Distributed virtual leader moving formation control using behavior-based MPC

In this paper we present a virtual leader formation control algorithm in which agents collaboratively adapt the formation and its motion in a distributed fashion. The formation is defined in terms of scaling and rotation parameters as well as agent offsets from of a virtual leader. The leader is virtual in that it is purely an artifact of a distributed optimization framework. Adaptation of the formation and control of the virtual leader is achieved through an MPC framework, which allows individual dynamic constraints to be considered in the optimization. By also defining the virtual leader motion in terms of a parameterized controller, distributed parameter optimization techniques are employed to solve the optimization problem at each iteration of the MPC algorithm. This allows agents to respect the underlying network constraints, forming a distributed implementation which scales well to large numbers of agents.