Skill Decision Transformer
This work addresses the challenge of learning diverse skills from offline data for reinforcement learning practitioners, though it builds incrementally on prior methods like Generalized Decision Transformers.
The paper tackles the problem of extracting diverse behaviors from offline trajectory data in reinforcement learning by proposing Skill Decision Transformer, which discovers primitive skills and performs state-marginal matching, achieving competitive performance with reward-free optimization on the D4RL benchmark.
Recent work has shown that Large Language Models (LLMs) can be incredibly effective for offline reinforcement learning (RL) by representing the traditional RL problem as a sequence modelling problem (Chen et al., 2021; Janner et al., 2021). However many of these methods only optimize for high returns, and may not extract much information from a diverse dataset of trajectories. Generalized Decision Transformers (GDTs) (Furuta et al., 2021) have shown that utilizing future trajectory information, in the form of information statistics, can help extract more information from offline trajectory data. Building upon this, we propose Skill Decision Transformer (Skill DT). Skill DT draws inspiration from hindsight relabelling (Andrychowicz et al., 2017) and skill discovery methods to discover a diverse set of primitive behaviors, or skills. We show that Skill DT can not only perform offline state-marginal matching (SMM), but can discovery descriptive behaviors that can be easily sampled. Furthermore, we show that through purely reward-free optimization, Skill DT is still competitive with supervised offline RL approaches on the D4RL benchmark. The code and videos can be found on our project page: https://github.com/shyamsn97/skill-dt