Reinforcement Learning on DMControl
995CartPole BalanceNCRL
Evaluation Results
| Method | Links | ||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NCRLEnvironment steps=150k, Observation space=pixel2025.02 | 995 | 84.6 | 12.2 | -17.1 | 974.1 | 960.5 | 707.7 | 266.1 | 887.6 | 842.8 | 366 | 543.8 | 317.6 | 462.3 | 521.6 | 614.2 | 855.6 | 941.4 | 766.9 | 820.2 | 948 | 758.9 | 617.73 | 733.3 | |
| DreamerV3Environment steps=500k, Observation space=pixel2025.02 | 994.3 | 222.1 | 2.5 | -0.2 | 965.7 | 949.2 | 616.6 | 293.6 | 942.9 | -2.1 | 363.8 | 843.7 | 473.8 | 657.4 | 693.8 | 597 | 369.3 | 746 | 328.1 | 689.6 | 663.9 | 508.8 | 541.81 | 606.8 | |
| DrQ-v2Environment steps=500k, Observation space=pixel2025.02 | 992.3 | 30.3 | 1.17 | 0.3 | 947.6 | 797.8 | 299.3 | 96.7 | 744.3 | -9.5 | 246 | 338.1 | 202.4 | 294.4 | 384.3 | 535.6 | 258.1 | 442.2 | 296.5 | 478.3 | 446 | 366.9 | 372.23 | 318.7 | |
| DrQ-v2Environment steps=150k, Observation space=pixel2025.02 | 983.3 | 20.8 | 1.5 | 0.4 | 742.9 | 280.1 | 143 | 91.7 | 470.9 | -17.1 | 167.4 | 251.2 | 108.4 | 171.2 | 335.6 | 251.8 | 76.5 | 318.9 | 102.8 | 190.5 | 126.2 | 164.7 | 226.49 | 166.05 | |
| DreamerV3Environment steps=150k, Observation space=pixel2025.02 | 955.8 | 85.2 | 1.7 | 2 | 946.2 | 808.9 | 224.4 | 128.2 | 625.9 | -4.7 | 229.4 | 621.4 | 143.1 | 407.6 | 626.6 | 200.8 | 145.2 | 227.2 | 183 | 168.3 | 207.9 | 124.8 | 320.86 | 204.35 |