Worst-case time-constrained reinforcement learning on HalfCheetah MuJoCo (test)
2.76Normalized Worst-Case RewardOracle-TC-M2TD3
Evaluation Results
| Method | Links | |
|---|---|---|
| Oracle-TC-M2TD3Oracle=true, Framework=TC2024.06 | 2.76 | |
| Oracle-TC-RARLOracle=true, Framework=TC2024.06 | 2.63 | |
| TC-M2TD3Framework=TC2024.06 | 2.42 | |
| Stacked-TC-M2TD3Architecture=Stacked, Framework=TC2024.06 | 2.4 | |
| Stacked-TC-RARLArchitecture=Stacked, Framework=TC2024.06 | 1.76 | |
| DRFramework=Domain Randomization2024.06 | 1.59 | |
| TC-RARLFramework=TC2024.06 | 1.54 | |
| M2TD3Framework=Robust2024.06 | 1 | |
| Oracle M2TD3Oracle=true2024.06 | 0.95 | |
| TD3Framework=Vanilla2024.06 | 0 | |
| RARLFramework=Robust2024.06 | -0.61 | |
| Oracle RARLOracle=true2024.06 | -0.71 |