Exploiting Symmetry in Dynamics for Model-Based Reinforcement Learning with Asymmetric Rewards

arXiv:2403.19024v1 Announce Type: new
Abstract: Recent work in reinforcement learning has leveraged symmetries in the model to improve sample efficiency in training a policy. A commonly used simplifying assumption is that the dynamics and reward both exhibit the same symmetry. However, in many real-world environments, the dynamical model exhibits symmetry independent of the reward model: the reward may not satisfy the same symmetries as the dynamics. In this paper, we investigate scenarios where only the dynamics are assumed to exhibit …

abstract arxiv cs.ai cs.lg cs.ro cs.sy dynamics eess.sy efficiency environments however policy reinforcement reinforcement learning sample simplifying symmetry training type work world