Congestion feedback control for computer networks with bandwidth estimation

Effective congestion control allows networks to operate in region of low delay and high throughput. These characteristics seem to be the most required from network environment properties. To achieve high transfer rates, variety of methods and algorithms are proposed in the literature. This paper focuses on congestion control using a method combining a particle swarm optimization algorithm and the use of finite impulse response filer for non-stationary, discrete, dynamical model of communication channel. Proposed control strategy tunes the model to alleviate the results of sudden, unexpected network state changes. That helps to avoid undesirable congestion effects like packet dropping, retransmissions, high transfer latency and low network throughput.