microduck-basketball

Balances on a free-rolling size-7 basketball and follows velocity commands from proprioception alone, through a one-layer LSTM with no ball state in the actor inputs. Simulation and export checks pass; onboard timing and real-robot behavior are untested.

Project Overview

--- thumbnail: https://huggingface.co/HannesVonEssen/microduck-basketball/resolve/main/media/social-preview.png pipeline tag: reinforcement-learning tags: [microduck, microduck-policy, mjlab, robotics, balance, basketball, lstm, recurrent] license: apache-2.0 ---

Microduck balances on a free-rolling size-7 basketball and follows velocity commands using proprioception and a one-layer LSTM , with no ball state in the actor inputs . The released checkpoint is b11, iteration 6,999 . This is an experimental policy for hardware testing : simulation and export checks passed; onboard timing and real-robot behavior are untested.

ONNX, full training checkpoint and configuration on Hugging Face · Training source · LSTM runtime support: PR 231

Task

Field	Value
Source	HUGGING FACE MODEL SOURCE
Runtime contract	61 observations / 14 actions / 50 Hz
Verification	Pending independent Duckio review
Artifact	Not indexed

Source Coverage

The upstream README includes: Video and measured performance, Architecture and deployment contract, Download and continue training.

Intended Use

Use this entry to assess the project and locate its canonical implementation. Read the complete README, inspect the manifest and artifacts, and reproduce the source's evaluation before assigning simulation or hardware verification.