Adaptive Safe Data Driven Control Strategy for Closed Loop System

Due to the widely studied data driven control in today’s new big data era, this paper combines other innovative considerations into our existed contributions about data driven control, for example, adaptation, safety and switching logic together, so that an innovative adaptive safe data driven control strategy is proposed to complete its entire research from the theoretical perspectively and practical application. More specifically, after reviewing one closed loop system with the unknown plant and unknown controller simultaneously, data driven control strategy is used to design one controller without any prior information about the unknown plant, while achieving the perfect tracking or model matching. Then the concept of safety is defined to constrain the unknown controller or its parameterized controller parameters, and consequently adaptive idea is introduced to design one controller, generating in its recursive form for nonparametric or parameterized controller respectively. Within our both defined safety and control mission, adaptive safe data driven control is turned into one inequality constrain optimization problem. Because of a special feasible region, i.e. a box constraint, the gradient projection algorithm is proposed to solve the optimal controller parameters, being dependent on the measured input-output data. Moreover from an applied point of view, we adjust our mentioned adaptive safe data driven control to its improved adaptive safe switching mechanism, more suited to practical engineering, i.e.unmanned helicopter automatic taking off and landing. Finally, some simulation results are given to show our theoretical results on adaptive safe data driven control and its improved switching form. Generally, this new paper extends our previous contributions about data driven control to complete its synthesis research, combining some advanced factors together from both the academic theory and practice, for example, adaptation, safety, optimization theory and algorithm, switching logic together.