Computation of safety control for hybrid system with applications to intersection collision avoidance system

Abstract

In this thesis, I consider the problem of designing a collision avoidance system for the scenario in which two cars approach an intersection from perpendicular directions. One of the cars is a human driven vehicle, and the other one is a semi-autonomous vehicle, equipped with a driver-assist system. The driver-assist system should warn the driver of the semi-autonomous vehicle to brake or accelerate if potential dangers of collision are detected. Then, if the system detects that the driver disobeys the warning, the system can override the behavior of the driver to guarantee safety if necessary. A hybrid automaton model with hidden modes is used to solve the problem. A disturbance estimator is used to estimate the driver's reaction to the warning. Then, with the help of a mode estimator, the hybrid system with hidden modes is translated to a hybrid system with perfect state information. Finally, we generalize the solution for the application example to the solution of safety control problem for general hybrid system with hidden modes when the hybrid system satisfies some proposed constraints and assumptions.