Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Rethinking Goal-conditioned Supervised Learning and Its Connection to Offline RL

About

Solving goal-conditioned tasks with sparse rewards using self-supervised learning is promising because of its simplicity and stability over current reinforcement learning (RL) algorithms. A recent work, called Goal-Conditioned Supervised Learning (GCSL), provides a new learning framework by iteratively relabeling and imitating self-generated experiences. In this paper, we revisit the theoretical property of GCSL -- optimizing a lower bound of the goal reaching objective, and extend GCSL as a novel offline goal-conditioned RL algorithm. The proposed method is named Weighted GCSL (WGCSL), in which we introduce an advanced compound weight consisting of three parts (1) discounted weight for goal relabeling, (2) goal-conditioned exponential advantage weight, and (3) best-advantage weight. Theoretically, WGCSL is proved to optimize an equivalent lower bound of the goal-conditioned RL objective and generates monotonically improved policies via an iterated scheme. The monotonic property holds for any behavior policies, and therefore WGCSL can be applied to both online and offline settings. To evaluate algorithms in the offline goal-conditioned RL setting, we provide a benchmark including a range of point and simulated robot domains. Experiments in the introduced benchmark demonstrate that WGCSL can consistently outperform GCSL and existing state-of-the-art offline methods in the fully offline goal-conditioned setting.

Rui Yang, Yiming Lu, Wenzhe Li, Hao Sun, Meng Fang, Yali Du, Xiu Li, Lei Han, Chongjie Zhang• 2022

Related benchmarks

TaskDatasetResultRank
Offline Reinforcement LearningKitchen Partial
Normalized Score75.2
62
Offline Reinforcement LearningD4RL antmaze-umaze (diverse)
Normalized Score55.6
40
Offline Reinforcement Learningantmaze medium-play
Score63.2
35
Offline Reinforcement Learningkitchen mixed
Normalized Score77.8
29
Offline Reinforcement LearningAntmaze umaze
Average Return90.8
24
Offline Reinforcement Learningantmaze medium-diverse
Score46
18
Offline Reinforcement Learningantmaze large-play
Score0.6
18
Offline Goal-Conditioned Reinforcement LearningFetchSlide (offline)
Discounted Return2.73
10
Offline Goal-Conditioned Reinforcement LearningFetchPush (offline)
Discounted Return14.7
10
Offline Goal-Conditioned Reinforcement LearningHandReach (offline)
Discounted Return5.97
10
Showing 10 of 16 rows

Other info

Follow for update