Design and implementation of a protocol for safe cooperation of self-driving cars

Tese de mestrado, Engenharia Informática (Arquitetura, Sistemas e Redes de Computadores) Universidade de Lisboa, Faculdade de Ciências, 2019Ever since its introduction, the car has fundamentally changed our society. Its popularity grew tremendously in the early 20th century, and today it is nearly ubiquitous. However, there are several problems related to automobiles, one of the major ones being road congestion. Drivers lose millions of dollars every year in fuel costs and time spent in day to day traffic congestion. Another major problem is emissions from vehicles, forming a significant percentage of greenhouse gas emissions. Automated driving systems currently rely on their own sensors to gather information from the real-world, and make informed decisions to keep their passengers safe. But sensors may not be sufficiently accurate during all conditions and can even fail, so automated driving systems take this into consideration when controlling the car, leading to larger safety margins. Vehicle-to-Vehicle communication can enable cooperation between vehicles which, among other things, can be sending or receiving information from other nearby vehicles, increasing the confidence level in the information gathered or even gathering information otherwise not obtainable. Cooperation opens the door to complex vehicular applications such as road-trains (or platooning) and virtual traffic lights, both of which have the potential to mitigate the problems mentioned before. These applications have tight safety requirements due to the context in which they operate: vehicles, possibly with human occupants, operating on roads traversed by non-autonomous vehicles and pedestrians. In this MSc Dissertation, we describe the implementation of a vehicular cooperation algorithm backed by both vehicle-to-vehicle communication and a cloud membership service. We then evaluate the implemented algorithm in a cooperative environment to conclude about its correctness, making use of the Robot Operating System middleware to implement a simulation and visualize a maneuver executed using the algorithm