The Grasp Planner and Execution Pipeline automates the identification and execution of optimal grasping strategies for both structured and unstructured environments, addressing the critical industry need for handling high-mix product varieties without manual reprogramming.

Built on ROS2, this module leverages Reinforcement Learning and physics-based simulation (MuJoCo) to dynamically compute grasp points using either CAD models for known parts or geometric primitives for unknown objects. This solution significantly enhances manufacturing flexibility and efficiency by reducing setup times and enabling robots to adapt to novel items autonomously.

Responsible partner: TEK