MODEL → EXPERIMENT → EVIDENCE
UR7e Policy Lab.
A frozen robot policy. A classical controller. The same trajectory.
REFERENCE EXPERIMENT
600 CPU episodes.
Both sides of the result.
100 paired trajectories × 2 controllers × 3 noise levels. No retraining. One frozen checkpoint.
| Controller | Strict completion | Tracking RMS* |
|---|---|---|
| Frozen SAC | 93 / 100 | 5.21 mm |
| Fixed DLS | 90 / 100 | 2.38 mm |
*Mean per-episode RMS over acquired trajectories, including failures. DLS tracks more accurately in this experiment. The completion difference does not establish statistical superiority. All conditions and metadata ↗