Best Paper Award

  In recent years, control and analysis over large-scale networks have had major impacts on everyday living. For example, the Internet of Things (IoT) and big-data analysis have become two of the most active research areas in engineering. Over the last 10 years, cooperative control of multi-agent systems has attracted considerable attention in control engineering as a key technology for distributed control of such large-scale network systems.
   The average consensus is a fundamental problem in the area of control of multi-agent systems. In average consensus problems, agents cooperatively update their states to achieve convergence to the average of the initial values through local communications with neighbor agents. There has been a substantial volume of research articles on average consensus problems such as vehicle formation, analysis of migration of wild animals, sensor networks, and smart-grids. However, convergence of existing average consensus dynamics is relatively slow. Thus, accelerated methods are required in a number of applications whilst maintaining distributed structures of local communications among agents.
   In this paper, we propose average consensus dynamics with 2-hop communications in which each agent uses the states of 2-hop neighbors as well as those of 1-hop neighbors. We show that the proposed dynamics improve convergence speed compared to that of conventional 1-hop dynamics. We also show that the convergence speed of the proposed dynamics depends on the algebraic connectivity (the second smallest eigenvalue) of a graph Laplacian that represents communications among agents. In addition, we show that the proposed 2-hop dynamics can be easily extended to a more general class of multi-hop dynamics.
   In summary, this paper proposes novel 2-hop average consensus dynamics for faster convergence. The proposed method is expected to give deep insights into cooperative control of multi-agent systems and emerging fields of network science.