Dexterous manipulation for assembly work using multilink dual arm manipulator with vision sensors

Abstract

This paper presents our work on realizing the manipulation of nut and bolt for assembly work using dual arm manipulator. The robot architecture consists of multiple active sensors and control procedures that are structured to work for real time operations. It is specifically designed to be similar to human arm which has seven joints for each arm to perform complicated assembly tasks on small parts using both hands that usually requires high cost human labour. Sensors are optimally used to provide data of the environment and the target object in variable conditions, shapes and sizes. Data between eye-in-hand and eye-to-hand cameras are fused and handled together with force sensors and encoders to calculate the required robot dynamics of the seven link arms for object manipulation. The first experiment that has been used as the basic platform in preparing our robot for assembly task consists of autonomous picking and placing an M8 sized bolt and hexagonal nuts of minimum 2 mm radius. Besides, based on the proposed vision processing, the robot hand could self adjust its wrist angle to avoid collision between the fingers and objects during grasping. Moreover, we showed application of vision to align M10 bolt and nut for complicated screwing operation. Experiment results are laid out to show the robot performance for above operations thus producing a dexterous dual arm robot for assembly work purposes.