Design and Control of an Articulated Robotic Arm Using Visual Inspection for Replacement Activities

Abstract

Design of robotic systems and their control for inspection and maintenance tasks is highly complex activity involving coordination of various sub-systems. In application like inspections in fusion reactor vessels and deep-mining works, a regular off-line maintenance is necessary in certain locations. Due to the hostile environments inside, robotic systems are to be deployed for such internal observations. In this regard, current work focuses on the methodology for maintenance of the first wall blanket modules in a fusion reactor vessel using a manipulator system. A design is proposed for wall tile inspections in an ideal environment in which vacuum and temperature conditions are not accounted and wall surface curvature is not accounted initially. The entire design has four important modules: (i) mathematical modelling (ii) control system design (iii) machine vision and image processing, (iv) hardware development and testing.
A five- axis articulated manipulator equipped with a vision camera in eye-to-hand configuration is designed for performing the pick and place operations of the defected tiles in a systematic manner. Kinematic and dynamics analysis of the system are first carried-out and a scaled prototype is fabricated for testing various operating issues. Forward kinematics of manipulator allows in estimation of robot workspace and in knowing the singular regions during operation, while the inverse kinematics of the manipulator would be needed for real time manipulator control task. Dynamics of manipulator is required for design of model-based controllers. Interactive programs are developed in Matlab for kinematics and dynamics and three-dimensional manipulator assembly configuration is modelled in SolidWorks software. Motion analysis is conducted in ADAMS software in order to compare the results obtained from the classical kinematics. Two types of model-based control schemes (namely Computed Torque Control and Proportional Derivative-Sliding Mode Control approach) with and without external disturbances are implemented to study trajectory tracking performance of the arm with different input trajectories. A disturbance observer model is employed in minimizing the tracking errors during the action of external disturbances such as joint friction and payload.
In order to experimentally understand the inspection and replacement activities, a test set-up is developed using vision camera and microcontroller platform to guide the robot joint servos so as to perform defected object replacement activity. Presence of crack and the coordinate of the region are indicated with the use of image-processing operations. Using a high resolution Basler camera mounted at fixed distance from the tile surface, the surface images are acquired and image processing module identifies the crack details using edge detection algorithms. Necessary motion of the end-effector will be provided based on the pose calculations using coordinate transformations. Both visual inspection and joint guidance are combined in a single application and the results are presented with a test case of tile replacement activity. The results are presented sequentially using a test surface with uniform rectangular tiles.