Abstract
Since congestion is the primary cause of computer network performance deterioration, controlling congestion has emerged as one of the core problems in computer networks. The method to stop the network's performance from deteriorating as a result of variations in its traffic load is called congestion control. Ineffective resource use might result in network failure in the absence of adequate congestion management techniques. Therefore, congestion management is an attempt to adjust a network's performance to variations in traffic load without negatively impacting users' perceived utility.
In order to enhance the efficacy of congestion control on TCP/IP networks by predicting the uncertainties and disturbances in the network, this study presents a unique technique for internet congestion management employing an Extended State Observer (ESO) in conjunction with proportional-derivative (PD) control.
The limitations of various classical observers, such as the Disturbance Observer (DO), are also covered in this study, along with how ESO overcomes them by applying the concept to real-world non-linear systems. The simulation demonstrates that the extended state observer performs much better when handling dynamic uncertainty and network parameter fluctuation.