Characterisation and Mitigation of Radar Interference in Automotive Applications

Radar has emerged as an integral sensor for imaging and scene perception to achieve full situational awareness for vehicle automation. The increase in radar penetration rate and a limited spectrum availability for automotive uses have given rise to coexistence problems,
where radars directly or indirectly interfere with each other. Interference causes a degradation in the radar performance, particularly an increased noise (interference) floor, making it difficult to detect, identify, and classify targets accurately, as well as perceive the vehicle’s
surroundings which is essential for trajectory estimation and path planning applications. This might lead to potential accidents especially in dense road conditions.

The work presented in this thesis investigates the mutual interference in 77 GHz millimetrewave frequency modulated continuous wave radar to develop suitable mitigation strategies without a detrimental effect on the performance of the radar. A mathematical model is
developed to systematically assess and quantify the performance of radars in an interference background. Various types of interference expected in automotive scenarios are first classified, such as fast-chirp, slow-chirp, synchronous, and asynchronous. The appearance and impact of interference are then evaluated at each stage of the victim radar’s signal processing and analysed in the time, range, range-Doppler, and range-azimuth profiles.

The characteristics of interference in each profile are exploited to develop appropriate mitigation techniques along with their comparative performance analysis. For pre-detection mitigation, waveform randomisation and spectral windowing – inherently used in radar
processing – are discussed to minimise the impact of interference. Techniques to detect interference and estimate the waveform parameters of the interfering radars are developed using the radar data from real-road measurements, which are used for cognitive interference mitigation. In the spatial domain, adaptive digital beamforming is discussed to suppress interference for various radar configurations and penetration rates. In imaging techniques, Doppler beam sharpening is developed to significantly reduce the impact of interference with complete removal of interference from the radar’s field of view in some cases.

A comprehensive statistical evaluation of interference and its comparison with the additive white Gaussian noise (AWGN) have demonstrated that most types of interference have statistics comparable to those of AWGN. Therefore, the currently existing detection techniques remain valid for target detection in an interference background.

Lastly, an experimentally verified holistic end-to-end system level radar simulator is developed to model, evaluate, and optimise radar performance in diverse scenarios with various traffic, roadside, and interference density conditions.