Adaptive rate‐based congestion control with weighted fairness through multi‐loop gradient projection internal model controller

Abstract

As a transmission control protocol (TCP) implementation imposes the sources with small round trip time (RTT) to allocate the bottleneck unfairly, an adaptive congestion control is necessary to avoid packet loss along with fairness. The main idea of this paper is to design an adaptive rate-based queue management scheme based on gradient projection method and internal model control for network sources with different RTT values. The goal is to achieve weighted fairness, maximum utilisation and compatibility with both large and small RTTs. The communication network consisting of a bottleneck link and N TCP sources is considered as a multi-input single-output system. In the proposed approach, the number of the active TCP sources is determined adaptively and the utilisation factors are updated. For this purpose, the gradient method is used to design the adaptation rule for the utilisation factors in order to obtain weighted fairness, and then the projection method is augmented with the gradient procedure to achieve maximum bottleneck utilisation. The proposed procedure can tolerate both large and small RTTs, and consequently, it can be used in a wide range of communication networks. Extensive simulations based on network simulator NS2 and Simulink validate the analytical results.