Neural-based Self-tuning Controller for AQM Router Supporting TCP with ECN

As an enhancement mechanism for the end-to-end congestion control, active queue management (AQM) can keep smaller queuing delay and higher throughput by purposefully dropping the packets at the intermediate routers. Comparing with RED algorithm, although the PI (proportional-integral) or PID (proportional-integral-differential) controller for AQM improves the stability, It is very difficulty in selecting a group of the parameters of the controller to guarantee the transient performance. Moreover, the case is even worse especially in rapidly changing and complex conditions, such as stochastically accessed users, unpredicted nonresponsive traffic and hard limitations on the network resources. In this paper, we propose a self-tuning controller for AQM router supporting TCP with ECN based on neural compensator. The effectiveness of integrated performance of the controller is demonstrated by the simulation results.