Multi-step Greedy Reinforcement Learning Algorithms

TL;DR

This paper introduces multi-step greedy algorithms for model-free reinforcement learning, demonstrating their ability to improve performance over standard methods like DQN and TRPO across Atari and MuJoCo benchmarks.

Contribution

It develops a general framework for multi-step greedy RL algorithms using surrogate problems, applicable with various existing RL methods, and provides insights on hyper-parameter tuning.

Findings

Algorithms outperform DQN and TRPO on benchmarks

Multi-step approach improves policy quality

Hyper-parameter tuning is crucial for performance

Abstract

Multi-step greedy policies have been extensively used in model-based reinforcement learning (RL), both when a model of the environment is available (e.g.,~in the game of Go) and when it is learned. In this paper, we explore their benefits in model-free RL, when employed using multi-step dynamic programming algorithms: $\kappa$-Policy Iteration ($\kappa$-PI) and $\kappa$-Value Iteration ($\kappa$-VI). These methods iteratively compute the next policy ($\kappa$-PI) and value function ($\kappa$-VI) by solving a surrogate decision problem with a shaped reward and a smaller discount factor. We derive model-free RL algorithms based on $\kappa$-PI and $\kappa$-VI in which the surrogate problem can be solved by any discrete or continuous action RL method, such as DQN and TRPO. We identify the importance of a hyper-parameter that controls the extent to which the surrogate problem is solved and suggest a way to set this parameter. When evaluated on a range of Atari and MuJoCo benchmark tasks, our results indicate that for the right range of $\kappa$, our algorithms outperform DQN and TRPO. This shows that our multi-step greedy algorithms are general enough to be applied over any existing RL algorithm and can significantly improve its performance.