Extended Object Tracking assisted Adaptive Clustering for Radar in Autonomous Driving Applications

Abstract

Multiple Extended Object Tracking in autonomous driving scenarios must be applicable in highly varying environments such as highway scenarios as well as in urban and rural environments. In this paper, a flexible UKF-based Interacting Multiple Motion (IMM) model extension for the Random Matrix Model (RMM) framework is introduced for nonlinear models. In addition to that, an adaptive clustering method where the provided tracking prior information is invoked to obtain stable clustering and tracking in varying environments with different objects and varying object types is derived. The effectiveness of the filter and clustering method is demonstrated in a real-world scenario.