Flocking of Multi-Agent Systems Via Model Predictive Control Based on Position-Only Measurements

The information of both the position and velocity of agents are required in most existing flocking algorithms. This paper studies the model predictive control (MPC) flocking of a networked multi-agent system based on position measurements only. We first propose a centralized impulsive MPC flocking algorithm and further develop a feasible sequential-negotiation based distributed impulsive MPC flocking algorithm, where each agent sequentially solves a local optimization control problem involving the states of its neighbors only. We prove that both the centralized and distributed impulsive MPC flocking algorithms lead to a stable flock by using geometric properties of the optimal path followed by individual agents and provide numerical simulation examples to illustrate their effectiveness and advantages in convergence rate and communication cost.