Continuous Control on DeepMind Control 1.5M steps
272Acrobot Swingup ScoreDrQv2
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| DrQv2Observation type=pixel, Random seeds=5, Test rollouts=100, Evaluation window (training steps)=150K, Training steps=1.5M2022.10 | 272 | 478 | 781 | 757 | 506 | 200 | 402 | 591 | 219 | 973 | 802 | 568 | 545.72 | 537.37 | |
| SSPGObservation type=pixel, Random seeds=5, Test rollouts=100, Evaluation window (training steps)=150K, Training steps=1.5M2022.10 | 218 | 797 | 755 | 794 | 637 | 246 | 570 | 855 | 221 | 978 | 913 | 634 | 634.79 | 695.85 | |
| DrQObservation type=pixel, Random seeds=5, Test rollouts=100, Evaluation window (training steps)=150K, Training steps=1.5M2022.10 | 24 | 321 | 777 | 184 | 89 | 272 | 135 | 147 | 9 | 587 | 343 | 477 | 280.34 | 228.13 | |
| SACObservation type=pixel, Random seeds=5, Test rollouts=100, Evaluation window (training steps)=150K, Training steps=1.5M2022.10 | 8 | 135 | 8 | 170 | 79 | 0 | 61 | 68 | 0 | 64 | 7 | 26 | 52.28 | 43.49 | |
| CURLObservation type=pixel, Random seeds=5, Test rollouts=100, Evaluation window (training steps)=150K, Training steps=1.5M2022.10 | 6 | 497 | 520 | 305 | 224 | 185 | 182 | 126 | 10 | 713 | 482 | 378 | 302.2 | 264.37 |