Information broadcasting algorithm for Finite-Time Reaching-at-Risk Consensus with application to weapon-target assignment

We propose a randomized algorithm aimed at reaching, in finite time, exact consensus among a set of agents that are linked though a connected, possibly time-varying, graph. The information exchanged among neighboring agents is limited in size, by randomly selecting the content from an agent internal state. The time needed to reach the agreement is a random variable, whose empirical cumulative distribution function is utilized to determine a stopping rule that ensures consensus is achieved with a prescribed confidence level. Simulation results show that the consensus-reaching time obtained with a prescribed confidence level compares relatively well to local-averaging-based algorithms. The algorithm is shown to be robust, to some extent, to lossy communications and is demonstrated on a weapon-target assignment problem.