Congestion Control Algorithm for the Internet Based on Adaptive Neural Network Terminal Sliding Mode

To constrain congestion phenomenon in the internet, we propose a congestion control algorithm based on terminal sliding mode control theory and radial basis function (RBF) neural network. Taking the variation of network parameters and the effect of non-transmission control protocol (TCP) data flows as the uncertainty existing in the system, we design a novel terminal sliding surface to shorten the finite convergence time for the sliding mode. Moreover, we prove the asymptotical stability of the sliding mode. Using the RBF neural network, we estimate the upper bound of the uncertainty in the controller, and then obtain an adaptive law of neural network weights based on Lyapunov stability theory. Simulation experiments on the same network conditions show that our algorithm has better robustness and faster convergence rate than the PID controller and the traditional terminal sliding mode controller.