Real-time Interference Mitigation for Automotive Radar

Automotive radar is an indispensable component of advanced driver assistant systems (ADASs). As more vehicles are equipped with automotive radars, interference among the radars becomes a problem and can severely degrade the performance of target detection. An interference mitigation algorithm must minimize the impact of interference under highly dynamic driving conditions while meeting stringent processing time constraint. In this paper, we present Soteria-a real-time compressed sensing based interference mitigation algorithm for Frequency Modulated Continuous Wave (FMCW) radar system. Soteria identifies the interference signal by exploiting the sparsity of the signal in frequency-time domain. It then separates the intended signal and interference signal based on the Compressive Sampling Matching Pursuit (CoSaMP) algorithm. By exploiting the intrinsic correlation between the input data from adjacent time slots, Soteria narrows down the search space for CoSaMP to find a solution. To further accelerate computation time, Soteria pursues a parallel implementation based on GPU computing architecture. Simulation experiments show that Soteria can achieve ~10ms processing time and outperforms the state-of-the-art algorithms in terms of target detection.