hill-improver

An entry for 1.5 m Hill Climb of the Microduck Arena. Simulation only.

A remix of jrubiosainz/microduck-game-hillclimber-remix5 (revision 7865b3d7c): a seed-3 branch in its free slot 3 (counter 25), whose table rows were re-searched by SPSA gradient descent (seed 3 crossing 0.75963 to 0.75204 s); the branch-table graph lineage goes back to dashi's microduck-hill-remix-r760. The network is frozen, and every other initializer and operator is the base's.

Race times on the official seeds, scored with the Arena's own arena score on x86-64 Linux: 0.74, 0.76, 0.76, 0.76, 0.76 s, a 0.756 s mean. These are in-sample numbers. The search scored candidates on the same seeds.

Not for a real robot

Trained and judged only in the Arena's simulator. Do not load it on a physical MicroDuck.

Files

policy.onnx: sha256 9afc9f02be990cf692992104a3e8c9f81d0eff871aeac2b69391aa9d02fb96b4. Input obs [1, 61], output actions [1, 14].
manifest.json: the policy manifest (schema 2).

Credits

jrubiosainz, whose policy this is built on, and the authors credited there, and Pollen Robotics. The base declares no licence, and this repository declares none either.