Safety of passengers and surrounding road users is the most important challenge in the design and deployment of autonomous driving technologies. In fact, the highest Automotive Safety Integrity Level (ASIL-D) will likely be required for autonomous driving functionalities. While fulfilling such safety requirements involves special design efforts at all levels of the autonomous driving stack, in this talk we will focus on the control design of a safe motion planner in urban environments. We will start by illustrating a model-based control design approach to the vehicle motion planning problem, in presence of human road users (pedestrians, cyclists, human-driven vehicles). We will show that, under mild assumptions, the vehicle behavior can be made cautious in presence of road users and guaranteed to be persistently safe. Experimental results obtained with a passenger vehicle negotiating an intersection with a simulated pedestrian, will be shown. An important ingredient of the proposed motion planning framework is a prediction model of the surrounding traffic. In the second part of the seminar, we will illustrate our ongoing research on humans’ intent prediction in traffic environments. We will show how the evolution of a traffic scene can be predicted using very simple models and motion data (position, velocity) of road users observed in similar traffic scenes.
Join at: imt.lu/aula2