Skip to Main Content
Model predictive controllers, due to their capability to control the time variant and delayed systems, are considered for congestion control in computer networks. To extend the application of these controllers as an active queue management (AQM) system in dynamic TCP/IP networks, a new rule-based predictive controller is proposed. This controller uses the small signal linearized fluid-flow model of the TCP/IP networks and simulates the future behavior of control system by applying few candidate control sequences. Using the extremes of the resulted predicted output sequences, optimal control signal is determined based on some appropriate rules. This approach benefits from the mentioned capabilities of the model predictive controllers while the computational complexity is reduced compared to the commonly used MPC schemes. Low queue fluctuation, fast response and good disturbance rejection are the features of the proposed controller which are compared via simulations with those of PI and RED, the two well-known AQM methods.