Waypoint Transformer: Reinforcement Learning via Supervised Learning with Intermediate Targets

TL;DR

The paper introduces the Waypoint Transformer, an enhanced offline reinforcement learning method that incorporates intermediate targets to improve performance and stability across challenging environments.

Contribution

It presents the Waypoint Transformer architecture, building on decision transformers, with automatic waypoint conditioning to better connect suboptimal trajectories.

Findings

Significant increase in final return over existing RvS methods

Performance on par or better than state-of-the-art temporal difference methods

Largest improvements in challenging environments and data configurations

Abstract

Despite the recent advancements in offline reinforcement learning via supervised learning (RvS) and the success of the decision transformer (DT) architecture in various domains, DTs have fallen short in several challenging benchmarks. The root cause of this underperformance lies in their inability to seamlessly connect segments of suboptimal trajectories. To overcome this limitation, we present a novel approach to enhance RvS methods by integrating intermediate targets. We introduce the Waypoint Transformer (WT), using an architecture that builds upon the DT framework and conditioned on automatically-generated waypoints. The results show a significant increase in the final return compared to existing RvS methods, with performance on par or greater than existing state-of-the-art temporal difference learning-based methods. Additionally, the performance and stability improvements are largest in the most challenging environments and data configurations, including AntMaze Large Play/Diverse and Kitchen Mixed/Partial.