Albarella, Nicola (2022) Control Architectures for Advanced Driving Assistance Systems. [Tesi di dottorato]

Preview

Text Albarella_Nicola_XXXV.pdf Download (10MB) | Preview

Collection description

The rapid economic growth has led to an increasing number of vehicles on the road, thus increasing the number of road accidents as well. This issue is a serious dilemma, laying economic burdens on governments, as well as, safety problems on people. Advanced Driving Assistance Systems (ADAS) are software modules assisting the driver, as the name suggests, in monitoring the environment and controlling the vehicle itself. These modules have been demonstrated to be effective in reducing the rate of collisions, and are the main focus of this thesis. Increasing the level of safety, thus bridging the gap between driving assistance and autonomous driving is a challenging task. While in the former, a safety driver is always there, and ready to intervene if needed, in the latter the driver could even not be on board at all. Therefore, the vehicle must be capable of driving itself, in any scenario, despite the adversity of the environment (e.g. road asphalt condition, weather, etc.), the uncertainty in sensor measurements and the complex interactions with other road users. This thesis tackles this problem from multiple points of view. First, it is shown how, by properly design state of the art ADAS, e.g. by endowing these with additional environment information, it is possible to enhance the overall safety. Moreover, a novel motion planning control architecture is presented. By properly combining the latest advancements in Machine Learning and Optimal Control, safe, effective and scalable driving policies can be learned from data. It will be shown how, by making safety formally explicit, constraints can be put on Machine Learning techniques, thus increasing both performances and safety.

Downloads

Actions (login required)

View Item