|SABUG LORENZO JR||Cycle: XXXV |
Section: Systems and Control
Tutor: PIRODDI LUIGI
Advisor: FAGIANO LORENZO MARIO Major Research topic
:On data-driven optimization techniques in the design and control of autonomous systemsAbstract:
In most cases, the design and planning of autonomous systems increase in complexity as more factors of different nature are taken into consideration. Actuator specifications, joint responses, and mechanical dimensions can have interactions which affect performance in a non-trivial manner. Hence, an explicit mathematical model linking these factors to the performance or an overall cost function can be difficult, or almost impossible. This is more often the case when electromagnetic, hydrodynamic, or chemical interactions are involved in the system process. In optimizing such systems, cost or objective functions are evaluated by performing time-consuming simulations, resource-intensive experiments, and/or empirical measurements, and are termed black-box functions. Because of the huge overhead in performing simulations and experiments to evaluate the black-box function at specific points, it is of best interest to strategically choose the next test point(s) according to previous data. The goal of this research is to investigate optimization techniques that can be used on such applications, through a blending of learning the black-box function and exploiting the available information from previous data.