Konsensusalgorithmen für vernetzte Regelungen

Abstract

This thesis deals with distributed consensus algorithms for multi-agent networks. The goal of consensus algorithms is to achieve a consensus for a specific task. In our case we consider a setting where network agents should meet at a single point. This rendezvous algorithm is well known in literature. We mainly focus on two dimensional random geometric networks, but one dimensional networks and regular scenarios are also considered. For the performance study we pay special attention to the convergence properties and the ability of the agents to reach the consensus point. All tasks carried out by the nodes are distributed, in other words, there is no need of external agents for coordinating or deciding. In addition to the consensus algorithms the effect of using control within these networks is also studied. With these control contributions we try to improve the overall performance of the system. We will try to control the dynamics of the nodes and also apply temporal delays to the movement. Finally, we present some numerical results where we can observe how different settings and parameters affect the behavior of the whole system.