Intelligent objects in simulation

In this paper, learning automata are incorporated as decision making agents (‘intelligent objects’) to handle the discrete type of decisions in simulation models. The integrated simulation system is viewed as a continuously evolving, closed system in which operational and design decisions can be made automatically. In this learning paradigm, the simulation model serves as an environment which provides performance evaluations for the decision policies formulated in succession by the learning automata. As the automata revise the action probabilities at regular intervals on the basis of these performance evaluations, they improve their decision making process continuously. The approach can be an excellent tool for the analysis and design of complex systems under uncertainty. An application is demonstrated in a manufacturing setting in which due-dates and machine assignments of jobs are the decisions to be made by the automata so as to minimize a nonlinear function of job tardiness and earliness subject to random job arrivals and machine breakdowns.