Fuzzy approximate fair dropping using genetic algorithm optimization

Active queue management (AQM) with fair bandwidth allocation is addressed in this paper. The fairness of the approximate fair dropping (AFD) (R. Pan et al., April 2003) scheme is shown by simulation to suffer under complex scenarios. A prediction-based fuzzy approximate fair dropping (FAFD) scheme is proposed to solve this problem. A least mean square (LMS) predictor is used to estimate the incoming traffic in the near future. Based on the current queue length and the predicted traffic intensity, the control parameters are adjusted by a fuzzy inference system , which is optimized by genetic algorithms (GA). Simulation results show that the prediction-based fuzzy adaptive control yields higher bandwidth utilization and reduces queuing delay jitter while achieving better fairness than the original AFD scheme.