CLUE: Calibrated Latent Guidance for Offline Reinforcement Learning
This addresses the challenge of reducing manual reward specification in offline RL for researchers and practitioners, though it appears incremental as it builds on existing methods like conditional variational auto-encoders.
The paper tackles the problem of labor-intensive reward labeling in offline reinforcement learning by proposing CLUE, which uses a few expert data to drive intrinsic rewards, eliminating the need for extrinsic rewards. The result shows that CLUE improves sparse-reward offline RL performance, outperforms state-of-the-art offline imitation learning baselines, and discovers diverse skills from static reward-free offline data.
Offline reinforcement learning (RL) aims to learn an optimal policy from pre-collected and labeled datasets, which eliminates the time-consuming data collection in online RL. However, offline RL still bears a large burden of specifying/handcrafting extrinsic rewards for each transition in the offline data. As a remedy for the labor-intensive labeling, we propose to endow offline RL tasks with a few expert data and utilize the limited expert data to drive intrinsic rewards, thus eliminating the need for extrinsic rewards. To achieve that, we introduce \textbf{C}alibrated \textbf{L}atent g\textbf{U}idanc\textbf{E} (CLUE), which utilizes a conditional variational auto-encoder to learn a latent space such that intrinsic rewards can be directly qualified over the latent space. CLUE's key idea is to align the intrinsic rewards consistent with the expert intention via enforcing the embeddings of expert data to a calibrated contextual representation. We instantiate the expert-driven intrinsic rewards in sparse-reward offline RL tasks, offline imitation learning (IL) tasks, and unsupervised offline RL tasks. Empirically, we find that CLUE can effectively improve the sparse-reward offline RL performance, outperform the state-of-the-art offline IL baselines, and discover diverse skills from static reward-free offline data.