Structure preserving model reduction for multi-agent systems with port-Hamiltonian systems dynamics

A multi-agent systems is a system containing multiple subsystems (agents). The behavior of a multi-agent system contains both the dynamics of the individual agents and the communication topology of the network that specifies interacting information among the agents. In the last decades, the problems of cooperative control and optimization for the multi-agent systems have received compelling attention and interest from the system and control field. The theoretic research on this topic has a broad application in industrial, military and civil areas, such as formation control of multiple unmanned aerial vehicles, spatial distribution of satellite constellations on certain orbits, management of wireless sensor networks or automation of the power grid. However, these systems increasingly lead to large-scale systems, which are difficult for model analysis, online simulation and system control if using their high-order mathematical models directly. Thus, it is of clear importance to have methodologies to replace such a complex and high order systems with lower order and simpler models in order to accelerate system simulations, simplify controller design and improve the feasibility of controller implementation. However, it is also desirable for many properties of the original system to be preserved in a low order approximation, both from an analysis and a control design point of view. Therefore, this research proposal aims at exploiting suitable model reduction techniques to obtain a simpler, lower order model for the multi-agent system that preserves the structure of the original system.