Elastic Decision Transformer
This addresses a key limitation in offline reinforcement learning for robotics and gaming, though it is an incremental improvement over existing Decision Transformer variants.
The paper tackles the problem of trajectory stitching in Decision Transformers, which struggle to generate optimal trajectories from sub-optimal parts, by introducing Elastic Decision Transformer (EDT) that adjusts history length during inference to enable stitching. The result shows EDT outperforms Q Learning-based methods in multi-task regimes on the D4RL locomotion benchmark and Atari games.
This paper introduces Elastic Decision Transformer (EDT), a significant advancement over the existing Decision Transformer (DT) and its variants. Although DT purports to generate an optimal trajectory, empirical evidence suggests it struggles with trajectory stitching, a process involving the generation of an optimal or near-optimal trajectory from the best parts of a set of sub-optimal trajectories. The proposed EDT differentiates itself by facilitating trajectory stitching during action inference at test time, achieved by adjusting the history length maintained in DT. Further, the EDT optimizes the trajectory by retaining a longer history when the previous trajectory is optimal and a shorter one when it is sub-optimal, enabling it to "stitch" with a more optimal trajectory. Extensive experimentation demonstrates EDT's ability to bridge the performance gap between DT-based and Q Learning-based approaches. In particular, the EDT outperforms Q Learning-based methods in a multi-task regime on the D4RL locomotion benchmark and Atari games. Videos are available at: https://kristery.github.io/edt/