Robust Algorithms For Wireless Networks And Intelligent Congestion Control

In recent years, the speed of the development of network technology is growing, hybrid wired-and-wireless access network have been used widely, the network congestion is the key factor affecting the network performance. The main issue of wireless network is how to avoid congestion problem. Due to the characteristics of wireless network, which make the existing traditional wire network congestion control methods can not be applied to wireless network directly. Therefore, it is significant to design congestion control methods which can be applied to the wireless access network. Aiming at the congestion problem in wire-and-wireless network, we first analyze the characteristics and shortcomings of the existing AQM algorithm, and combined with the wireless access network TCP/AQM model, design new wireless AQM algorithm with robust control and neural network theory. At last, verify the effectiveness of these algorithms by simulation. The main work of this paper includes:(1) Take full account of network parameters changes and problems caused by uncertainty, proposed AQM (SFC, State Feedback Control) with parameter uncertainties and time delay model, and prove the stability of the algorithm by Lyapunov. Using LMI to calculate the control law parameters. As it’s not easy to get the value of the network window, we use state estimation method to avoid the tedious calculations. Finally, using NS2to simulate and contrast with the traditional AQM.(2) Using state-space approach to describe the problem as a linear system containing state and input delay. Based on the network parameters changes, using the state-observer-based AQM algorithm(OBSF, Observer Based State Feedback) to reduce the errors caused by approximation. Simulating and verifying the performance of the algorithm by NS2.(3) As there is no exact TCP/AQM flow model for wireless access network, proposed PID neural AQM based on BP learning rule (NNPID, Neural Network PID),and applied it on single bottle neck link in wireless access network. Using gradient on-line adjustment to get the weights. Meanwhile, proving the stability of the algorithm Simulation result shows that the proposed algorithm has obvious advantages in rapid response capability and volatility.