Autonomous systems such as self-driving cars and drones are going to transform our lives in many ways. One fundamental element of these systems is the accurate and robust sensing of the surrounding environments. Current solutions involve a multi-modal approach where varieties of sensors are utilized to achieve the best performance. Among all the sensing modalities, millimeter wave radar is playing a critical role due to its high resolution and the capability of sensing in the dark environment. As the cost, size and weight of millimeter wave radars decrease prominently in recent years, there is a significant opportunity to widely adopt it for a variety of sensing tasks.This thesis presents a MIMO FMCW millimeter wave radar system at 77-81 GHz band. It is a small-size and high-accuracy solution for sensing objects in the environment. Our system is designed to estimate the object's surface shape, orientation, curvature, boundaries, and 2D location by detecting the object's surface at multiple positions on a planned trajectory. We evaluated our system on a metal rectangular box and a metal cylinder. The experiment results show that our system is able to determine the surface type (planar or curved) correctly and achieve cm-level accuracy on the boundaries and location estimation. The orientation error is bounded to 2.14 degrees. The curvature estimation is highly accurate for the middle part of the curved surface.
Read
