Robust Control Of Single Conveyor-Serviced Production Station

Conveyor-Serviced Production Station (CSPS) is a typical automated manufacturing system, in which parts are delivered to the production station with a finite-capacity buffer to be further processed.Interference of the production environment and abrasion of processing equipment will cause variation in the service rate.Therefore, studying the optimal control problem of such a system under the case of uncertain parameters has theoretical and practical significances.This dissertation considers the case where the service rate of each state has same bounds. The main objective is to find an optimal robust policy which optimizes the system weight performance under worst cases of the service rate, by using vacancies of the buffer as system states and look-ahead range as control variable. The optimal robust control problem could be described as a mini-max problem by employing semi-Markov decision process (SMDP) with uncertain parameters. Firstly, the worst performance value was obtained under fixed policy by genetic algorithm. Secondly, based on the worst performance value, the optimal robust control policy was achieved with simulated annealing algorithm. Simulation results show that there is little difference between optimal performance cost of the system whose service rate is fixed as the mean of interval and optimal robust performance cost of the CSPS system with uncertain service rate. Moreover, the difference between the optimal robust performance cost and the optimal performance cost is getting smaller when the uncertain interval narrows which means that the global optimization algorithm works effectively.In addition, we consider case where states of the system correspond to different ranges of service rate. The main objective, still, is to find an optimal robust policy which optimizes the system performance under worst cases of the service rate.Then the optimal robust control problem could be described as a mini-max problem by using semi-Markov decision process (SMDP) with uncertain parameters. Policy iteration and Q-learning algorithm are combined to obtain the optimal robust policy.Simulation results show thatrobust optimal control of the system could achieve better optimization results, indicating the effectiveness of our proposed algorithm.