Distributed collision avoidance for autonomous vehicles: World automata representation

The automatic control of interacting autonomous vehicles (AVs) is one of the problems that engineers are currently trying to solve. The present paper deals with the design of local control laws governing the movement and collision avoidance of such groups of AVs, enforcing the safeness of the operations as well as task completions. This problem is inspired by the automation of a container terminal where each AV executes tasks assigned by a supervisor. A task involves moving an AV from an assigned origin to an assigned destination by a given deadline. The constraints imposed by the bounded workspace (a long and narrow quay in the container terminal example), the deadlines assigned to each task, and the uncertainty in the detection and communication make the problem difficult to solve in a centralized way. Therefore a distributed control approach is preferred with a local control agent in each AV adjusting its trajectory, so that its task is completed without collisions. By applying a fixed set of priority rules the computational complexity for each agent is reduced compared to the centralized case. Whenever an AV detects a possible conflict, i.e. the estimated position of another AV within the detection range, it must adjust its own speed and trajectory in order to avoid a future collision, reducing the number of cases where a supervisor has to intervene in order to resolve conflicts that degenerate in dead-locks. The modelling and validation of the system is performed by using the world automata theory.