Towards Human-Like Table Tennis Serving: Preliminary Exploration with Simplified Serving Motion Using an Industrial Robotic Manipulator in NVIDIA Isaac Sim
Po-Chuan Chiou, Jing-Chen Hong
AI summary
Problem
Current table tennis serving robots lack human-like motion planning, limiting their realism and utility for training and biomechanical research.
Approach
The authors built a digital twin in NVIDIA Isaac Sim and simulated simplified one-, two-, and three-joint forward kinematics to evaluate how joint coordination impacts ball trajectory and velocity.
Key results
- Successful ball delivery to the opposite side across all joint configurations
- Modest velocity increase with additional active joints
- Peak end-effector speed achieved in three-joint planning
- Identification of insufficient biomimetic coordination in simplified models
Why it matters
Provides a foundational simulation framework and kinematic baseline for developing human-like robotic table tennis servers, relevant to robotics, sports science, and AI training.
Abstract
In this report, we present our latest work-in- progress result of table tennis serving simulation. The ultimate goal for our research is realization of human-like table tennis serving with a 6-joint robotic arm. The preliminary evaluation indicates the potential of multi-joint bionic serve motion planning with extended topics and future directions discussed.