Consensus in multi-agent systems with switching topologies using position based back-stepping

This paper focuses on the consensus problem in a decentralized multi-agent system with switching communication topologies. Consensus of agents can be reached for centralized dynamically switching topologies when the union of all topologies forms a spanning tree graph. In case of decentralized control, the topology must form a strongly connected graph to reach consensus. A new algorithm, which uses “back-stepping” of agents, is proposed in this work which assures consensus whenever the topology is not strongly connected. The algorithm ensures that when an agent reaches a position where if loses communication with others and the topology is no more a strongly connected graph, it back-tracks and take a different path to reach consensus. Simulation results show that after introducing back-stepping, each of the agent´s path is chosen to ensure that communication with other agents is maintained and they come to a consensus in minimum possible time.