One step ahead quadratic receding horizon distributed control of multiagent formations

A one step ahead optimization based quadratic receding horizon control of autonomous agents is proposed. Each agent is modeled as a double integrator. The goal is to force the agents to converge to a formation specified by a set of linear algebraic constraints. We specifically extend our earlier work that involved trajectories in which velocity and position constraints are independently specified. In this paper we seek to widen this class of trajectories to settings where the underlying constraints involve position and velocity interdependence. We characterize the class of such trajectories that can be stably attained and maintained by one step ahead optimization based quadratic receding horizon control. As in our earlier work we only define a geometric topology for the agent formation, and by correctly choosing the cost function, show that our algorithm produces a communication topology mirroring the geometric topology. By providing some redundancy in the formation topology it is possible for the system to survive the loss of an agent. Other attractions of the scheme are scalability, the requirement of only local knowledge of the desired formation topology and ease of reconfiguration in the face of loss of agents and/or channels.