Optimal Motion Control of Mechatronic Systems: Contributions to Iterative Learning Control and Efficient Trajectory Optimization (Optimale bewegingscontrole voor mechatronische systemen: bijdragen tot iteratief lerende controle en efficiënte trajectoptimalisatie)
Optimal Motion Control of Mechatronic Systems: Contributions to Iterative Learning Control and Efficient Trajectory Optimization
As a result of increasing customer expectations, the continuing miniaturization of components, and the constant pursuit for cost-effective production methods, many mechatronic systems are facing ever more stringent specifications with respect to positioning accuracy, production speed, energy consumption, etc. This evolution has led to the introduction of optimization methods in machine, controller, and motion trajectory design to obtain superior mechatronic systems that operate at a maximum efficiency. In optimal motion control, the goal is to optimize the control signals and the corresponding motion trajectory according to a certain performance criterion while considering the system dynamics and limitations. At the same time, the repetitive nature of many industrial processes has led to the development of iterative learning control (ILC) techniques, which aim at improving the tracking accuracy of a system performing the same task repeatedly. ILC algorithms iteratively solve an optimal motion control problem by measuring the tracking error from one iteration and using it to update the system input for the next iteration. The main advantage of ILC over model-based trajectory optimization is that ILC allows to obtain accurate tracking without having an accurate system model. Although the successful applications of ILC are numerous, traditional ILC algorithms still have several drawbacks, which limit the applicability to systems operating in a repetitive manner.The first part of this research contributes to the field of optimal motion control of mechatronic systems by extending and facilitating the applicability of ILC techniques in three different ways. First, a data-driven version of the norm-optimal ILC framework for constrained linear time-invariant systems is presented. The data-driven norm-optimal ILC algorithm does not assume any a priori system information aside from the fact that the plant is linear time-invariant and hereby eliminates the need for a time-consuming identification procedure. Second, the applicability of ILC is extended to other optimal control problems than tracking control, such as energy-optimal and time-optimal point-to-point motion control problems, and time-optimal path following problems. Third, an ILC initialization method based on a previously learned similar trajectory is developed to obtain better performance from the first iteration and hence, also faster convergence of the ILC.The second part of this research contributes to the field of optimal motion control of mechatronic systems by developing computationally efficient trajectory optimization methods. The first contribution focusses on energy-optimal control of linear-time invariant systems performing a series of point-to-point motions. Not only the system input, but also the time allocated to each motion is optimized such that the complete series of point-to-point motions is performed in an energy-optimal way. The second contribution is the development of a computationally efficient algorithm to solve time-optimal point-to-point motion control problems for discrete-time linear time-invariant systems with linear system constraints. This algorithm can, for instance, be applied in norm-optimal ILC for time-optimal point-to-point motions to reduce the computational cost of updating the system input.