Adaptive critic neural network-based object grasping control using a three-finger gripper

Robotic grippers that are capable of manipulating objects such as plant trays, fruits, vegetable and so on are required in MARS´ greenhouse operation. Grasping and manipulation of objects have been a challenging task for robots. It is important that the manipulator performs these tasks accurately and faster without damaging the object. The complex grasping task can be defined as object contact control and manipulation subtasks. In this paper, object contact task is defined for the fingers in terms of following a trajectory accurately. On the other hand, the grasping task is defined in terms of maintaining a predefined applied force by the fingers so that the object is properly secured. A sophisticated controller is required for the grasping task since the process of grasping an object without apriori knowledge of the object´s size, texture, and softness is rather difficult task. The proposed scheme consists of a feedforward action generating neural network (NN) that compensates for the nonlinear gripper and contact dynamics. The learning of this NN is performed on-line based on a critic signal so that a three-finger gripper track a predefined desired trajectory, which is specified in terms of a desired position and velocity for object contact control while it applies a desired force on the object for grasping. Novel NN weight tuning updates are derived for the action generating NN and a Lyapunov-based stability analysis is presented. Simulation results are shown for a three-finger gripper grasping an object