Time? Yes Does 2nd List y Yes Have Least Next Bus Time?’ No Does (n—1)th u Yes ist Have Least Next Bus Time? No Select This List For Execution Select 'N'th List _ 1 Return ] 156 Subroutine Slice 2 Enter Slice 2 Delete From Demand List and Print Message Yes NO '1‘ Read Demand One by One & Accumulate Has Max. Waiting Time Constraint Been Violated? NO Is Bus Occ. Les— No Return with Sliced List NO Than Capacity of Bus? Yes Is Bus Occ. Less Than Dispatch Logic? Yes Is Abss Clock ime of Gen. Less Than Hdwy + Last Bus Time? Yes Are There Any . More Demand? Yes