Vehicular Radar Network: Cross-layer Design Optimization

Abstract

With the increasing proliferation of radars on vehicles, interference among vehicular radars is becoming a serious issue. In this thesis, we consider the most widely adopted frequency modulated continuous wave (FMCW) automotive radars. We analyze and propose methods to mitigate two types of interference, coherent interference and incoherent interference. Coherent radar interference is the interference with the same FMCW waveform parameters of victim radar, while incoherent radar interference is the interference with different FMCW waveform parameters. For dealing with coherent interference, we propose to adopt media access control (MAC) protocols for mutual radar interference mitigation. We propose two cross-layer performance metrics - multiple access capacity and probability of target misdetection to quantify the network performance under each MAC protocol. Based on our analysis and extensive simulations, we find that pure random access achieves a very poor trade-off between multiple access capacity and probability of target misdetection. However, such a trade-off can be significantly improved by frequency hopping and phase coding. This shows that proper MAC protocols can achieve very good coherent interference mitigation performance even without synchronization and coordination. We describe new insights behind such gains that can be valuable for FMCW radar MAC design. To improve the range-velocity detection performance under incoherent interference, we exploit distinguished features of target and interference components in the fast-time-frequency (fTF) representation and propose to directly recover the underwhelmed target component via the fast-time-frequency mode retrieval (fTFMR). This is achieved by utilizing the Fourier synchrosqueezed transform (FSST) and introducing robust ridge detection that, in combination, guarantees that the recovered fast-time errors of the target signal are bounded at separable time intervals. Comprehensive performance comparison with a list of baseline methods shows that the fTFMR method yields higher output signal-to-interference-noise ratios (SINRs) at both the range and velocity domains and reduces the false alarm compared to the state-of-the-art fast-time interference mitigation methods. To improve the angle detection performance under incoherent interference, we formulate a spatial multiple-input multiple-output (MIMO) detection problem that counts for interference structure and propose a subspace-based detector in the framework of generalized likelihood ratio test (GLRT). We derive the exact theoretical performance of the proposed subspace-based GLRT detector and show that it’s a constant false alarm rate (CFAR) detector. Further, we extend our analysis to massive MIMO radar and show that the proposed subspace-based GLRT detector has perfect angle detection performance under interference when the array size goes to infinity. We theoretically derive the massive MIMO regime where the proposed subspace-based detector is guaranteed to achieve a high enough detection probability under interference. Extensive simulations show the superior performance of our proposed subspace-based GLRT detection scheme compared to the classic detector and validate the effectiveness of massive MIMO for interference mitigation.