Validation of a virtual prototyping method using computed torque and ILC as controllers on a parallel robot

摘要

This thesis validates a new method of virtual prototyping: a macro that automatically generates a dynamic model from CATIA V6 and exports the model to SimMechanics. Until now, engineers who need to verify their design with a dynamic model have either had to calculate it by hand, or manually input all the necessary parameters in SimMechanics or other software. Both of these methods are time-consuming and often lead to mistakes. To demonstrate the relative difficulty of the hand-calculation method, we begin by presenting our dynamic model of a four-bar parallel robot that we calculated using the Lagrange method. At this stage we also calculated a computed torque controller that is used later when comparing the performance of the macro-generated dynamic model to the actual robot. The rest of our thesis shows how the new macro replaces both the hand-calculation and manualentry methods of dynamic modeling, and vastly simplifies the task of calculating the computed-torque controller. We used the macro to export a dynamic model to SimMechanic, where we then added a sensor and actuator to complete the dynamic model of our four-bar parallel robot. The macro-generated model was tested in simulations using three combinations of controller: computed torque alone, computed torque with P-type iterative learning control (ILC), and computed torque with PD-type ILC. The third combination produced the best results, that is, the lowest error value. To validate the performance of the model, we then tested the performance of the real robot using the same three combinations of controllers. Note that the computed torque used here is the one that was generated by the macro. Again the results show that the combination of PD-type ILC and computed torque functions best. However, the error was bigger in the practical experiment than it was in thesimulation, which used the macro-generated dynamic model. This difference is likely because our dynamic model does not consider several factors that affect the real robot, such as friction, the mass of screws and bolts, the moment of inertia of the rotor and pulleys, and the stiffness of the timing belt. For the sake of simplicity, our dynamic model was not intended to take all these factors into account. Therefore, the fact that the simulation results closely match those of the practical experiment serves to validate this new method of virtual prototyping.