An improved congestion avoidance control model for TCP Vegas based on Ad Hoc networks

The congestion control of TCP Vegas is more stable and brilliant than other TCP versions such as Westwood Plus because of the congestion control before collision in Ad Hoc networks with frequent topology changes. Nevertheless, Vegas has less competitive ability when its data flow coexists with other protocols´. In addition, it cannot make full advantage of available bandwidth to transmit packets. The research aims to propose an improved TCP Vegas, ICATCP, which has three enhanced aspects in congestion avoidance phases. I. The lower layers´ parameters will be considered in throughput model to improve the accuracy of theoretical throughput. II. Forward throughput prediction mechanism with revision based on Grey Prediction is used to promote the online cwnd control. III. The optimal exploring mechanism based on Q-Learning is applied to search the more reasonable changing size of congestion window. The simulation results show that the ICATCP has higher throughput, lower delay, more fair allocation of bandwidth than Vegas and Westwood Plus in multi-hops Ad Hoc scenarios.