06 Juillet – Thesis defense - Mélodie Ruhnke

09 h Amphi Jean-Paul Dom - laboratory IMS / building A31 (University of Bordeaux - Talence campus)

Study of diagnosticability for the reconfiguration of a hierarchical and generic control law for autonomous vehicles in urban environment.

The main subject of this thesis is the development and design of supervising algorithms for automated driving (of level 3 or more) in urban environment. Increasing autonomy of vehicles has progressively reduced the role of the driver. If a driving function is unable to correctly deal with a given situation, detection of the system’s inadequate operation becomes a main need in order to guarantee a reconfiguration towards a safe operating mode. The main objective of the present work is to develop supervision functions for a given autonomous vehicle architecture in order to achieve two tasks: to guarantee functional safety of the autonomous driving functions and to also guarantee passenger’s safety. The main emphasis is proposing fault detection functions and reconfiguration in the specific case of path planning and speed control for an ACC.
The first part of this thesis deals with diagnosis theory as well as reconfiguration. A state of art presents different fault detection and isolation (FDI) methods, which is then completed by a feasibility analysis of such methods for the automobile case. Analysis for the two chosen applications in this thesis leads to the choice of a diagnosis based upon interval construction for path planning and dynamic parity space for speed control.
Reconfiguration problem implies taking into account conformity, signal continuity and/or stability. These three aspects must be considered for decision making, path planning and control algorithms. Essential tools for reconfiguration handling are presented, most notably principal theorems concerning switched system’s stability. This allows to state a theoretical background that will then lead to the demonstration of reconfigured speed controller’s stability, which is the conclusion of this section.
The second part of this thesis is the application of diagnosis methods and reconfiguration principles presented in previous chapters for two cases: speed control and path planning. For the speed control scenario, a fault on a CRONE controller is detected by using dynamic parity space method. Once the default is detected, a reconfiguration towards a PI controller is performed in order to guarantee the safety of the speed control. An analysis on both passivity and stability allows to conclude this study by assuring reconfigured controller’s stability.
Three scenarios are analyzed for path planning: a fault case and two cases of operational domain exit scenarios. For the first case, a calculator overload unable the path planning on the allowed time. The objective is to detect such fault and reconfigure towards an unloaded calculator, which keeps path continuity. On the second case, path calculation algorithm (A*) proposes a solution that implies leaving the road limits. The detection function supervises the distance between a point of the vehicle and the edge of the road. Reconfiguration allows to change the path generation algorithm and keep path continuity. The third scenario deals with a speed profile which disrespects admissible comfort accelerations, which is dangerous for road tracking. The detection function must identify an operational domain exit for the transversal acceleration in the case of a strong curvature turn. Reconfiguration reduces the acceleration in order to keep road tracking and respect acceleration limits.

Event localization