Cooperative Convex Optimization in Networked Systems: Augmented Lagrangian Algorithms With Directed Gossip Communication
Recently, there has been increased interest in large scale networked systems including networks of agents, wireless ad-hoc networks, and Wireless Sensor Networks (WSNs.) Typically, these systems lack a central unit, and the inter-node communication is prone to random failures (e.g., random packet dropouts in WSNs.) In this paper, the authors consider a generic computational model that captures many applications in networked systems. With this model, nodes cooperate to find the optimal parameter (scalar or vector) of common interest, e.g., the optimal operating point of the network. Each node i has a private (known only at node cost function of x, e.g., a loss at node i if operating at x. The total cost is the sum over the individual nodes' costs.