Adaptive Control with Multiple Models, Switching and Tuning
Keywords:Learning and adaptive control
Description:Logic-based adaptive control is emerging as an alternative to adaptive control with continuous tuning, thanks to its capability of guaranteeing fast adaptation together with scalability and modularity properties. Adaptive control with multiple models is a particular form of logic-based adaptive control, where an uncertain plant is controlled with a supervisory logic by switching in feedback with it one element from a family of (precomputed) candidate controllers.
A high degree of adaptation and reconfiguration is a fundamental challenge in the control of uncertain systems composed of multiple operating modes, possibly subjected to faults or changes of the system dynamics and of the operating conditions. Applications of these systems cover control over networks (e.g. congestion control over computer networks), fault-tolerant control (e.g. in unmanned autonomous vehicles) and robotics (e.g. self-reconfiguring robots changing their shape to adapt to the operating environment).
Adaptive control with multiple models has shown the potentialities to provide with adaptation and reconfiguration capabilities in systems working in multiple operating regimes. This methodology, although promising, requires further investigations with respect to several aspects. The aim of this MSc proposal is the development of novel logics-based adaptive methods for control of uncertain systems possibly enhanced with: systematic methods for synthesizing in a structurally optimal sense the multiple models for systems of arbitrary scale and complexity; or learning methods to efficiently cope with new operating modes for which no multiple models had been synthesized. Further investigations are also required to extend the proposed methods to systems subjected to time delays, input nonlinearities and other input/state constraints.
Prerequisites: research oriented attitude