We are very proud to congratulate Evyatar Matmon on successfully passing his MSc examination!
Evyatar’s thesis, Leader Identification of Multi-Agents Systems Under Semi-Autonomous Consensus Protocol, contributes to the problem of network identification in multi-agent networks. A challenging yet important problem in the network sciences is to be able to reconstruct the interaction topology of a multi- agent system using only observed data from the agent trajectories. Evyatar’s thesis focuses on a sub-problem of the network identification problem, aimed at identifying the leader agents in a semi-autonomous network. The setup assumes a connected network with agents split between leader nodes and follower nodes. It is further assumed that all the follower nodes are running a consensus algorithm.
The contributions of the thesis can be split into two parts. The first studies the structure of leader-follower graphs and characterizes a class of graphs that ensure a sufficient separation between the entries of the Fiedler vector corresponding to the leader nodes and the follower nodes. In this direction, an algorithm is proposed that generates such graphs with these conditions. This is an interesting and original contribution providing new insights on the relationship between graph structures and Fiedler vectors. The second part of the thesis leverages an existing result from the literature where the Fiedler vector can be estimated by data generated from consensus protocols using the notion of the relative tempo of the graph. This leads to a natural algorithm for identifying the leader nodes based on the earlier results of the network.
Congratulations, Evyatar, on this excellent work!