pith. machine review for the scientific record. sign in

arxiv: 2604.03999 · v1 · submitted 2026-04-05 · 💻 cs.RO

Recognition: no theorem link

Dynamic Whole-Body Dancing with Humanoid Robots -- A Model-Based Control Approach

Shibowen Zhang , Jiayang Wu , Guannan Liu , Helin Zhu , Junjie Liu , Zhehan Li , Junhong Guo , Xiaokun Leng
show 7 more authors
Hangxin Liu Jingwen Zhang Jikai Wang Zonghai Chen Zhicheng He Jiayi Wang Yao Su
Authors on Pith no claims yet

Pith reviewed 2026-05-13 17:38 UTC · model grok-4.3

classification 💻 cs.RO
keywords motiondynamicdanceframeworkmotionsexecutionhumanoidrobots
0
0 comments X

The pith

A two-stage framework retargets human dance via QP and trajectory optimization then executes it with centroidal MPC on the Kuavo 4Pro robot, enabling four-minute live multi-robot performances where longer prediction horizons improve both expressiveness and stability.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The system works in two parts. First, human dancers are recorded with motion capture. The recorded motions are adjusted to the robot's body shape by solving a quadratic program that respects joint limits and balance. These motions are then refined with trajectory optimization so the robot's center of mass and foot placements stay dynamically feasible. Second, while the robot dances, a model-predictive controller based on simplified centroidal dynamics follows the planned trajectory in real time. It looks ahead several steps and changes where the swing foot lands to counteract pushes, uneven floors, or other disturbances. The authors tested the full pipeline on the Kuavo 4Pro humanoid both in simulation and in a live four-minute public performance with four robots dancing together. They report that increasing the MPC prediction horizon made the motions look more expressive during planning and more stable during execution.

Core claim

We validate our framework on the full-size humanoid robot Kuavo 4Pro, demonstrating the dynamic dance motions both in simulation and in a four-minute live public performance with a team of four robots. Experimental results show that longer prediction horizons improve both motion expressiveness in planning and stability in execution.

Load-bearing premise

The offline trajectory optimization produces motions that remain dynamically feasible and trackable by the online centroidal MPC when real-world disturbances occur; if the simplified centroidal model deviates substantially from the full robot dynamics, the foot-placement corrections may fail to maintain balance.

read the original abstract

This paper presents an integrated model-based framework for generating and executing dynamic whole-body dance motions on humanoid robots. The framework operates in two stages: offline motion generation and online motion execution, both leveraging future state prediction to enable robust and dynamic dance motions in real-world environments. In the offline motion generation stage, human dance demonstrations are captured via a motion capture (MoCap) system, retargeted to the robot by solving a Quadratic Programming (QP) problem, and further refined using Trajectory Optimization (TO) to ensure dynamic feasibility. In the online motion execution stage, a centroidal dynamics-based Model Predictive Control (MPC) framework tracks the planned motions in real time and proactively adjusts swing foot placement to adapt to real world disturbances. We validate our framework on the full-size humanoid robot Kuavo 4Pro, demonstrating the dynamic dance motions both in simulation and in a four-minute live public performance with a team of four robots. Experimental results show that longer prediction horizons improve both motion expressiveness in planning and stability in execution.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard robotics modeling assumptions plus a small number of tunable weights; no new physical entities are postulated.

free parameters (2)
  • prediction horizon length
    Tuned and compared experimentally; longer values reported to improve expressiveness and stability.
  • QP and MPC cost weights
    Hand-tuned parameters that balance tracking accuracy, balance, and smoothness.
axioms (1)
  • domain assumption Centroidal dynamics provide a sufficiently accurate reduced-order model for real-time foot-placement correction.
    Invoked in the online MPC stage without additional validation against full-body dynamics.

pith-pipeline@v0.9.0 · 5527 in / 1411 out tokens · 57271 ms · 2026-05-13T17:38:01.993449+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.