“Around-the-Corner” Radar: Particle Filters for Non-Line-of-Sight Target Tracking in the Presence of Ambiguities

Recent research works show growing interest in exploiting multipath radar signals, produced by reflections on surrounding surfaces to detect and locate non-line-of-sight (NLOS) targets. However, the localization accuracy is often compromised by ambiguities, stemming from limitations of hardware and processing to distinguish between similar paths across different target positions. To address this issue, tracking has been shown efficient in resolving ambiguities through long-time integration. However, tracking estimates still rely on ambiguous likelihood information, which can potentially cause the filter to converge onto incorrect trajectories. This article proposes two improvement axes for tracking estimation in the presence of ambiguities. The first aspect involves refining the measurement model by managing the number of model paths. The second aspect tackles the problem of the presence and absence of a target using a track-before-detect approach to improve the tracking robustness in weak path strength scenarios. These solutions are evaluated in synthetic signal simulations and real-data experiments to show improvements in tracking estimation.