Softmax Deep Double Deterministic Policy Gradients

Abstract

A widely-used actor-critic reinforcement learning algorithm for continuous control, Deep Deterministic Policy Gradients (DDPG), suffers from the overestimation problem, which can negatively affect the performance. Although the state-of-the-art Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm mitigates the overestimation issue, it can lead to a large underestimation bias. In this paper, we propose to use the Boltzmann softmax operator for value function estimation in continuous control. We first theoretically analyze the softmax operator in continuous action space. Then, we uncover an important property of the softmax operator in actor-critic algorithms, i.e., it helps to smooth the optimization landscape, which sheds new light on the benefits of the operator. We also design two new algorithms, Softmax Deep Deterministic Policy Gradients (SD2) and Softmax Deep Double Deterministic Policy Gradients (SD3), by building the softmax operator upon single and double estimators, which can effectively improve the overestimation and underestimation bias. We conduct extensive experiments on challenging continuous control tasks, and results show that SD3 outperforms state-of-the-art methods.

arXiv commentNeurIPS 2020

References in corpus (13)

• Continuous control with deep reinforcement learning
• Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor
• Addressing Function Approximation Error in Actor-Critic Methods
• Distributed Prioritized Experience Replay
• Identifying and attacking the saddle point problem in high-dimensional non-convex optimization
• Distributed Distributional Deterministic Policy Gradients
• Qualitatively characterizing neural network optimization problems
• Averaged-DQN: Variance Reduction and Stabilization for Deep Reinforcement Learning
• Model-Based Value Estimation for Efficient Model-Free Reinforcement Learning
• Better Exploration with Optimistic Actor-Critic
• Revisiting the Softmax Bellman Operator: New Benefits and New Perspective
• Boltzmann Exploration Done Right
• Smoothed Action Value Functions for Learning Gaussian Policies

Cited by in corpus (4)

• Regularized Softmax Deep Multi-Agent $Q$-Learning
• Future Impact Decomposition in Request-level Recommendations
• PAC-Bayesian Soft Actor-Critic Learning
• Value Function Decomposition in Markov Recommendation Process