Joint Detection and Cancellation of Cross-Interference in Automotive FMCW Radars

Abstract

Frequency Modulated Continuous Wave (FMCW)radars have emerged as a promis ing sensing modality for implementing Advanced Driver Assistance Systems (ADAS) for autonomous vehicles due to their capability of all-weather, precise, and high-resolution detection of objects. Dense deployment of FMCW radar equipped autonomous vehicles in urban scenarios can increase the risk of cross-interference between multiple radars operating in close proximity. Cross-interference leaking into the received baseband sig nal of the victim radar can increase its perceived noise floor which can significantly degrade the detection of weak objects/obstacles. In this thesis, we study the structure of cross-radar interference and its impact on detection of weak objects. We propose an algorithmic framework for canceling the cross-interference in the received baseband signal at the victim radar. We develop a novel structure-aware proximal subgradient descent algorithm to blindly separate the interference-free received baseband signal consisting of complex sinusoids and the temporally-sparse cross-interference signal. We conduct experiments using both simulated and real-world FMCW radar measure ments to demonstrate the accurate, robust and real-time cross-interference cancellation performance of our proposed algorithm compared to existing interference mitigation methods. Lastly, we also discuss the techniques to accelerate the convergence and fur ther reduce the runtime of our proposed interference cancellation algorithm.