A direct adaptive neural controller for flow control in computer networks

This paper presents a direct adaptive neural network control strategy for flow control in computer networks. The system to be controlled is modeled by a neural network and the control signals are directly obtained by minimizing a cost function which represents the difference between a reference and the output of the neural model. This model which can be cast in the framework of a general quality-of-service control problem, allows for the design of network access flow control mechanisms that can account for the nonlinear phenomena existing in computer networks. A number of simulation examples are given to illustrate the capability and flexibility of the flow control scheme. The results show that the flow control scheme is able to regulate the traffic loads to meet the system performance requirements.