On Measures of Cooperation in Distributed Systems Under Uncertainties

We study the evolution of distributed multiagent systems under uncertainty where the autonomous agents may cooperate with each other, and/or with supervisor/operator, in order to achieve the system's objective. The cooperation is facilitated by means of information sharing among the autonomous agents and/or supervisor/operator, which has the purpose of improving the effectiveness of the autonomous agents. The evolution of cooperative systems is modeled using discrete-state, continuous-time Markov processes. To measure and quantify the degree of cooperation within such systems, we introduce the concept of coefficient of cooperation, which is obtained by minimizing the Kullback-Leibler or 1-norm distances between nonstationary probability distributions. The presented techniques are illustrated on several different types of multiagent search systems.