Distributed Omniscient Observers for Multi-Agent Systems: Design and Applications

TL;DR

This paper introduces distributed omniscient observers enabling each agent in multi-agent systems to estimate all states using local information, facilitating applications like game equilibrium and social behavior emergence.

Contribution

It presents a novel observer design that operates without global graph knowledge, applicable to heterogeneous and homogeneous systems.

Findings

Observers enable accurate state estimation in multi-agent systems.

Simulations show emergence of social behaviors like herding and navigation.

Supports distributed Nash equilibrium seeking.

Abstract

This paper proposes distributed omniscient observers for both heterogeneous and homogeneous linear multi-agent systems, such that each agent can correctly estimate the states of all agents. The observer design is based on local input-output information available to each agent, and knowledge of the global communication graph among agents is not necessarily required. The proposed observers can contribute to distributed Nash equilibrium seeking in multi-player games and the emergence of self-organized social behaviors in artificial swarms. Simulation results demonstrate that artificial swarms can emulate animal social behaviors, including sheepdog herding and honeybee dance-based navigation.