Design of autonomous control systems for a Driverless race car: Remote Emergency System and Autonomous State Machine

Autonomous driving is the most challenging research field in automotive in recent years, alongside emissions reduction. Particularly, the progressive reduction of the human driver’s contribution in the vehicle control loop, started with the first milestones in the active safety field as ABS and ESP and with the aim of completely get rid of the human driving, is increasing constantly the road safety eliminating the human error. Since 2017, Formula Student competition is including a Driverless class, too, in order to stimulate students and future engineers to face the design and realization of a Driverless race car prototype. The specific rules of this class regulate the design and implementation of the crucial subsystems for a driverless vehicle: the perception, the actuators, and the safety components. This thesis work is a contribution to the development of the Driverless prototype by the Polytechnic University of Turin. Most specifically, the first part concentrates on the communication and testing of the Remote Emergency System (RES), which is a radio component useful to give remote commands to the vehicle. It is really important since it can activate the Emergency Braking System (EBS), if the team remotely notices issues during the autonomous mission. The second part focuses on the design of the Autonomous System state machine, which is the upper supervisor of the vehicle decision-making part because, on the basis of the inputs from sensors and remote commands, it is in charge of activating the actuators and the specific control algorithm for the selected mission. The work followed a step-by-step methodology, starting from the theoretical bases and the simplest model, and adding improvements just once the previous updates were verified. However, the work on RES started directly with physical testing to rebuild its behaviour using software as Busmaster, Simulink Vehicle Network toolbox, and Vector CANdb++. Concerning the state machine instead, the design is all virtual in MATLAB and Simulink Stateflow environments, giving physical meaning to the variables in order to simplify the subsequent implementation in the vehicle control unit.