Stitching Sub-Trajectories with Conditional Diffusion Model for Goal-Conditioned Offline RL
This addresses the challenge of long-horizon planning in offline RL for robotics or autonomous systems, though it is incremental as it builds on existing diffusion model approaches.
The paper tackles the problem of learning goal-oriented policies from suboptimal offline data in Goal-Conditioned Reinforcement Learning by proposing SSD, a method that uses a conditional diffusion model to stitch sub-trajectories, achieving state-of-the-art performance on standard benchmarks.
Offline Goal-Conditioned Reinforcement Learning (Offline GCRL) is an important problem in RL that focuses on acquiring diverse goal-oriented skills solely from pre-collected behavior datasets. In this setting, the reward feedback is typically absent except when the goal is achieved, which makes it difficult to learn policies especially from a finite dataset of suboptimal behaviors. In addition, realistic scenarios involve long-horizon planning, which necessitates the extraction of useful skills within sub-trajectories. Recently, the conditional diffusion model has been shown to be a promising approach to generate high-quality long-horizon plans for RL. However, their practicality for the goal-conditioned setting is still limited due to a number of technical assumptions made by the methods. In this paper, we propose SSD (Sub-trajectory Stitching with Diffusion), a model-based offline GCRL method that leverages the conditional diffusion model to address these limitations. In summary, we use the diffusion model that generates future plans conditioned on the target goal and value, with the target value estimated from the goal-relabeled offline dataset. We report state-of-the-art performance in the standard benchmark set of GCRL tasks, and demonstrate the capability to successfully stitch the segments of suboptimal trajectories in the offline data to generate high-quality plans.