Robotic Manipulation on Adroit Door (DTW and Success Metrics)
0.87DTW Score (DTWs)HiMAQ+IQL
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HiMAQ+IQLBaseline Algorithm=IQL2026.05 | 0.87 | 0.96 | 0.77 | 0.82 | 95 | |
| MAQ+IQLBaseline Algorithm=IQL2026.05 | 0.84 | 0.95 | 0.75 | 0.81 | 93 | |
| HiMAQ+SACBaseline Algorithm=SAC, Macro Action Length (H)=82026.05 | 0.83 | 0.92 | 0.75 | 0.8 | 50 | |
| MAQ+SACBaseline Algorithm=SAC, Macro Action Length (H)=82026.05 | 0.8 | 0.91 | 0.71 | 0.77 | 56 | |
| IQLBaseline Algorithm=IQL2026.05 | 0.43 | 0.61 | 0.48 | 0.5 | 16 | |
| BCBaseline Algorithm=Behavioral Cloning, Macro Action Length (H)=12026.05 | 0.18 | 0.42 | 0.32 | 0.41 | 2 | |
| BCBaseline Algorithm=BC2026.05 | 0.18 | 0.42 | 0.32 | 0.41 | 2 | |
| SACBaseline Algorithm=SAC, Macro Action Length (H)=12026.05 | -0.39 | -0.25 | -0.28 | -0.15 | 43 |