Designing adaptive PI algorithm based on single neuron

Authors:
Li Qing;Qingxin Zhu;Mingwen Wang
Affiliations:
School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu, P. R. China;School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu, P. R. China;School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu, P. R. China
Venue:
ICCNMC'05 Proceedings of the Third international conference on Networking and Mobile Computing
Year:
2005

Citing 2
Cited 1

Random early detection gateways for congestion avoidance

IEEE/ACM Transactions on Networking (TON)
Fluid-based analysis of a network of AQM routers supporting TCP flows with an application to RED

Proceedings of the conference on Applications, Technologies, Architectures, and Protocols for Computer Communication

IAPI: An intelligent adaptive PI active queue management scheme

Computer Communications

Quantified Score

Hi-index	0.00

Visualization

Abstract

PI is a newly proposed Active Queue Management algorithm that has many important applications. But in case of heavy congestion its response is sluggish, and because of its static parameter's setting PI is sensitive to network status, such as RTT, capacity of neck-link, and number of TCP flows. To overcome these shortcomings of PI algorithm, we propose a novel AQM scheme, called Neuron based PI or NPI. NPI takes PI controller as an ADALINE with two inputs, and the proportional and integral factors of the controller are adjusted online by LMS algorithm. Simulation results show that under NPI the queue length converges to the desired value quickly and the oscillation is small.