Leveraging Reward Gradients For Reinforcement Learning in Differentiable Physics Simulations

A novel algorithm, Cross Entropy Analytic Policy Gradients (CE-APG), that is able to leverage analytic gradients to outperform state of the art deep reinforcement learning on a certain set of challenging nonlinear control problems.

Data and Resources

Cite this as

Sean Gillen, Katie Byl (2025). Dataset: Leveraging Reward Gradients For Reinforcement Learning in Differentiable Physics Simulations. https://doi.org/10.57702/yfvc8ug7

DOI retrieved: January 3, 2025

Additional Info

Field Value
Created January 3, 2025
Last update January 3, 2025
Defined In https://doi.org/10.48550/arXiv.2203.02857
Author Sean Gillen
More Authors
Katie Byl
Homepage https://github.com/sgillen/apg