Energy-based Potential Games for Joint Motion Forecasting and Control

Christopher Diehl, Tobias Klosek, Martin Krueger, Nils Murzyn, Timo Osterburg, Torsten Bertram
Proceedings of The 7th Conference on Robot Learning, PMLR 229:3112-3141, 2023.

Abstract

This work uses game theory as a mathematical framework to address interaction modeling in multi-agent motion forecasting and control. Despite its interpretability, applying game theory to real-world robotics, like automated driving, faces challenges such as unknown game parameters. To tackle these, we establish a connection between differential games, optimal control, and energy-based models, demonstrating how existing approaches can be unified under our proposed Energy-based Potential Game formulation. Building upon this, we introduce a new end-to-end learning application that combines neural networks for game-parameter inference with a differentiable game-theoretic optimization layer, acting as an inductive bias. The analysis provides empirical evidence that the game-theoretic layer adds interpretability and improves the predictive performance of various neural network backbones using two simulations and two real-world driving datasets.

Cite this Paper

BibTeX
@InProceedings{pmlr-v229-diehl23a, title = {Energy-based Potential Games for Joint Motion Forecasting and Control}, author = {Diehl, Christopher and Klosek, Tobias and Krueger, Martin and Murzyn, Nils and Osterburg, Timo and Bertram, Torsten}, booktitle = {Proceedings of The 7th Conference on Robot Learning}, pages = {3112--3141}, year = {2023}, editor = {Tan, Jie and Toussaint, Marc and Darvish, Kourosh}, volume = {229}, series = {Proceedings of Machine Learning Research}, month = {06--09 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v229/diehl23a/diehl23a.pdf}, url = {https://proceedings.mlr.press/v229/diehl23a.html}, abstract = {This work uses game theory as a mathematical framework to address interaction modeling in multi-agent motion forecasting and control. Despite its interpretability, applying game theory to real-world robotics, like automated driving, faces challenges such as unknown game parameters. To tackle these, we establish a connection between differential games, optimal control, and energy-based models, demonstrating how existing approaches can be unified under our proposed Energy-based Potential Game formulation. Building upon this, we introduce a new end-to-end learning application that combines neural networks for game-parameter inference with a differentiable game-theoretic optimization layer, acting as an inductive bias. The analysis provides empirical evidence that the game-theoretic layer adds interpretability and improves the predictive performance of various neural network backbones using two simulations and two real-world driving datasets.} }
Endnote
%0 Conference Paper %T Energy-based Potential Games for Joint Motion Forecasting and Control %A Christopher Diehl %A Tobias Klosek %A Martin Krueger %A Nils Murzyn %A Timo Osterburg %A Torsten Bertram %B Proceedings of The 7th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2023 %E Jie Tan %E Marc Toussaint %E Kourosh Darvish %F pmlr-v229-diehl23a %I PMLR %P 3112--3141 %U https://proceedings.mlr.press/v229/diehl23a.html %V 229 %X This work uses game theory as a mathematical framework to address interaction modeling in multi-agent motion forecasting and control. Despite its interpretability, applying game theory to real-world robotics, like automated driving, faces challenges such as unknown game parameters. To tackle these, we establish a connection between differential games, optimal control, and energy-based models, demonstrating how existing approaches can be unified under our proposed Energy-based Potential Game formulation. Building upon this, we introduce a new end-to-end learning application that combines neural networks for game-parameter inference with a differentiable game-theoretic optimization layer, acting as an inductive bias. The analysis provides empirical evidence that the game-theoretic layer adds interpretability and improves the predictive performance of various neural network backbones using two simulations and two real-world driving datasets.
APA
Diehl, C., Klosek, T., Krueger, M., Murzyn, N., Osterburg, T. & Bertram, T.. (2023). Energy-based Potential Games for Joint Motion Forecasting and Control. Proceedings of The 7th Conference on Robot Learning, in Proceedings of Machine Learning Research 229:3112-3141 Available from https://proceedings.mlr.press/v229/diehl23a.html.

Related Material