SUAT · EMI LAB / RESEARCH MAP
SUAT · Force-Aware VLA Research
Successful manipulation does not describe contact quality. Could recent force-torque and motion history predict near-future instability before an action chunk has finished?
Make trajectory quality inspectable
I generated Franka trajectory candidates, calculated six cost components and trained a neural ranker. A GUI exposed grasp stability, stress/deformation, smoothness, acceleration, obstacle clearance and path length. This was relative ranking, not a calibrated safety score.
Change the route when simulation became impractical
I explored simplified physical proxies, Isaac Lab/TacEx and MuJoCo replay. Contact fidelity, data-generation speed and available compute prevented a credible new dataset within the placement, so I moved to existing ForceVLA trajectories.
Test the warning signal before integrating it
A standalone GRU used temporal history from 244 real inputForce trajectories across five tasks. At a five-step horizon, AP was 0.736 versus 0.571 for a current-force heuristic. The reported test set contained 20,475 windows.
Separate learning risk from improving actions
I added an auxiliary MLP risk head during ForceVLA LoRA training. Limited offline subsets produced AUROC 0.759 at h=5 and 0.748 at h=15 against force-derived proxy labels. No robust action-loss improvement was established and no closed-loop rollout was conducted.
Two experiments, read separately
| Experiment | Metric / horizon | Result |
|---|
| Standalone GRU / current-force heuristic | AP · h=5 | 0.736 / 0.571 |
| ForceVLA + MLP | AUROC · h=5 / h=15 | 0.759 / 0.748 |
Source notes
Placement Interim Report · pp. 2–6
Research context, individual responsibilities and reasons for the simulation-to-ForceVLA pivot.
Placement Review SUAT / Dobot v9 · slides 3–9
Later results and explicit limits; read alongside the interim report.
Temporal Force Risk Head — progress (fixed) · slide 1
Standalone GRU / heuristic comparison; integration status on slide 2 predates the later review.