Robot Navigation and Treasure Collection on Treasure-Collecting Robot Final-stage 2.7M-3.0M steps
100Goal Reach RateBaseline policy
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Baseline policy2026.06 | 100 | 17.3 | 58.65 | |
| Agency-Transferring Model-Free Policy Enhancement TechniqueBackbone=TD32026.06 | 94.11 | 99.78 | 96.95 | |
| Agency-Transferring Model-Free Policy Enhancement TechniqueBackbone=SAC2026.06 | 85.05 | 99.56 | 92.31 | |
| Residual TD3Backbone=TD32026.06 | 76.15 | 99.27 | 87.71 | |
| Residual SACBackbone=SAC2026.06 | 66.62 | 98.86 | 82.74 | |
| TD3Backbone=TD32026.06 | 31.69 | 99.89 | 65.79 | |
| SACBackbone=SAC2026.06 | 11.04 | 99.78 | 55.41 |