When stabilisation and optimisation meet: a codesign approach
Project Leader: Dragan Nesic
Collaborators: Romain Postoyan (CNRS, Nancy, France) Lucian Busoniu (TU Cluj-Napoca, Romania)
Primary Contact: Dragan Nesic (firstname.lastname@example.org)
Keywords: control and signal processing; defence
Disciplines: Electrical & Electronic Engineering
The next generation of engineered systems need to perform complex tasks with precision, and be robust, resilient and adaptive to their environment enabled by the confluence of control, optimisation, learning and computation. Understanding the interplay between robust stability and optimisation is key to this endeavor. Many techniques, such as model predictive control and reinforcement learning, rely on an intricate interplay between an optimisation-based control algorithm and an optimisation routine used to calculate the control law. This project aims to develop a general design framework for stability, suboptimality and robustness of such algorithms, that can be used in range of novel applications, such as driverless cars and drones.
This project will enable the design of advanced control algorithms used in cyber-physical systems by understanding in depth the stabilising, near-optimality and robustness features of optimisation-based control algorithms, such as model predictive control and approximate dynamic programming. The capabilities of these cyber-physical systems will be enhanced considerably by careful designs of their ‘brain’, which can learn about their environment, adapt to it and perform complex tasks with precision.
Operating autonomously, the next generation of engineered systems will be essential for smart highways, driverless cars, swarms of drones, various types of robots and advanced manufacturing systems to name a few examples. It will impact transportation, environmental monitoring and defence, improving our quality of life in overpopulated cities, providing better use of our energy and water, reduce pollution and waste as well as maintaining a competitive edge in the global market.
This project is funded by the Australian Research Council Discovery Grants Program.