Adaptive neuro-fuzzy controller trained by genetic-particle swarm for active queue management in internet congestion

Abstract

Routers are vital during network congestion. All routers have input and output packet buffers. Various congestion control strategies have been suggested. Some controller-based proportional-integral derivative (PIDs) have recently been offered as active queue management (AQM) solutions to alleviate the deterioration of transmission control protocol (TCP) congestion management system performance. However, the time delay is large, the data retention decreases, and oscillation occurs, suggesting that the present PID-controller is unable to fulfill quality of service (QoS) criteria. Some research is developed on new control technologies such as neural networks and fuzzy logic. This paper proposes the adaptive neuro-fuzzy inference system (ANFIS) like PID controller for AQM. This model employs genetic algorithms (GAs) and particle swarm optimization (PSO) to learn and optimize all variables for ANFIS like PID controller. Simulations were used to investigate the effects of using fuzzy like PID based on single sign-on (SSO), and (ANFIS like PI, ANFIS like PID with GA-PSO) controllers on the length of the queue for an AQM router, respectively. Then we compared the findings to see which approach should be utilized to manage the queue length for AQM routers. In simulations, ANFIS like PID has superior stability, convergence, resilience, loss ratio, goodput, lowest rising time, overshoot, and settling time.