Adaptive control of service in queueing systems (Q790703)

The authors consider the combined problem of parameter estimation and of optimal control of the service rate in an M/G/1 queueing system with unknown (but constant) arrival rate \(\lambda\) and with general cost structure. An optimal adaptive policy is determined using recently developed results on parameter estimation and adaptive control of semi- Markov processes. The authors show that under some general conditions the optimal adaptive control scheme is as following: (i) determine an optimal stationary policy (OSP) for the average cost problem for each value of the parameter \(\lambda\), (ii) at each decision point \(T_ n\) (assumed to be given), get a strong consistent estimate \(\lambda_ n\) of the true parameter value, (iii) define the adaptive policy as the value of OSP, obtained in step (i), in point \(\lambda =\lambda_ n\) at each decision moment \(T_ n\).