15 Décembre – Thesis defense - Julien Moreau.
09 h30 Amphi Jean-Paul Dom - Laboratory IMS / Building A31 (Talence campus)
Driver assistance systems and connected autonomous vehicle at intersections.
The main goal of this thesis is to help removing a scientific barrier in urban environment navigation. Thus, the trajectory planning must allow the generation of reference to be followed by the autonomous vehicle respecting the constraints of the road infrastructure, while guaranteeing a capacity to react to potential obstacles. Therefore, the trajectory planning must allow reference generation to be followed by the autonomous vehicle respecting road infrastructure constraints, while guaranteeing a capacity to react to potential obstacles.
After a brief history on autonomous driving, the specificities of the urban environment are listed. It is shown that the urban environment is an uncertain environment, with high curvature roads and low speed traffic. The objectives and constraints associated with trajectory planning are outlined and a path and speed decomposition approach combined with a complementarity between predictive and reactive methods is proposed.
Predictive path planning is concerned with determining the path to be followed by the autonomous vehicle within a number of known constraints in advance. A reference path, reachable by the nonholonomous vehicule and guaranteed in the free space at all time, is generated. The generated path does not take into account the variation of the environment and the vehicle is then led to avoid obstacles preventing the follow-up of this planned path. These obstacles can be dynamical or static, for example a parked vehicle encroaching on the track.
In the case of static obstacles, a new path to be followed by the vehicle is generated through a path rescheduling algorithm. A speed profile is associated with the reference path, taking into account the safety, passenger comfort and vehicle limitations, including the actuator ones.
The autonomous vehicle may also have to avoid a number of dynamic obstacles. In this case, the speed of the ego-vehicle is adapted according to the situation is then paramount in order to avoid a possible collision.
Finally, a simulation platform developed with the SCANeR Studio driving simulation software and Matlab/Simulink software is used to test these algorithms in complex situations.