Robotic Dance

In plain words: Familiar: Robotic dance (idea)

Designed 2 min read 3 glossary terms 4 connected pages
The short version written for the Snayu team

A candidate use case where a person wearing motion trackers leads a group of robots in a dance, with each robot adapting the movement to its body, plus related research.

Why read it: Read this if exploring motion retargeting or multi-robot synchronisation ideas.
  1. 1Modes: Mirror (follow together), Echo (deliberate delay) and Ensemble (variations per body).
  2. 2Live Kinesis would retarget motion; each robot's local controller keeps balance and safety.
  3. 3Adds no published workflow or supported hardware; research checked 2026-09-17.

Designed Candidate use case only.

Words on this page, decoded click one for its card

The original text Underlined words open their plain-English card.

Put on motion trackers, step into the admitted dance space, and lead a group of physical familiars with your own body. Each robot adapts the movement to its proportions and capabilities. Candidate modes are MirrorPersona Keeper (follow together), EchoVoice (a deliberate time offset), and Ensemble (declared variations for different bodies).

This is a candidate Familiar use case, with research checked on 2026-09-17. It suggests a dance PatternWorkflow and exact body/controller profiles; it adds no published PatternWorkflow or supported hardware to the Familiar contract. State of Work keeps the shared delivery boundary.

How it could fit

armed wearable / optical mocap source
→ calibrated human motion with timestamps
→ Live Kinesis retargeting for each exact robot rig
→ separately admitted local tracking controller
→ Familiar mission observations, effects, and settlement

Live Kinesis is the proposed seam for source epochs, clocks, retargeting, gaps, and expiring motion segments. A pinned GMR adapter is one candidate for its technical mapping. Familiar would bind each FamiliarBody@3 to a finite dance mission, the designated consenting performer, dance mode, space, duration, and stop policy. The body's controller retains balance, contacts, joint limits, collision response, and fast actuation; an LLM can select an admitted mode or request a stop while this local loop runs independently.

Admit and map each robot separately before using shared source timing to synchronize the group. The group policy must say what happens when one member drops out, and intentional EchoVoice delay must remain distinguishable from stale tracking. Lost tracking, changed calibration, or an expired segment invokes the commissioned local response; reconnect requires fresh admission.

At settlement, record performed intervals, deviations, stopped members, and unknown effects. Participation alone does not authorize all uses of the source: capture, recording, and reuse of the performer's motion remain separate consent choices.

Research and existing demonstrations

These are integration leads, with different evidence boundaries:

Reference What it contributes
TWIST and TWIST2 Whole-body human-to-humanoid teleoperation. TWIST2 uses a PICO 4 Ultra and two PICO Motion Trackers, making it a concrete wearable-input reference for a Unitree G1 profile.
GMR — General Motion Retargeting Maps human motion onto different humanoid bodies and supports live teleoperation. A candidate retargeting component; mapping a pose does not establish balance or successful physical execution.
ASAP Demonstrates learned agile skills, including APT Dance, on a real G1. Useful for controller training and a rehearsed-clip route; it does not establish live imitation of an arbitrary dancer.
CHINGMU × Unitree, WRC 2026 The vendor's September 3 report describes a dancer's captured movement driving a humanoid synchronously. A direct demonstration of the intended experience, without an independently verified latency or general compatibility claim.

Selecting a profile would still require its exact capture interface, robot/controller revision, measured timing, and code, weights, body-model, and data-use terms. None of these references is an installed LychD dependency.

First bounded probe

Start with one recorded performer trace and two simulated bodies with different mappings. Exercise MirrorPersona Keeper and EchoVoice, then inject stale frames, calibration changes, one-body refusal, and an operator stop. Check timing and per-body partial settlement before attempting a separately commissioned single-robot live trial. Ensemble variations can follow; the fixture establishes no physical safety or hardware compatibility.

Return to Familiar or Embodiment.