Neighbor Selections and Impacts on Vehicle Flocking Control with Time Delays and Unknown Disturbances

Vehicle flocking is an advanced cooperative control strategy that enables a group of vehicles to move cohesively on multi-lane roads. A neighbor selection rule defines how to identify interactive neighbors to shape the flocking topology and thus influence the flocking dynamics. This paper investigates, for the first time, three different neighbor selection methods and their impacts on vehicle flocking control by considering time-delayed information and unknown environmental disturbances. The K-nearest rule generally gives a dynamic flocking topology as neighbors change over time. When K ≥ 6, the K-nearest rule is equivalent to the well-known distance-based rule, which creates an undirected and fully connected graph that allows individual oscillations to spread system-wide. The field-of-view (FOV)-based rule forms a directed graph with fewer graph edges and disturbance propagations than those of the distance-based rule. When time delays and external disturbances affect 5 randomly selected agents out of 9, simulation results show that the FOV-based rule is more robust because the directed graph structure in FOV inherently curtails the propagation of these oscillations from the rear to the front agents.