Environment Based Multi-Object Grasp Planning in Bin Picking Process

Given the potential of bin picking to eliminate manual handing practices in manufacturing and logistics industries, this paper presents an environment-based grasp planning pipeline to address the challenges posed by complex object arrangements in cluttered environments. The pipeline integrates offline grasp planning system with online grasp filter, designed specifically for multi-object cluttered bin picking tasks. In the offline phase, the pipeline utilizes the improved Particle Swarm Optimization (PSO) algorithm to generate the grasp solutions by incorporating gripper and object models. In the online phase, the system considers grasp and robot kinematics to prioritize objects and determine the best grasp based on the grasp task scene model and the precomputed solutions from the offline phase. The proposed pipeline is validated through both simulated and experimental trials using the Franka Emika Panda robotic arm. The results demonstrate the effectiveness of the pipeline in handling multi-object cluttered bin picking tasks.