Locally Persistent Exploration in Continuous Control Tasks with Sparse Rewards
This paper tackles the problem of efficient exploration in sparse reward continuous control environments, which is a common challenge for reinforcement learning agents.
This paper addresses the challenge of exploration in continuous control tasks with sparse rewards by proposing a new exploration method. The method generates persistent, locally self-avoiding trajectories in state space, drawing inspiration from statistical physics, and is evaluated in a 2D navigation task and higher-dimensional MuJoCo locomotion tasks.
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.