Research On TCP Network Congestion Control Based On Hamiltonian Theory

With the development of science and technology, the emergence of new network applications and an increasing number of Internet users in the network, the congestion become a problem to be solved. The combination of active queue management mechanism based on the router and TCP congestion control is a major way to resolve the problem of TCP network congestion. Considering the strong nonlinear TCP network, this thesis uses the theory of Hamiltonian to design the controller. Hamiltonian system is a kind of control method based on energy, not only the structure is simple, in the controller design and stability analysis also has a huge advantage. Therefore, in view of the TCP network system, this thesis try to use the Hamiltonian theory for the first time in control method to design controller, successfully solved the problem of the network congestion. In this thesis the main research contents include:First of all, TCP network congestion control research’s background and significance were introduced. The network congestion’s reasons were elaborated. The current researches of congestion control algorithm were summarized. Aiming at the existing problem of network congestion, the existing difficulties of network congestion control algorithm design was analyzed and the researches of Hamiltonian theory were illustrated.Secondly, expounds the basic theory of the Hamiltonian control method, structural characteristics and the structure of the Hamiltonian model. At the same time, considering the parameter time-varying, strong nonlinear dynamic network system, in order to adapt changes in the network, this chapter uses Hamiltonian system which is a kind of control method based on the energy function to handle the characteristics of nonlinear. Preset feedback method is used to establish the network congestion control model.Thirdly, the Hamiltonian controller was designed and its stability in theory was analyzed. Then, the simulation was analyzed and the traditional PI controller simulation diagrams were compared. The results show that Hamilton’s controller has the small overshoot, fast regulating time. Its speediness and stability is better than PI controller, which can effectively avoid network congestion.Fourthly, the interferences such as the UDP flows were disposed and the performance indicators related inspection function was constructed with the Minimax control method. Estimating the influence of interferences on the system, the conservatism of the traditional interference treatment was reduced effectively. For parameter uncertainties in TCP network system, the adaptive control was joined and the adaptive law was designed. The simulation analysis shows that the proposed control strategy is effective to improve the transient stability and robustness of the system.At last, a summary and outlook were made.the Contents of this research were summarized and the prospects in the Hamiltonian theory of congestion control were analyzed.