TR2021-138

Extended Object Tracking with Spatial Model Adaptation Using Automotive Radar


    •  Yao, G., WANG, P., Berntorp, K., Mansour, H., Boufounos, P.T., Orlik, P.V., "Extended Object Tracking with Spatial Model Adaptation Using Automotive Radar", International Conference on Information Fusion (FUSION), November 2021.
      BibTeX TR2021-138 PDF
      • @inproceedings{Yao2021nov,
      • author = {Yao, Gang and WANG, PU and Berntorp, Karl and Mansour, Hassan and Boufounos, Petros T. and Orlik, Philip V.},
      • title = {Extended Object Tracking with Spatial Model Adaptation Using Automotive Radar},
      • booktitle = {International Conference on Information Fusion (FUSION)},
      • year = 2021,
      • month = nov,
      • url = {https://www.merl.com/publications/TR2021-138}
      • }
  • MERL Contacts:
  • Research Areas:

    Computational Sensing, Dynamical Systems, Optimization, Signal Processing

Abstract:

This paper considers extended object tracking (EOT) using high-resolution automotive radar measurements with spatial model adaptation. This is motivated by the fact that offline learned spatial models maybe over-smoothed due to insufficient offline labels and can be mismatched to onboard radar sensors due to different specifications. To refine the offline learned spatial representation in an online setting, we first derived a modified unscented Rauch-Tung-Striebel (RTS) smoother that explicitly accounts for the offline learned model (i.e., the B-spline chained ellipses model). The smoothed state estimates are then used to create an online batch of \emph{state-independent} training data which is finally utilized by an expectation-maximization algorithm to update the spatial model parameters. Numerical validation is provided to verify the effectiveness of the proposed online model adaptation scheme.