A Framework With Local Interaction Topology And Extended Consensus Protocol For Multi-agent Multi-tasks Control

The multi-agent system refers to the system consisting of multiple agents,who work towards a same goal through cooperation with each other.Multi-agent systems are widely used in many fields such as intelligent transportation,monitoring and inspection,and agricultural production.In practical applications,multi-agent systems often face multi-task requirements in complex scenarios.However,if multiple tasks are modeled separately,the multiagent system would confront the challenges of increasing controllability and cost constraints.Therefore,multi-agent system needs a consistency framework,so as to analyze the stability of multi-task,to save the development cost of multi-agent multitask,and to increase the scope of applications.This paper proposes a framework with local interaction topology and extended consensus protocol for multi-agent multitasks control.First,this paper combines the local interaction topology with the distributed controller.In the local interaction topology,the state information of neighborhood agent and leader is comprehensively considered.In addition,the local interaction coefficient is defined to measure the tightness of agent connection.Compared with the fully connected structure,the local interactive topology achieves lower cost without losing controllability.Second,this paper designs an extended consensus protocol.The relationship between the convergence states of agents is regarded as a rigid body.Using the properties of linear and translation of rigid body,the multi-dimensional convergence state in multi-task is uniformly defined,and an extended consensus protocol is constructed.Unlike the traditional consensus protocol which ignores the transformation relationship between convergence states,the extended consensus protocol is more general and can be extended to multi-dimensional convergence states without redesigning the controlled items of the controller.The stability of the multiagent system is proved with Lyapunov stability principle.Final,this paper presents the experiment of multi-agent formation and rendezvous tasks to verify the effectiveness of the proposed framework.Experimental results show that this framework can uniformly describe different task scenarios.Under various interactive topologies,the experimental results on the error of multiagent system is consistent with that from the theoretical analysis.In addition,experimental results show that the local interactive topology can effectively reduce the communication cost by 69.09% compared with the fully connected topology,while maintaining the controllability of the system.