A Framework of Multiagent Systems Behavioral Control Under State-Dependent Network Protocols

Abstract

In this investigation, a unifying framework for designing distributed linear-like and nonlinear state-dependent protocols to control the behavior of the multiagent systems over communication networks is presented. Various behaviors of the connected agents, mainly on undirected graphs, are achieved by well-designed couplings where consensus, formation, and deployment are some examples of the achievable behaviors. Composite behaviors stored in a behavioral bank can be selected by a suitable behavior selection mechanism that is controlled directly by the agent embedded artificial intelligence or indirectly through a mission planning utility. The detailed structure of the framework is provided, where the integration of first integral and nonlinear eigenvalue approaches constitute its core is presented. The nonlinear protocols can be converted into their equivalent linear-like protocols by applying a simple transformation. However, the needed conditions will be the same in both cases. Consensus protocols-both linear like and nonlinear-to achieve arithmetic, geometric, and harmonic means are covered. A generalization using the mean-of-order-p is also provided. Stability and convergence issues are mainly handled using properties of M-matrices and Lasalle's principle. The use of state-dependent parameterization to control behaviors is presented. The multitude of primitive behaviors is used to build a more sophisticated behavioral bank that resides in each agent such that an agent can choose or follow the recently active behavior.