Controlling QoS at the application level for multimedia applications using artificial neural networks: Experimental results

In this paper we implement a recurrent high order neural network based controller (built into a Client-Server architecture), in order to provide QoS control. The scheme is based on a recently proposed QoS control algorithm with proven stability and robustness properties. Mild assumptions have been used to construct a user satisfaction function that depends on two media characteristics (i.e. color depth and frame rate). The above, along with bandwidth measurements obtained at per frame basis, provide the essential information to the controller, which outputs the appropriate values for the media characteristics in order to achieve the required user satisfaction without violating the available bandwidth constraint. The implementation is performed on a TCP/IP network and the QoS control, which is applied at the application level, is shown to hold a real time property. Experimental results highlight its performance.