Consensus of linear multi-agent systems with multilayer network using output feedback controller

Abstract

Recently, there is a need for consensus of Multi-agent systems (MAS) in various engineering fields such as unmanned vehicles, cluster robot control, and sensor network. This paper is inspired by the disconnected autonomous vehicle and Connected Car. Connected cars not only get information through sensors like ordinary cars, but also receives various information of roads and traffic situation by wireless communication between vehicles or infra to increase safety of unmanned autonomous driving. However, as a common sensors in typical automobiles have limitations in physical situations such as a limitation of a measurement distance, it is common that not all agents can get information even though all the agents have same sensors. If information between vehicles are not obtained, it may cause great danger during autonomous navigation. Therefore, when the sensor can not be measured, information should be exchanged through wireless communication to realize stable unmanned autonomous driving. Therefore, it is necessary to study the control of MAS consensus considering these realistic problem situation. Many recent studies on consensus of MAS solve the problem assume all sensors or agents have a single communication graph without considering the characteristics of sensors, and it does not take into account realistic limitations. A communication graph is a structure of connectivity between agents. In order to solve the consensus problem in reality, it is necessary to consider the structure and connectivity of the communication graph according to the type of sensors. Therefore, in this thesis, we present how to set up situations that can occur when controlling MAS to achieve consensus by using a multilayer network. By using structure of the multilayer network, it is possible to separate and analyze several graphs having different structures. In this thesis, we classify sensors as sensors that measure relative information and sensors that provide absolute information of agents through wireless communication such as GPS. We refer to them as relative sensors and absolute sensors. Various types of relative sensors have different physical limitations and therefore have different communication graphs depending on the types of sensors. Also, due to these physical limitations, the connectivity of the communication graphs can not be guaranteed either. Therefore, we introduced the concept of multilayer network to analyze several unconnected graphs and showed consensus of MAS by designing a dynamic controller. In this thesis, general MAS are analyzed using Multilayer network, and consensus problem is solved mathematically by presenting a dynamic controller. The contribution of this thesis can be summarized as follows. • Typical results of consensus mainly considered one communication graph. However, this paper considers multiple communication graphs that are not guaranteed to be connected. • In MAS, only two layers are considered in the result of using structure of multilayer network. In this thesis, under suggested assumption, the number of layers is not limited. • Dynamic controller using output information is presented for consensus. Through these theoretical results, we expect to provide theoretical results solve consensus problem more efficiently.