Leave no Trace: Learning to Reset for Safe and Autonomous Reinforcement Learning

TL;DR

This paper introduces an autonomous approach for reinforcement learning that learns both task-performing and environment-reset policies, reducing manual resets and enhancing safety in real-world applications.

Contribution

It proposes a joint learning framework for forward and reset policies with uncertainty-aware safety, enabling autonomous resets and safer exploration in reinforcement learning.

Findings

Reduces manual resets in learning tasks

Decreases unsafe actions leading to non-reversible states

Automatically induces a curriculum for learning

Abstract

Deep reinforcement learning algorithms can learn complex behavioral skills, but real-world application of these methods requires a large amount of experience to be collected by the agent. In practical settings, such as robotics, this involves repeatedly attempting a task, resetting the environment between each attempt. However, not all tasks are easily or automatically reversible. In practice, this learning process requires extensive human intervention. In this work, we propose an autonomous method for safe and efficient reinforcement learning that simultaneously learns a forward and reset policy, with the reset policy resetting the environment for a subsequent attempt. By learning a value function for the reset policy, we can automatically determine when the forward policy is about to enter a non-reversible state, providing for uncertainty-aware safety aborts. Our experiments illustrate that proper use of the reset policy can greatly reduce the number of manual resets required to learn a task, can reduce the number of unsafe actions that lead to non-reversible states, and can automatically induce a curriculum.