AI-Powered Delta Robot System for Real-Time Table Tennis Ball Tracking and Striking

The integration of AI and robotics has advanced sports automation, with table tennis serving as a challenging testbed due to its fast-paced dynamics. While systems like OMRON's FORPHEUS and PongBot's M-One demonstrate highend capabilities, their complexity limits accessibility. This study develops a cost-effective delta parallel robot system for table tennis ball striking, aimed at simplifying training applications. The system uses YOLOv5 for real-time ball tracking and trajectory prediction, MATLAB for motion planning, and TwinCAT with Scurve trajectory planning for smooth robotic arm control. A 3Dprinted paddle and ROS-based communication further enhance the system. Experiments achieved a 65.5 % success rate, with higher accuracy (74 %) for diagonal trajectories due to extended observation time. Challenges include image noise and simplified physics models affecting trajectory predictions. Future improvements will focus on advanced filtering and real-time adjustments. This research highlights the potential for integrating AI and robotics in dynamic object interaction tasks.