Dual-arm Object Manipulation by a Hybrid Controller with Kalman-Filter-based Inputs Fusion

Grasping and manipulation of various objects by arms are essential robotic tasks. The dual-arm system in comparison with a single-arm system can grasp objects with larger sizes and shape variations owing to its high degree of task-space redundancy. We report on the development of a dual-arm manipulation strategy suitable for object holding and moving. The controller has master-slave structure. It hybridizes effects of the desired position trajectory (spatial relation) and the force interaction between the arm and the object (compliance), which includes maintaining adequate surface contact and normal force. The compensations of the position and force errors from the two controllers are fused in a Kalman filter to yield the control input for final motion correction. A dual-arm robot is built. The performance of the controller is experimentally evaluated, and the results conclude that the proposed control strategy can deal with position error, external force disturbance, and object weight variation.