Tracking on Bipedal Walking hardware
-0.15X Position Error (m)Proprioceptive Distribution Matching (π₁)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Proprioceptive Distribution Matching (π₁)Velocity Command=y +0.5m/s, Finetuning with Residual Actuator Model=true2026.04 | -0.15 | -0.1 | 3 | |
| Finetune w/ Residual Actuator Model (π1)Velocity Command=y +0.5m/s2026.04 | -0.15 | -0.1 | 3 | |
| Proprioceptive Distribution Matching (π₁)Velocity Command=x -0.5m/s, Finetuning with Residual Actuator Model=true2026.04 | -0.025 | 0.025 | 0 | |
| Finetune w/ Residual Actuator Model (π1)Velocity Command=x -0.5m/s2026.04 | -0.025 | 0.025 | 0 | |
| Proprioceptive Distribution Matching (π₁)Velocity Command=x +0.5m/s, Finetuning with Residual Actuator Model=true2026.04 | 0 | 0.05 | 2 | |
| Original Policy (π0)Velocity Command=x -0.5m/s2026.04 | 0 | -0.4 | 3 | |
| Original Policy (π0)Velocity Command=yaw -0.3925m/s2026.04 | 0 | 0.025 | -12 | |
| Finetune w/ Residual Actuator Model (π1)Velocity Command=x +0.5m/s2026.04 | 0 | 0.05 | 2 | |
| Original Policy (π0)Velocity Command=yaw +0.3925m/s2026.04 | 0.06 | 0.17 | 3 | |
| Proprioceptive Distribution Matching (π₁)Velocity Command=yaw +0.3925m/s, Finetuning with Residual Actuator Model=true2026.04 | 0.1 | 0 | 0 | |
| Finetune w/ Residual Actuator Model (π1)Velocity Command=yaw +0.3925m/s2026.04 | 0.1 | 0 | 0 | |
| Proprioceptive Distribution Matching (π₁)Velocity Command=y -0.5m/s, Finetuning with Residual Actuator Model=true2026.04 | 0.2 | 0 | 5 | |
| Original Policy (π0)Velocity Command=y +0.5m/s2026.04 | 0.2 | -0.25 | 3 | |
| Finetune w/ Residual Actuator Model (π1)Velocity Command=y -0.5m/s2026.04 | 0.2 | 0 | 5 | |
| Proprioceptive Distribution Matching (π₁)Velocity Command=yaw -0.3925m/s, Finetuning with Residual Actuator Model=true2026.04 | 0.25 | 0.15 | -3 | |
| Finetune w/ Residual Actuator Model (π1)Velocity Command=yaw -0.3925m/s2026.04 | 0.25 | 0.15 | -3 | |
| Original Policy (π0)Velocity Command=x +0.5m/s2026.04 | 0.6 | 0.35 | 5 | |
| Original Policy (π0)Velocity Command=y -0.5m/s2026.04 | 1.2 | -0.5 | 10 |