Robust Control for Autonomous Vehicles based on Integral Quadratic Constraints

Abstract

This dissertation focuses on designing robust controllers for autonomous vehicles based on integral quadratic constraints (IQCs). In real-world applications, it is essential to consider dynamic uncertainties and environmental disturbances during controller design. The IQC theorem is a useful tool for dealing with system uncertainties. This theorem unifies all the classical results of robust control theory and can be easily combined with optimization-based control such as linear matrix inequality (LMI)-based control. By using IQC, it is possible to analyze robust stability against a wide variety of uncertainty classes and investigate system performance, such as reachable sets.

This dissertation explores the combination of the IQC theorem with optimal control methods, such as model predictive control (MPC) and control barrier function (CBF)-based control methods. Additionally, the author discusses the robustness and stability of neural network (NN) combined controllers using IQC. These IQC-combined robust controllers have applications in lateral and longitudinal control of autonomous vehicles, including adaptive cruise control (ACC) and lane-keeping systems (LKS). The IQC-based optimal control methods demonstrate their efficiency and scalability in dealing with various uncertainties.