How would you predict grasp poses and plan collision-free trajectories for Berkshire Grey's robotic picking system?
