Multiagent Coordination Optimization: A Control-Theoretic Perspective of Swarm Intelligence Algorithms

In this paper, a new swarm optimization algorithm, called Multiagent Coordination Optimization (MCO) is developed, which is based on the Particle Swarm Optimization (PSO) and cooperative control of multiple agents, to optimize general nonlinear objective functions for unconstrained optimization problems. The standard PSO algorithm needs a local optimal position to achieve the global optimal position, which is generated by the minimum objective value between the current local optimal position and current actual position. Different from this, here we use a consensus-based term to calculate the local optimal position for MCO, which is widely studied in the network consensus problem. More importantly, the global convergence result for MCO is presented by aid of some tools from dynamical systems and control theory. We propose two deterministic dynamic models to approximate the intrinsic, averaging dynamics of MCO. Some standard objective test functions are used to evaluate the performance of our algorithm and the proposed dynamic models. Finally, by using this algorithm, we solve an optimal distributed linear averaging problem and a sensor network based parameter estimation problem for threat localization.