Offline Reinforcement Learning on Hopper v2 (offline)
15Observation ErrorUWMSG
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| UWMSGCorruption type=Adversarial, Corruption rate (c)=0.3, Corruption scale (epsilon)=1.02024.11 | 15 | 27.7 | 30.3 | 0.7 | |
| MSGCorruption type=Adversarial, Corruption rate (c)=0.3, Corruption scale (epsilon)=1.02024.11 | 16 | 23 | 22.6 | 0.6 | |
| BCCorruption type=Adversarial, Corruption rate (c)=0.3, Corruption scale (epsilon)=1.02024.11 | 21.6 | 15.5 | 19.5 | 19.5 | |
| IQLCorruption type=Adversarial, Corruption rate (c)=0.3, Corruption scale (epsilon)=1.02024.11 | 32.8 | 37.9 | 57.3 | 1.3 | |
| EDACCorruption type=Adversarial, Corruption rate (c)=0.3, Corruption scale (epsilon)=1.02024.11 | 36.2 | 25.7 | 21.2 | 0.6 | |
| RIQLCorruption type=Adversarial, Corruption rate (c)=0.3, Corruption scale (epsilon)=1.02024.11 | 50.8 | 63.6 | 65.8 | 65.7 | |
| TRACERCorruption type=Adversarial, Corruption rate (c)=0.3, Corruption scale (epsilon)=1.02024.11 | 64.5 | 67.2 | 64.3 | 61.1 | |
| CQLCorruption type=Adversarial, Corruption rate (c)=0.3, Corruption scale (epsilon)=1.02024.11 | 78 | 32.2 | 49.6 | 0.6 |