Optimization Framework of a Robotic Pick and Place System for Waste Sorting

This work proposes an open, distributed framework for planning and controlling robotic systems capable of separating waste items based on their material and/or geometry. A central planner derives valid picking poses and efficiently generates feasible, collision-free trajectories for picking and placing items into the corresponding bins. The planned trajectories are then transferred to a controller tasked with overseeing the overall process, controlling the robot’s motion, and managing the gripper. Robot Operating System (ROS) serves as the middleware, and Gazebo is selected as the simulator, enabling the integration of a wide variety of robots, grippers, path planners, control and planning methods. Finally, numerous valuable metrics are leveraged to demonstrate how different hardware and algorithmic choices of the robotic cell affect its performance, enabling the selection of an optimal configuration.