Extended Object Tracking with IMM Filter for Automotive Pre-Crash Safety Applications
Anusha Hanumegowda, Soumya Dewangan, Srihari Bhupala, Frank Gruson, Dagmar Steinhauser · 2021 18th European Radar Conference (EuRAD) · 2022
While automotive radar is widely used for comfort applications such as ACC (Adaptive Cruise Control) and safety applications such as EBA (Emergency Brake Assist) since more than 20 years, it's recently also gaining interest in applications such as 360° near-range sensing for pre-crash applications. Pre-crash scenarios exhibit multiple motion mode changes within a fraction of seconds and as such they demand for faster data acquisition and more efficient tracking mechanisms compared to the traditional applications mentioned above. In this work, fast mode pre-crash radar detections are tracked using a combination of Interacting Multiple Model (IMM) and Extended Object Tracking (EOT). The IMM-EOT approach outperforms tracking using either IMM or EOT alone in terms of accuracy, adaptability to motion mode changes and ability to provide additional details of the track such as extent information. The method is verified on radar measurements from both indoor and outdoor test facilities. The measurement set up resembles accident scenarios with both full and partial overlap of the target with respect to ego vehicle. The effect of reduced chirp count and critical maneuvering on tracker performance is studied.