Risk management system based on uncertainty estimation by multi-agent

Abstract

The purpose of this paper is to create risk management system that multi-agent patrols for maintaining security. We consider that maintaining security is equal to relieving uncertainty. Therefore, we formulate uncertainty of places needed to maintain security by the entropy in an information theory. We call these places ”check point i(i=1,2,…,n)”. Agents patrol and observe check point's condition value with updating patrol schedule on the basis of estimating uncertainty of check points in real time. We propose the method ”Earliest Deadline First Scheduling with Adaptive Risk Estimation(EDFRE)” to relieve uncertainty. Then we compare the proposaled method with simpleEDF(Earliest Deadline First) scheduling. The result indicated that our proposal method was effective for dynamic situation.