Adaptive Conformal Prediction for Motion Planning among Dynamic Agents

Anushri Dixit, Lars Lindemann, Skylar X Wei, Matthew Cleaveland, George J. Pappas, Joel W. Burdick
Proceedings of The 5th Annual Learning for Dynamics and Control Conference, PMLR 211:300-314, 2023.

Abstract

This paper proposes an algorithm for motion planning among dynamic agents using adaptive conformal prediction. We consider a deterministic control system and use trajectory predictors to predict the dynamic agents’ future motion, which is assumed to follow an unknown distribution. We then leverage ideas from adaptive conformal prediction to dynamically quantify prediction uncertainty from an online data stream. Particularly, we provide an online algorithm that uses delayed agent observations to obtain uncertainty sets for multistep-ahead predictions with probabilistic coverage. These uncertainty sets are used within a model predictive controller to safely navigate among dynamic agents. While most existing data-driven prediction approaches quantify prediction uncertainty heuristically, we quantify the true prediction uncertainty in a distribution-free, adaptive manner that even allows to capture changes in prediction quality and the agents’ motion. We empirically evaluate our algorithm on a case study where a drone avoids a flying frisbee.

Cite this Paper

BibTeX
@InProceedings{pmlr-v211-dixit23a, title = {Adaptive Conformal Prediction for Motion Planning among Dynamic Agents}, author = {Dixit, Anushri and Lindemann, Lars and Wei, Skylar X and Cleaveland, Matthew and Pappas, George J. and Burdick, Joel W.}, booktitle = {Proceedings of The 5th Annual Learning for Dynamics and Control Conference}, pages = {300--314}, year = {2023}, editor = {Matni, Nikolai and Morari, Manfred and Pappas, George J.}, volume = {211}, series = {Proceedings of Machine Learning Research}, month = {15--16 Jun}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v211/dixit23a/dixit23a.pdf}, url = {https://proceedings.mlr.press/v211/dixit23a.html}, abstract = {This paper proposes an algorithm for motion planning among dynamic agents using adaptive conformal prediction. We consider a deterministic control system and use trajectory predictors to predict the dynamic agents’ future motion, which is assumed to follow an unknown distribution. We then leverage ideas from adaptive conformal prediction to dynamically quantify prediction uncertainty from an online data stream. Particularly, we provide an online algorithm that uses delayed agent observations to obtain uncertainty sets for multistep-ahead predictions with probabilistic coverage. These uncertainty sets are used within a model predictive controller to safely navigate among dynamic agents. While most existing data-driven prediction approaches quantify prediction uncertainty heuristically, we quantify the true prediction uncertainty in a distribution-free, adaptive manner that even allows to capture changes in prediction quality and the agents’ motion. We empirically evaluate our algorithm on a case study where a drone avoids a flying frisbee.} }
Endnote
%0 Conference Paper %T Adaptive Conformal Prediction for Motion Planning among Dynamic Agents %A Anushri Dixit %A Lars Lindemann %A Skylar X Wei %A Matthew Cleaveland %A George J. Pappas %A Joel W. Burdick %B Proceedings of The 5th Annual Learning for Dynamics and Control Conference %C Proceedings of Machine Learning Research %D 2023 %E Nikolai Matni %E Manfred Morari %E George J. Pappas %F pmlr-v211-dixit23a %I PMLR %P 300--314 %U https://proceedings.mlr.press/v211/dixit23a.html %V 211 %X This paper proposes an algorithm for motion planning among dynamic agents using adaptive conformal prediction. We consider a deterministic control system and use trajectory predictors to predict the dynamic agents’ future motion, which is assumed to follow an unknown distribution. We then leverage ideas from adaptive conformal prediction to dynamically quantify prediction uncertainty from an online data stream. Particularly, we provide an online algorithm that uses delayed agent observations to obtain uncertainty sets for multistep-ahead predictions with probabilistic coverage. These uncertainty sets are used within a model predictive controller to safely navigate among dynamic agents. While most existing data-driven prediction approaches quantify prediction uncertainty heuristically, we quantify the true prediction uncertainty in a distribution-free, adaptive manner that even allows to capture changes in prediction quality and the agents’ motion. We empirically evaluate our algorithm on a case study where a drone avoids a flying frisbee.
APA
Dixit, A., Lindemann, L., Wei, S.X., Cleaveland, M., Pappas, G.J. & Burdick, J.W.. (2023). Adaptive Conformal Prediction for Motion Planning among Dynamic Agents. Proceedings of The 5th Annual Learning for Dynamics and Control Conference, in Proceedings of Machine Learning Research 211:300-314 Available from https://proceedings.mlr.press/v211/dixit23a.html.

Related Material