GA-Based RED and PID Design for Minimization of Queue Length Variance in TCP/AQM Networks

Congestion control in TCP networks can be achieved by adjusting the transmission rate at the source or controlling incoming packets to the router at the link side. Active Queue Management (AQM) is a congestion control strategy implemented at the router to improve the Quality of service (QoS). The main objectives of AQM are to maximize link utilization and to minimize queuing delay by stabilizing the queue length at a desired target. This paper proposes Genetic Algorithm (GA) based Random Early Detection (RED) and PID controllers to improve the performance of congestion control in TCP networks. The parameters of the RED and the PID controllers are adjusted by minimizing an objective function using GA. The objective function is the variance of the error between output of the TCP network (queue length) and desired queue length. These GA-Based controllers minimize the fluctuations of queue length error which results in minimum jitter (the variance of delay). The performance of the proposed approach is examined by simulations.