NEUROCONTROLLER FOR BUFFER OVERLOAD CONTROL IN A PACKET SWITCH

The authors describe a buffer overload control scheme, which adaptivel y matches the buffer input rate to prevailing conditions in the queuei ng system, for efficient utilisation of the system resources. They pro pose an adaptive feedback control algorithm which uses neural network modelling of the system dynamics and a gradient-based control optimisa tion technique for buffer overload control. Using a simple dynamic que ueing model, it is shown that this technique can regulate buffer occup ancy levels (or queueing delays) to minimise congestion in the queuein g system.