Continuous Control on HalfCheetah (Robustness Perturbations)
1,834Score (p_act=0.1)Self-Paced
Evaluation Results
| Method | Links | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Self-PacedTraining steps=1 million, Evaluation episodes=1002025.11 | 1,834 | 1,182.1 | 800.6 | 507.7 | 266.3 | 2,184.5 | 1,570.3 | 1,131.1 | 761.4 | 545.5 | 2,338.5 | 2,017.2 | 1,601.8 | 1,215.1 | 935.4 | |
| Dom. Rand.Training steps=1 million, Evaluation episodes=1002025.11 | 1,243.1 | 843.1 | 527.6 | 297.7 | 114.1 | 1,396.7 | 997 | 579.9 | 395.8 | 242.3 | 1,717 | 1,385.8 | 1,082.6 | 722.5 | 448.4 | |
| PPOTraining steps=1 million, Evaluation episodes=1002025.11 | 1,168 | 869.3 | 633.1 | 424.3 | 249.5 | 1,279.6 | 882.5 | 566 | 337.9 | 175 | 1,467.6 | 1,222.8 | 855.7 | 602.3 | 385.5 | |
| FixedTraining steps=1 million, Evaluation episodes=1002025.11 | 1,158.1 | 760.7 | 513 | 246.5 | 105.3 | 1,040.5 | 633.7 | 321.3 | 162 | 34 | 1,516.9 | 1,262.7 | 997.7 | 709.4 | 477.9 | |
| ACCELTraining steps=1 million, Evaluation episodes=1002025.11 | 1,110.5 | 782 | 520.9 | 319.4 | 160 | 1,263 | 867.6 | 509.1 | 279 | 117.6 | 1,448.3 | 1,213.8 | 903.7 | 636.7 | 347.4 | |
| SPACETraining steps=1 million, Evaluation episodes=1002025.11 | 1,005.8 | 759.7 | 541.9 | 349.4 | 187 | 960.8 | 541.6 | 235.1 | 52.3 | 53 | 1,218.6 | 979.4 | 665.4 | 366.9 | 180 | |
| LinearTraining steps=1 million, Evaluation episodes=1002025.11 | 985.2 | 753.2 | 535.7 | 356.9 | 201.1 | 945.9 | 522.9 | 218.1 | 27.5 | 82 | 1,167.4 | 921.4 | 617.4 | 367.8 | 133.7 |