PRED-RAG: a Predictive Radial Grid for Automotive Radar Multipath - Identification Identification of objects created by the radar multipath phenomenon, with focus on low computational complexity.

Sammanfattning

Automotive radar sensors are crucial for advanced driver assistance systems butare susceptible to the multipath phenomenon, where radio waves reflect multipletimes between surfaces, creating false "ghost" objects that can trigger unnecessarysafety interventions. Previous work relies on restrictive assumptions about reflectionsurfaces and environmental conditions, yielding solutions that perform well in specificscenarios but demonstrate limited generalization capabilities in the complex, diversesituations encountered during real-world driving. This thesis addresses the challengeof identifying radar multipath objects in real-time environments, focusing on developingan algorithm that maintains low computational complexity while achievinghigh accuracy. We established a development and evaluation pipeline using syntheticdata together with a simulation framework, enabling data driven development of ouralgorithm. We propose the PRED-RAG algorithm, a novel approach that utilizesa radial grid structure combined with host motion prediction of static detectionsfor enhanced high-level environment mapping. The algorithm identifies tripletsconsisting of a ghost object, reflection point and true object, then evaluates themusing velocity-based criteria. When compared to a state-of-the-art algorithm, ourapproach demonstrates superior performance in both accuracy and computationalefficiency across various driving scenarios. The PRED-RAG algorithm achieves94.43% accuracy for high-priority objects compared to 39.26% for the baseline, withsignificantly better generalization capabilities, particularly in complex environments.The geometric properties employed in the grid-based approach effectively separateghost objects from true objects while maintaining runtime performance suitablefor real-time automotive applications. This work contributes to safer autonomousdriving systems by reducing false objects that could lead to unnecessary emergencyinterventions.