Abstract
We consider optimum feedback control strategy for computer
communication network, in particular, the access control
mechanism. The dynamic model
representing the source and the access control system is
described by a system of stochastic differential equations
developed in our previous works. Simulated annealing (SA) was used
to optimize the parameters of the control law based on neural
network. This technique was found to be computationally
intensive. In this paper, we have proposed to use a more powerful
algorithm known as recursive random search (RRS). By using this
technique, we have been able to reduce the computation time by a
factor of five without compromising the optimality. This is very
important for optimization of high-dimensional systems serving a
large number of aggregate users. The results show that the
proposed control law can improve the network performance by
improving throughput, reducing multiplexor and TB losses, and
relaxing, not avoiding, congestion.