Smart Spectrum Access: Multi-Arm Bandit Algorithms in an Adaptive Interference Mitigation Strategy for Automotive Radar

Since frequency-modulated continuous-wave (FMCW) radars are widely used in automotive applications, mutual interference has become a critical challenge in dense traffic scenarios. Increasing radar density degrades target detection performance and limits the effectiveness of existing mitigation strategies. This paper proposes BanditFMCW, a learning-based smart spectrum access strategy that detects interference in real time and adaptively reallocates the radar operating band using multi-arm bandit (MAB) algorithms. In particular, a sliding-window upper confidence bound (SW-UCB) algorithm with side observations is developed to cope with non-stationary interference conditions. BanditFMCW is evaluated through Monte Carlo simulations and realistic highway traffic simulations. It is compared with the random orthogonalization (RO) bandit algorithm and five state-of-the-art mitigation strategies using four figures of merit (FOMs). The results show that SW-UCB with side observations reaches orthogonal transmit signals faster than RO while reducing both the likelihood of operating in interfered bands and increasing the expected signal-to-interference ratio (SIR) in local and global deployment settings.