Planning to Explore via Self-Supervised World Models
This work addresses sample efficiency and task generalization in reinforcement learning, representing an incremental improvement over existing self-supervised exploration methods.
The paper tackles the challenges of task-specific learning and sample inefficiency in reinforcement learning by introducing Plan2Explore, a self-supervised agent that uses planning to seek out expected future novelty during exploration and adapts quickly to new tasks. It outperforms prior self-supervised methods on high-dimensional image control tasks, nearly matching the performance of an oracle with access to rewards.
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/