Locally Persistent Exploration in Continuous Control Tasks with Sparse Rewards

TL;DR

This paper introduces a novel exploration strategy for reinforcement learning in sparse reward environments, utilizing trajectory-dependent actions and statistical physics concepts to generate persistent, self-avoiding trajectories, improving exploration efficiency.

Contribution

The paper proposes a new exploration method based on locally self-avoiding trajectories that depend on the agent's history, enhancing exploration in continuous control tasks with sparse rewards.

Findings

Effective in 2D navigation tasks

Improves exploration in MuJoCo locomotion tasks

Provides theoretical insights into trajectory properties

Abstract

A major challenge in reinforcement learning is the design of exploration strategies, especially for environments with sparse reward structures and continuous state and action spaces. Intuitively, if the reinforcement signal is very scarce, the agent should rely on some form of short-term memory in order to cover its environment efficiently. We propose a new exploration method, based on two intuitions: (1) the choice of the next exploratory action should depend not only on the (Markovian) state of the environment, but also on the agent's trajectory so far, and (2) the agent should utilize a measure of spread in the state space to avoid getting stuck in a small region. Our method leverages concepts often used in statistical physics to provide explanations for the behavior of simplified (polymer) chains in order to generate persistent (locally self-avoiding) trajectories in state space. We discuss the theoretical properties of locally self-avoiding walks and their ability to provide a kind of short-term memory through a decaying temporal correlation within the trajectory. We provide empirical evaluations of our approach in a simulated 2D navigation task, as well as higher-dimensional MuJoCo continuous control locomotion tasks with sparse rewards.