Mobility study / 28 September 2026
Move. Turn.
Recover.
A pinned-leg simulation study of forward and reverse travel, both turns, rotation in place and bounded disturbances.
Simulated evidence · physical qualification open01 / Inspect a run
Every trace has a result.
Replay a nominal run or a challenge. The table below includes every parameter variation.
Ground path
Meters · circle marks the endpoint tolerance
Body response
Pitch / roll, degrees · dashed lines at ±12° pitch gate
These are recorded physics results, with assumed sensor and actuator behavior. The simulation has no raw IMU fusion, tire compliance, thermal model or powered leg motion. A saved pass is a candidate for physical testing.
02 / Parameter coverage
All qualification cases.
Each row includes nominal, twelve seeded parameter sets and four explicit corner/pose configurations. Counts are simulation outcomes, not real-world success probabilities.
| Scenario | Passed | Worst endpoint | Worst heading | Worst pitch |
|---|
03 / What changed
Position matters as much as balance.
The first controller passed the nominal cases but only 139 of 391 parameter runs. Small center-of-mass offsets caused drift and missed endpoints. Stronger position control was tested against the same cases and thresholds.
Strong shoves, a 20 mm square obstacle and excessive delay remain separate challenges. Their failures stay in the evidence.
Source: raw results with model hashes · reproduce the study · physical acceptance protocol