Microduck Circus

Three ducks learn to skip a shared long rope, two turning and one jumping, through an act-fail-practice-adapt agent loop with a holdout round. Sim-only.

Project Overview

MicroDuck is a simulation project for studying contact-aware locomotion and coordination between three bipedal robots. Two Microduck robots drive a flexible rope through handles attached to their mouths; a third robot learns to jump over it. The system combines reinforcement-learned motor policies, explicit phase feedback, and a geometric/contact-based evaluator.

The central problem is to make rope rotation, foot clearance, and supported landing work together under physical contact . A correctly timed jump alone is insufficient: the rope must pass below both feet, the robot must land cleanly, and the rope must remain attached to its turners.

This repository provides the deployment runtime, exported policies, evaluation evidence, rendering tools, and a reproducible patch containing the training implementation. It is a MuJoCo simulation , not a demonstrated hardware deployment or an end-to-end multi-agent RL system.

Task

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

Source Coverage

The upstream README includes: Demonstration, Method, 1. Shared physical environment, 2. Motor-policy training, 3. Closed-loop coordination, Evaluation, What counts as a successful skip?, Measured results.

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.