Declarative Models for Self-Calibrating Robots

The use of industrial robots in manufacturing often requires good calibration approaches and tools to obtain and maintain the desirable accuracy of the robot, for the robot to be able to execute a task in a satisfactory manner. For high-accuracy methods of calibration, it is necessary to first identify the robot’s stiffness parameters. This thesis will focus on a method that is based on a clamping procedure. In this procedure, the robot’s end effector is clamped to a fixed object and load is applied by the motors of the joints while measurements of torque and position of the motors are performed by the robot’s own sensors. These measurements are then used to identify the stiffness parameters of the robot. The purpose of this thesis is to develop a model library written in the modeling language Modelica to be used for robot simulations, and especially to simulate these clamping experiments. In the modeling, focus will lie on dynamical behaviour like compliance, backlash, and friction and how these effect the results of the clamping experiments. The modularity aspect has been a central part in the progress and the model library developed is component based, i.e., the parts can be connected in various ways to model different types of robots. For model validation, many simulations have been performed with different dynamic model parameters to evaluate the results and how they compare to the expected dynamic behaviour. Also, experimental data from experiments with ABB’s robot IRB140 have been used to judge how well the models can reproduce results from real experiments. The simulation results indicate that the models capture much of the dynamical behaviour in a satisfying way. The effects of joint compliance, friction, backlash, compliance in the clamping device, and variations in the clamping configuration are all of importance for the results of the clamping simulations. When comparing to data provided from clamping experiments of the IRB140 robot, the results of the simulations show that the models are able to reproduce much of the robot’s behaviour observed in the experiments