notesum.ai

Published at December 5

Action Mapping for Reinforcement Learning in Continuous Environments with Constraints

cs.LG

cs.AI

cs.SY

eess.SY

Released Date: December 5, 2024

Authors: Mirco Theile¹, Lukas Dirnberger, Raphael Trumpp, Marco Caccamo, Alberto L. Sangiovanni-Vincentelli²

Aff.: ¹TUM School of Engineering and Design, Technical University of Munich, Germany; ²Department of Electrical and Computer Engineering, University of California, Berkeley, USA

Arxiv: http://arxiv.org/pdf/2412.04327v1

Parameter	AM-SAC	AM-PPO
Discount factor cost $\gamma_{C}$	$0.9$	$0$ (Not needed)
Safety delta $\delta_{C}$	$0.05$	-
Safety critic learning rate	$1\times 10^{-4}$	$1\times 10^{-4}$
Lagrangian mult. learning rate	$0.01$	$0.01$