Improving Stochastic Action-Constrained Reinforcement Learning via Truncated Distributions
This addresses computational inefficiencies for researchers and practitioners in RL, but it is incremental as it builds on existing truncated distribution methods.
The paper tackled the problem of accurately computing key characteristics like entropy and log-probability for truncated normal distributions in action-constrained reinforcement learning, which prior approximations degraded performance, and demonstrated significant improvements in benchmark environments.
In reinforcement learning (RL), it is often advantageous to consider additional constraints on the action space to ensure safety or action relevance. Existing work on such action-constrained RL faces challenges regarding effective policy updates, computational efficiency, and predictable runtime. Recent work proposes to use truncated normal distributions for stochastic policy gradient methods. However, the computation of key characteristics, such as the entropy, log-probability, and their gradients, becomes intractable under complex constraints. Hence, prior work approximates these using the non-truncated distributions, which severely degrades performance. We argue that accurate estimation of these characteristics is crucial in the action-constrained RL setting, and propose efficient numerical approximations for them. We also provide an efficient sampling strategy for truncated policy distributions and validate our approach on three benchmark environments, which demonstrate significant performance improvements when using accurate estimations.