A Policy Efficient Reduction Approach to Convex Constrained Deep Reinforcement Learning

Abstract

Although well-established in general reinforcement learning (RL), value-based methods are rarely explored in constrained RL (CRL) for their incapability of finding policies that can randomize among multiple actions. To apply value-based methods to CRL, a recent groundbreaking line of game-theoretic approaches uses the mixed policy that randomizes among a set of carefully generated policies to converge to the desired constraint-satisfying policy. However, these approaches require storing a large set of policies, which is not policy efficient, and may incur prohibitive memory costs in constrained deep RL. To address this problem, we propose an alternative approach. Our approach first reformulates the CRL to an equivalent distance optimization problem. With a specially designed linear optimization oracle, we derive a meta-algorithm that solves it using any off-the-shelf RL algorithm and any conditional gradient (CG) type algorithm as subroutines. We then propose a new variant of the CG-type algorithm, which generalizes the minimum norm point (MNP) method. The proposed method matches the convergence rate of the existing game-theoretic approaches and achieves the worst-case optimal policy efficiency. The experiments on a navigation task show that our method reduces the memory costs by an order of magnitude, and meanwhile achieves better performance, demonstrating both its effectiveness and efficiency.

Reinforcement Learning for Real Life (RL4RealLife) Workshop in the 38th International Conference on Machine Learning, 2021

References in corpus (14)

• Language Models are Few-Shot Learners
• Addressing Function Approximation Error in Actor-Critic Methods
• Dueling Network Architectures for Deep Reinforcement Learning
• Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems
• Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity
• Safe, Multi-Agent, Reinforcement Learning for Autonomous Driving
• Distributed Distributional Deterministic Policy Gradients
• Doubly Robust Policy Evaluation and Learning
• Reward Constrained Policy Optimization
• Benchmarking Batch Deep Reinforcement Learning Algorithms
• Diversity-Driven Exploration Strategy for Deep Reinforcement Learning
• Algorithms for CVaR Optimization in MDPs
• Reinforcement Learning with Convex Constraints
• Provable Submodular Minimization using Wolfe's Algorithm