###Learning-based Model Predictive Control and its application to Smart Mobility Systems
Project ID: 2228bd1217 (You will need this ID for your application)
Research Theme: Engineering
UCL Lead department: Electronic and Electrical Engineering (EEE)
Lead Supervisor: Boli Chen
Project Summary:
Control, in the era of Cyber-physical Systems, faces new opportunities and challenges. As systems are becoming more complex and data is becoming more readily available, scientists are beginning to bypass classical model-based techniques in favour of data-driven methods. Data-driven methods are suitable for applications where first-principle models are non-trivial to identify accurately (e.g., in human-in-the-loop applications). Nevertheless, data-driven control methods usually cannot provide stability and robust guarantees during the learning and decision-making processes, which is the key to safety-critical applications. Model predictive control (MPC) has been successfully applied to modern control applications due to its ability to provide high performance and explicitly consider constraints. However, MPC relies on an accurate parametric state space model of the system, which may hinder its successful implementations in practice. In this project, we will investigate how to integrate data-driven and learning-based methods in MPC so that an accurate parametric system model is not required, and how to ensure stability and robustness when both are combined. Furthermore, when systems exhibit fast time-varying dynamics, online methods are particularly suitable because they operate on a limited amount of data and adapt quickly to new system dynamics. The intended research will bridge the gap between two different control philosophies – adaptive control and optimal control. Learning-based MPC methods will find many real-world applications. Particular emphasis will be given to control of connected and autonomous vehicle (CAV)-dominated traffic network. The gradual deployment of CAV leads to the transition phase of mixed traffic with the coexistence of human-driven vehicles and AVs, where the human-in-the-loop vehicles are hard to model due to the uncertain and stochastic human driver behaviours. The proposed control solution with non-parametric learning offers great flexibility in dealing with such a challenge and can strike a balance between multiple objectives, including energy efficiency, mobility, and safety guarantees.