Section: New Results

Stability analysis for controller switching in autonomous vehicles

Participants : Francisco Navas, Imane Mahtout, Fawzi Nashashibi.

This work investigates the Youla-Kucera (YK) parameterization to provide stable responses for autonomous vehicles when dynamics or environmental changes occur. This work explores the use of the YK parameterization in dynamics systems such as vehicles, with special emphasis on stability when some dynamic change or the traffic situation demands controller reconfiguration:

• YK parameterization provides all stabilizing controllers for a given plant. This is used in order to perform stable controller reconfiguration. Different YK-based control structures are obtained for dealing with problems such order complexity, plant disconnection or matrix inversability. Stability properties are preserved even if different structures are employed, but transient behavior between controllers changes depending on the employed YK-based structure. One of the structures presents the best transient behavior without oscillations, a lower order controller complexity and no need to disconnect the initial controller, which would be important if the system shutdown is very expensive, or the initial controller is part of a safety circuit [28]. This structure is used together with CACC applications improving CACC state-of-the-art. An hybrid behavior between two CACC controllers with different time gaps is explored by means of the YK parameterization, in order to avoid ACC degradation when communication link with preceding vehicle is lost. The proposed system uses YK parameterization and communication with a vehicle ahead (different from the preceding one) providing stable responses and, more interestingly, reducing intervehicle distances in comparison with an ACC degradation. A similar idea of hybrid behavior between CACC controller with different time gap is developed for entering/exiting vehicles in the string. In that case, YK parameterization is able to ensure stability of these merging/splitting maneuvers.

• Dual YK parameterization provides all the plants stabilized by a controller. This is employed for solving CL identification problems, or adaptive control solutions, which integrate identification and controller reconfiguration processes. YK-based CL identification uses classical OL identification algorithms, providing better results than if it is used alone. Results in a CACC-equipped vehicle prove how CL nature of the data affects a classical OL identification algorithm, and how dual YK parameterization helps to mitigate these effects. Finally, an adaptive control application is developed by using MMAC. Longitudinal dynamics of two vehicles in a CACC string are estimated within a model set, so the good CACC sytem can be chosen even if a heterogeneous string of vehicles is considered. Dynamics estimation results much more faster than other estimation processes in the literature.

• Different types of controllers and structures are used throughout Francisco Navas thesis ([10]), proving the adaptability of the YK parameterization to any type of controller. Simulation and experimental results demonstrate real implementation of stable controller reconfiguration, CL identification and adaptive control solutions dealing with dynamics changes or different traffic situations. The author thinks that YK is a suitable control framework able to ensure responses in autonomous driving.

• In [27] a design and implementation of a novel lateral control approach is proposed within Imane Mahtout thesis work. The control strategy is based on Youla-Kucera parametrization to switch gradually between controllers that are designed separately for big and small lateral errors. The presented approach studies the critical problem of initial lateral error in line following. It ensures smooth and stable transitions between controllers and provides a smooth vehicle response regardless of the lateral error. For an initial validation the work was tested in simulation, implementing a dynamic bicycle model. It has also been tested in real platforms implementing an electric Renault ZOE, with good results when activating the system at different lateral errors. Current work is focused on using YK-parametrization in estimating lateral vehicle dynamics.