Waypoint Transformer: Reinforcement Learning via Supervised Learning with Intermediate Targets
This work addresses a key limitation in offline reinforcement learning for AI systems, offering improved stability and performance in complex environments, though it is incremental as it builds upon existing transformer-based methods.
The paper tackled the underperformance of decision transformers in offline reinforcement learning by introducing the Waypoint Transformer, which integrates intermediate targets to connect suboptimal trajectory segments, resulting in significant performance gains on challenging benchmarks like AntMaze Large Play/Diverse and Kitchen Mixed/Partial.
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.