Tracking performance on Spring Joint hardware setting (evaluation)
-2X Axis Error (m)Original Policy
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Original PolicyCommand=x +1m/s2026.04 | -2 | 0.483 | 40 | |
| Finetune w/ Residual Actuator ModelCommand=x +1m/s2026.04 | -0.177 | 0.25 | 0 | |
| Finetune w/ Residual Actuator ModelCommand=y +1m/s2026.04 | -0.11 | 0.21 | 22 | |
| Finetune w/ Residual Actuator ModelCommand=x -1m/s2026.04 | -0.051 | -0.4 | 8 | |
| Original PolicyCommand=yaw -0.3925m/s2026.04 | -0.037 | 0.22 | -11 | |
| Finetune w/ Residual Actuator ModelCommand=yaw +0.3925m/s2026.04 | 0.05 | 0.051 | 0 | |
| Original PolicyCommand=yaw +0.3925m/s2026.04 | 0.083 | 0.121 | 3 | |
| Original PolicyCommand=x -1m/s2026.04 | 0.154 | -2.184 | 6 | |
| Finetune w/ Residual Actuator ModelCommand=yaw -0.3925m/s2026.04 | 0.17 | -0.1 | 5 | |
| Finetune w/ Residual Actuator ModelCommand=y -1m/s2026.04 | 0.2286 | -0.492 | 13 | |
| Original PolicyCommand=y -1m/s2026.04 | 0.279 | -0.898 | -6 |