MATS: An Interpretable Trajectory Forecasting Representation for Planning and Control

Boris Ivanovic, Amine Elhafsi, Guy Rosman, Adrien Gaidon, Marco Pavone
Proceedings of the 2020 Conference on Robot Learning, PMLR 155:2243-2256, 2021.

Abstract

Reasoning about human motion is a core component of modern human-robot interactive systems. In particular, one of the main uses of behavior prediction in autonomous systems is to inform robot motion planning and control. However, a majority of planning and control algorithms reason about system dynamics rather than the predicted agent tracklets (i.e., ordered sets of waypoints) that are commonly output by trajectory forecasting methods, which can hinder their integration. Towards this end, we propose Mixtures of Affine Time-varying Systems (MATS) as an output representation for trajectory forecasting that is more amenable to downstream planning and control use. Our approach leverages successful ideas from probabilistic trajectory forecasting works to learn dynamical system representations that are well-studied in the planning and control literature. We integrate our predictions with a proposed multimodal planning methodology and demonstrate significant computational efficiency improvements on a large-scale autonomous driving dataset.

Cite this Paper

BibTeX
@InProceedings{pmlr-v155-ivanovic21a, title = {MATS: An Interpretable Trajectory Forecasting Representation for Planning and Control}, author = {Ivanovic, Boris and Elhafsi, Amine and Rosman, Guy and Gaidon, Adrien and Pavone, Marco}, booktitle = {Proceedings of the 2020 Conference on Robot Learning}, pages = {2243--2256}, year = {2021}, editor = {Kober, Jens and Ramos, Fabio and Tomlin, Claire}, volume = {155}, series = {Proceedings of Machine Learning Research}, month = {16--18 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v155/ivanovic21a/ivanovic21a.pdf}, url = {https://proceedings.mlr.press/v155/ivanovic21a.html}, abstract = {Reasoning about human motion is a core component of modern human-robot interactive systems. In particular, one of the main uses of behavior prediction in autonomous systems is to inform robot motion planning and control. However, a majority of planning and control algorithms reason about system dynamics rather than the predicted agent tracklets (i.e., ordered sets of waypoints) that are commonly output by trajectory forecasting methods, which can hinder their integration. Towards this end, we propose Mixtures of Affine Time-varying Systems (MATS) as an output representation for trajectory forecasting that is more amenable to downstream planning and control use. Our approach leverages successful ideas from probabilistic trajectory forecasting works to learn dynamical system representations that are well-studied in the planning and control literature. We integrate our predictions with a proposed multimodal planning methodology and demonstrate significant computational efficiency improvements on a large-scale autonomous driving dataset.} }
Endnote
%0 Conference Paper %T MATS: An Interpretable Trajectory Forecasting Representation for Planning and Control %A Boris Ivanovic %A Amine Elhafsi %A Guy Rosman %A Adrien Gaidon %A Marco Pavone %B Proceedings of the 2020 Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2021 %E Jens Kober %E Fabio Ramos %E Claire Tomlin %F pmlr-v155-ivanovic21a %I PMLR %P 2243--2256 %U https://proceedings.mlr.press/v155/ivanovic21a.html %V 155 %X Reasoning about human motion is a core component of modern human-robot interactive systems. In particular, one of the main uses of behavior prediction in autonomous systems is to inform robot motion planning and control. However, a majority of planning and control algorithms reason about system dynamics rather than the predicted agent tracklets (i.e., ordered sets of waypoints) that are commonly output by trajectory forecasting methods, which can hinder their integration. Towards this end, we propose Mixtures of Affine Time-varying Systems (MATS) as an output representation for trajectory forecasting that is more amenable to downstream planning and control use. Our approach leverages successful ideas from probabilistic trajectory forecasting works to learn dynamical system representations that are well-studied in the planning and control literature. We integrate our predictions with a proposed multimodal planning methodology and demonstrate significant computational efficiency improvements on a large-scale autonomous driving dataset.
APA
Ivanovic, B., Elhafsi, A., Rosman, G., Gaidon, A. & Pavone, M.. (2021). MATS: An Interpretable Trajectory Forecasting Representation for Planning and Control. Proceedings of the 2020 Conference on Robot Learning, in Proceedings of Machine Learning Research 155:2243-2256 Available from https://proceedings.mlr.press/v155/ivanovic21a.html.

Related Material