Real-time path correction of industrial robots in machining of large-scale components based on model and data hybrid drive

Industrial robots are increasingly used in machining of large-scale components due to advantages of high re-peatability, large workspace and low cost. Nevertheless, applying industrial robots to high-accuracy machining of large-scale components remains a challenge, where the major hurdle is the insufficient manipulator stiffness due to joint flexibility. When the robot is performing a machining task, joint flexibility-induced position errors between motor and link called joint position errors (JPEs), as the main source of robot deformations, make the robot deviate from the desired path. For most industrial robots, due to the lack of link-side encoders, it is difficult to obtain the JPEs by direct measurement and compensate them in the controller, which deteriorates the path accuracy of the robot during machining greatly. To address this problem, this paper presents a real-time path correction approach of industrial robots based on JPE estimation and compensation with requiring only motor-side measurements and external wrenches. The proposed approach is divided into three steps. First, to estimate the actual link position of the robot in real-time, the dynamics of a manipulator with joint flexibility called flexible dynamics (FD) is introduced. Second, by taking both FD and disturbance dynamics into account, a novel link state estimator called flexible-dynamics based disturbance Kalman filter (FDBDKF) is developed, and thus JPEs can be estimated in real-time. Third, a data-driven locally weighted projection regression (LWPR)-based JPE prediction and compensation method is developed to further improve the compensation accuracy of the JPEs. Simulation and experimental results, obtained on a 6-DOF industrial robot, demonstrate the feasibility and effectiveness of the proposed approach. Experimental results show significant improvement (>80%) in the path accuracy of a simple material removal process corrected using the proposed approach.