Roller feedback ES generation 16
Archived derivative-free antithetic evolution strategy selected generation 16. A learned 126-parameter latent-feedback head adds bounded corrections over the frozen attributed Pollen neural backbone and prior own physical-state command adapter. This is not PPO or from-scratch network training.
Event: roller-sprint-2m. Fixed external command: 1.0. Accessories: rollers. Input [1,61], output [1,14], float32 ONNX, 50 Hz.
Local validation
At fixed command 1.0, the local canonical five-start result was 2.918 seconds, 5/5 finishes. New ordinary-start validation finished 128/128; command-stress validation finished 64/64 at commands 0.9 and 1.1. The comparison for this archive screen was its own historical predecessor, not the current Roller champion. It does not improve the published Roller best of 2.570 seconds.
This is a distinct previously unsubmitted research checkpoint for daily-run points, not a personal-best or podium claim. Official Arena acceptance, completion and points are not guaranteed. Local Apple Silicon results may differ from server results. No Tug policy or cup entrant is changed by publication of this repository.
Inherited upstream terms and attribution are retained in upstream-base-card.md and LICENSE-NOTE.txt. No new blanket redistribution license is asserted.
Model SHA256: 783ce05087de5af856411458f8bbd2b4a4bcd73bd75cdf9104afcdf8e8a41374.
Simulation only. No real-robot safety or reliability guarantee.