Design and Control of a Quasi-Direct Drive Robotic Gripper for Collision Tolerant Picking At High Speed
Frederik Ostyn, Bram Vanderborght, Guillaume Crevecoeur
Abstract
Faster robotic picking of objects can improve industrial production throughput. A gripper design with appropriate control strategy is demonstrated to pick objects at high speed, while being tolerant to unintentional collisions. The design has a quasi–direct drive with backdrivable, rigid finger mechanism to transfer collision forces with minimal delay towards the motor side. These forces are detected based on band–pass filtered momentum and motor speed monitoring. The motor velocity is observed through Kalman filtering in order to reduce noise in the low–level control loop and band–pass momentum observer, resulting in a higher collision detection sensitivity. On a higher level, the path of the gripper is planned with respect to the trade–off picking speed versus collision tolerance, quantified by the remaining gripper stroke upon impact. The gripper collision tolerance was experimentally verified in several scenarios during high–speed object picking where it mitigated collisions up to 0.6 m/s.