Walk
An official Microduck move: Pollen Robotics' walk,
retrained from scratch by Microduck Academy with the upstream recipe
as-is (task Mjlab-Velocity-Flat-MicroDuck of pollen-robotics/microduck_rl at 2b25a48,
2999 iterations, 4096 envs, no reward edit), so that it carries the
base/model.pt checkpoint the Academy needs to remix it. The policy Pollen ships is
alpha_walking.onnx.
The walk is the Academy's base walk (spikes/base_walk, Velocity task, 3000 iterations): the recipe behind alpha_walking.onnx. Pollen's default velstand.onnx adds a protective fall + recovery layer warm-started from a private checkpoint, which cannot be retrained faithfully; the walk itself is the same.
Pollen Robotics' walk: the velocity-commanded gait every microduck ships with, standing still at zero command.
Fidelity check (this move vs Pollen's policy, same scene, same command, take vx0.35)
| number | this move | Pollen's alpha_walking.onnx |
|---|---|---|
| final_upright | True | True |
| max_tilt | 0.254 | 0.259 |
| base_z_mean | 0.1227 | 0.1183 |
| base_z_min | 0.1127 | 0.1126 |
| travel_m | 1.94 | 1.207 |
| travel_x_m | 1.132 | 1.159 |
| speed_m_s | 0.243 | 0.151 |
| cadence_steps_s | 5.14 | 5.01 |
| fell | False | False |
Recorded with space/payload/render/rollout.py on the upstream robot MJCF (spikes/base_moves/fidelity.py).
Run it on a robot
sudo robotctl policy add walk tfrere/microduck-move-base-walk
robotctl robot do walk
Files
policy.onnx (play it, normalizer baked in), base/model.pt + base/agent.yaml + base/env.yaml (the rsl_rl
checkpoint and configs: remix = fine-tune from it), rollouts/0.traj (trajectory.v1 showcase take, with the
commands), video.mp4 + poster.jpg (the card clip), manifest.json
(schema 2 + academy block: origin, source task and policy, training facts, fidelity numbers), fidelity.json,
train.json.
Credits and license
The task, the reward design, the robot model and the training recipe are Pollen Robotics' work (pollen-robotics/microduck_rl, Apache-2.0). This repo republishes a retrained policy of that recipe under the same Apache-2.0 license, with the checkpoint, so the Academy's users can start from it.