Continuous Control Reinforcement Learning on DeepMind Control Suite (test)
917.4Cheetah RunSAC
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SACBackpropagation usage=w/ back-prop2025.10 | 917.4 | — | 319.46 | — | — | — | 980.01 | 895.02 | 980.43 | — | — | — | |
| ARQBackpropagation usage=w/o back-prop2025.10 | 880.61 | — | 516.23 | — | — | — | 973.66 | 771.15 | 976.33 | — | — | — | |
| PERevaluation episodes=10, random seeds=5, inputs=proprioceptive2022.08 | 831.87 | 975.4 | 217.35 | 496.42 | 766.3 | 981.58 | 935.08 | 755.49 | 957.38 | — | — | — | |
| ADBackpropagation usage=w/o back-prop2025.10 | 831.57 | — | 470.95 | — | — | — | 955.93 | 762.51 | 975.3 | — | — | — | |
| TD-MPC2Backpropagation usage=w/ back-prop2025.10 | 808.46 | — | 348.55 | — | — | — | 934.84 | 834.07 | 958.8 | — | — | — | |
| BASELINEevaluation episodes=10, random seeds=5, inputs=proprioceptive2022.08 | 761.89 | 966.68 | 264.75 | 612.67 | 831.92 | 983.06 | 955.08 | 759.13 | 943.67 | — | — | — | |
| ReLoevaluation episodes=10, random seeds=5, inputs=proprioceptive2022.08 | 660.29 | 978.78 | 247.81 | 833.92 | 942.64 | 979.08 | 956.8 | 795.14 | 963.28 | — | — | — | |
| LABERevaluation episodes=10, random seeds=5, inputs=proprioceptive2022.08 | 579.8 | 884.46 | 90.85 | 544.8 | 716.61 | 947.2 | 951.08 | 551.81 | 863.89 | — | — | — | |
| DrQ-Random ViewBackbone=DrQ, Sensory control strategy=Random View, Peripheral observation=false2023.06 | — | — | — | — | — | — | — | — | — | 0.569 | 0.591 | 0.768 | |
| DrQ-Raster ScanningBackbone=DrQ, Sensory control strategy=Raster Scanning, Peripheral observation=false2023.06 | — | — | — | — | — | — | — | — | — | 0.609 | 0.566 | 0.913 | |
| DrQ-Single PolicyBackbone=DrQ, Sensory modeling=Joint modeling, Peripheral observation=false2023.06 | — | — | — | — | — | — | — | — | — | 0.54 | 0.57 | 0.776 | |
| SUGARL w/o Joint LearningBackbone=DrQ, Joint Learning=false, Peripheral observation=false2023.06 | — | — | — | — | — | — | — | — | — | 0.377 | 0.446 | 0.355 | |
| SUGARL-DrQBackbone=DrQ, Peripheral observation=false2023.06 | — | — | — | — | — | — | — | — | — | 0.686 | 0.717 | 1.052 | |
| SUGARL-DrQ w/o PVMBackbone=DrQ, Partial Visual Memory (PVM)=false, Peripheral observation=false2023.06 | — | — | — | — | — | — | — | — | — | 0.672 | 0.62 | 0.93 |