Better Exploration with Optimistic Actor-Critic

Abstract

Actor-critic methods, a type of model-free Reinforcement Learning, have been successfully applied to challenging tasks in continuous control, often achieving state-of-the art performance. However, wide-scale adoption of these methods in real-world domains is made difficult by their poor sample efficiency. We address this problem both theoretically and empirically. On the theoretical side, we identify two phenomena preventing efficient exploration in existing state-of-the-art algorithms such as Soft Actor Critic. First, combining a greedy actor update with a pessimistic estimate of the critic leads to the avoidance of actions that the agent does not know about, a phenomenon we call pessimistic underexploration. Second, current algorithms are directionally uninformed, sampling actions with equal probability in opposite directions from the current mean. This is wasteful, since we typically need actions taken along certain directions much more than others. To address both of these phenomena, we introduce a new algorithm, Optimistic Actor Critic, which approximates a lower and upper confidence bound on the state-action value function. This allows us to apply the principle of optimism in the face of uncertainty to perform directed exploration using the upper bound while still using the lower bound to avoid overestimation. We evaluate OAC in several challenging continuous control tasks, achieving state-of the art sample efficiency.

20 pages (including supplement)

Cited by in corpus (23)

• Uncertainty-Based Offline Reinforcement Learning with Diversified Q-Ensemble
• Softmax Deep Double Deterministic Policy Gradients
• Effective Diversity in Population Based Reinforcement Learning
• ReLMoGen: Leveraging Motion Generation in Reinforcement Learning for Mobile Manipulation
• Variational Dynamic for Self-Supervised Exploration in Deep Reinforcement Learning
• Local policy search with Bayesian optimization
• On the Estimation Bias in Double Q-Learning
• Tactical Optimism and Pessimism for Deep Reinforcement Learning
• Principled Exploration via Optimistic Bootstrapping and Backward Induction
• Locally Persistent Exploration in Continuous Control Tasks with Sparse Rewards
• A Survey of Exploration Methods in Reinforcement Learning
• Non-local Policy Optimization via Diversity-regularized Collaborative Exploration
• Iterative Amortized Policy Optimization
• Machine Learning for Robotic Manipulation
• Novel Policy Seeking with Constrained Optimization
• Zeroth-Order Supervised Policy Improvement
• OPAC: Opportunistic Actor-Critic
• ADER:Adapting between Exploration and Robustness for Actor-Critic Methods
• Learning Pessimism for Robust and Efficient Off-Policy Reinforcement Learning
• Dr Jekyll and Mr Hyde: the Strange Case of Off-Policy Policy Updates
• Towards Tractable Optimism in Model-Based Reinforcement Learning
• Bayesian Distributional Policy Gradients
• Uncertainty-aware Low-Rank Q-Matrix Estimation for Deep Reinforcement Learning