Planning to Explore via Self-Supervised World Models

TL;DR

Plan2Explore introduces a self-supervised reinforcement learning approach that uses planning to efficiently explore and quickly adapt to new tasks without prior task knowledge, outperforming previous methods on high-dimensional control tasks.

Contribution

The paper proposes a novel planning-based exploration method enabling fast adaptation to multiple tasks without task-specific training or supervision.

Findings

Outperforms prior self-supervised exploration methods.

Almost matches the performance of an oracle with access to rewards.

Effective on high-dimensional image-based control tasks.

Abstract

Reinforcement learning allows solving complex tasks, however, the learning tends to be task-specific and the sample efficiency remains a challenge. We present Plan2Explore, a self-supervised reinforcement learning agent that tackles both these challenges through a new approach to self-supervised exploration and fast adaptation to new tasks, which need not be known during exploration. During exploration, unlike prior methods which retrospectively compute the novelty of observations after the agent has already reached them, our agent acts efficiently by leveraging planning to seek out expected future novelty. After exploration, the agent quickly adapts to multiple downstream tasks in a zero or a few-shot manner. We evaluate on challenging control tasks from high-dimensional image inputs. Without any training supervision or task-specific interaction, Plan2Explore outperforms prior self-supervised exploration methods, and in fact, almost matches the performances oracle which has access to rewards. Videos and code at https://ramanans1.github.io/plan2explore/