Abstract : 

Cooperative control of multiagent systems has been heavily studied in the past decade. Due to their ability to model many real world systems, it has been a popular trend to use nonlinear Euler-Lagrange equations to model the dynamics of each agent. In the past, feedback linearization or a linear parametrization has been used to control networks of Euler-Lagrange agents.

This presentation explores model-independent algorithms for coordinating networks where the Euler-Lagrange agents have unidirectional interaction constraints (directed graphs). It turns out that for stability, and due to the non-Lipschitz, nonlinear agent dynamics and the directed communications, we must centrally design sufficiently large control gains using global information. The disadvantage of centralised design is offset by the robustness and simplicity of the algorithms, and exponential stability properties. Algorithms for leaderless consensus, rendezvous to a stationary leader, and trajectory tracking of a moving leader are presented. Extensions to the fundamental algorithms are presented, involving event-based control and coordination on dynamic topologies.