AI-driven Real-time Robotic Ball Catching System using Ur5e Robotic Arm

An autonomous UR5e robotic arm performs dynamic ball catching using real-time 3D vision and predictive motion planning.

This project showcases a full perception–planning–actuation pipeline implemented entirely in ROS 2 C++, leveraging an Intel RealSense D415 camera for on-arm depth sensing and advanced trajectory prediction for high-speed interception.

Key features demonstrated include:

• Real-time ball detection and 3D localisation
• Trajectory estimation using minimal observations
• Fast, constraint-aware motion planning
• Safety-focused, reliable actuation with a net basket tool

The system achieves robust catching performance within hardware-imposed limits, highlighting the importance of tightly integrated sensing and control for practical autonomous robotics.

Tools: ROS2, UR5e Cobot, Inverse Kinematics, C++, Python