Robot dynamic path modification via multi-sensor data tracking

Abstract

This doctoral thesis proposes and validates experimentally nonlinear control strategies for a six-axis industrial robot Meca500 from Mecademic company using two approaches. The first approach aims to provide a precise dynamic trajectory tracking in Cartesian velocity space in the presence of parameter uncertainty and undesirable disturbances. While robot calibration can improve position and orientation (pose) accuracy, the only way for a user to improve the dynamic path accuracy is by “guiding” the robot with the help of an external sensor and a control algorithm running on a separate computer. For this purpose, industrial robots, which are normally controlled with pre-programmed position-mode instructions, offer sometimes the possibility to modify the pose of the robot end-effector on the fly. In the case of Mecademic’s Meca500 industrial robot, operators can indirectly modify the end-effector pose by controlling the robot joint or Cartesian velocity. In this thesis, a practical application of an active disturbance rejection control (ADRC) scheme is presented to improve the path accuracy of the Meca500. The dynamic path correction is achieved by first measuring the distance between a fixed point and the robot tooltip with a linear transducer (Renishaw’s QC20-W ballbar), and then feeding the tooltip velocity vector to the robot (via Ethernet TCP/IP). The (circular) path accuracy of the robot is significantly improved for different robot TCP velocities. For example, at 50 mm/s, the maximum radial error is less than 0.100 mm, and the mean error is 0.015 mm. The second approach ensures manipulation of Meca500 with a six-axis finger joystick Space- Mouse in a very soft and smooth manner avoiding human-robot cross-coupling, robot’s motor perturbations, oscillation, noise, or aggressive robot response since the system dynamics in this mutual tracking control problem as well as its rate of change may significantly change in amplitude, time, and direction during the normal robot’s movements. To this end, a novel concept of a multi-stage controller connected in series is proposed and verified practically by plotting the mouse-robot mutual response along with capturing a video for all types of movements. The multi stage controller starts with a fuzzy logic control block based on a natural description of the mouse-robot system by humans, which is better than creating a mathematical model. This provides a more feasible solution that integrates human knowledge about the relation between the input and the output at each operating point. The next stage is an admittance control block, connected in series and aiming to damp some oscillations that appear during the robot’s manipulation. Finally, an ADRC tracking differentiator is incorporated as a third and final stage to filter the noise caused by the admittance control, add an intended delay in some robot’s motion situations, and transform the position signal into velocity to be sent to the robot in velocity mode. The experimental results show the enhanced robot performance after adding the external multi-stage controller.

Date	19 Jan 2021
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Ouassima Akhrif (Supervisor) & Ilian Bonev (Co-supervisor)