AdaPT Towards Professional Tennis Styles for Humanoid Robots
with Adaptive Motion Planning and Tracking

Tao Huang1,2,4,*,‡ Ruofei Liu1,4,* Xuchen Tang1 Xinyin Zhang1 Junli Ren2 Huayi Wang2 Feiyu Jia2
Yukai Qi1 Kangning Yin2 Weishuai Zeng2 Lipeng Chen4 Xi Li3 Ting Wu1 Kailin Li2 Ruoli Dai1
Jingbo Wang2,† Lei Han1,†,‡ Jiangmiao Pang2,†

1 Noitom Robotics    2 Shanghai AI Laboratory    3 Dobot Robotics    4 Shanghai Jiao Tong University

* Equal Contribution    † Equal Advising    ‡ Project Lead

We propose AdaPT, an Adaptive motion Planning and Tracking framework
that learns professional tennis serve and rally styles.

Tennis Motions from Video & MoCap · Rally on G1 & Atom P3 · In-the-wild Serve

Paper GitHub Code arXiv YouTube

Overview

Tennis Rally

Rally hitting styles from different players, deployed on Unitree G1 and Atom P3.

Player
Robot

Tennis Serve

Tennis serving styles from multiple players, validated in both Mocap systems and in-the-wild settings.

Human vs. Robot

Dataset

Video motions are recovered and retargeted using GVHMR and GMR. Mocap motions are collected by the Noitom Robotics and retargeted using UMR (Unified Motion Retargeting, coming soon).

Duration by Motion Style

21.5 h

Total motion duration

6

Athlete styles

7

Stroke and serve types

30/120 Hz

Capture rate

GLB previews cannot load from file://. In this folder run python -m http.server 8000, then open http://localhost:8000/.

Forehand
Backhand
Slice
Drop
Volley
Flat serve
Topspin serve

Baselines

Player
Task

Citation

@article{huang2026adapt,
      title={Towards Professional Tennis Styles for Humanoid Robots with Adaptive Motion Planning and Tracking},
      author={Tao Huang, Ruofei Liu, Xuchen Tang, Xinyin Zhang, Junli Ren, Huayi Wang, Feiyu Jia, Yukai Qi, Kangning Yin, Weishuai Zeng, Lipeng Chen, Xi Li, Ting Wu, Kailin Li, Ruoli Dai, Jingbo Wang, Lei Han, Jiangmiao Pang},
      journal={arXiv preprint arXiv:2608.20087},
      year={2026},
      url={https://arxiv.org/abs/2608.20087}
}