Gossip Based Consensus Algorithm for Multi-Agent Systems

Multi-agent systems presently find extensive applicability across diverse domains. Central to these applications is the imperative for agents to exchange information regarding their experiences, knowledge, and actions. Distributed algorithms, denoting algorithms employed by multiple agents within a system to collaboratively execute a particular task, play a pivotal role. Within multi-agent systems, these algorithms facilitate the coordination of agents that are typically autonomous and possess decision-making capabilities. Given the diverse communication graph structures and communication challenges inherent in multi-agent systems employing distributed algorithms, agents necessitate a consensus phase. Conversely, the implementation of distributed algorithms within graphs featuring a substantial number of edges can exert significant strain on the limited resources of wireless nodes. This study proposes a gossip algorithm as a viable solution to the requirement for consensus among agents engaged in decentralized decision-making within multi-agent systems. The suggested algorithm reduce the graph through its scoring structure and fosters consensus among agents via the ranked gossip structure constructed upon this graph. Through simulation studies, the feasibility and effectiveness of the proposed algorithm across varied graph structures are demonstrated.