Performance Analysis of Centralized and Partially Decentralized Co-Operative Networks

• Published in: IEEE transactions on communications Vol. 64; no. 2; pp. 863 - 875
• Main Authors: Senanayake, Rajitha, Yeoh, Phee Lep, Evans, Jamie S.
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.02.2016; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Approximation methods; Base station cooperation; Base stations; Clusters; Computer architecture; Decentralized; Estimation; Fading; Interference; LMMSE; Mathematical models; Microprocessors; Multiuser detection; Networks; Probability density functions; Product development; Rayleigh fading; Receiving antennas; Signal processing; Signal to noise ratio; Rayleigh fading; multiuser detection; LMMSE; Base station co-operation
• ISSN: 0090-6778; 1558-0857
• DOI: 10.1109/TCOMM.2015.2512918

We consider cellular networks with co-operative clusters of neighboring base stations detecting multiple in-cluster users subject to interference from out-of-cluster users. We assume that the base stations, equipped with multiple antennas, are connected to a central processor in each cluster. For such a network, we first consider centralized processing where all the in-cluster user signals are sent to the central processor for linear minimum mean squared error (LMMSE) estimation. Next, we consider partially decentralized processing where the in-cluster user signals are locally estimated at each base station, and the local estimates are combined at the central processor. For both processing architectures, we derive new expressions for the achievable rate of an in-cluster user when the channels between the users and base stations are subject to independent Rayleigh fading and distance-dependent path loss. The solutions are based on accurate approximations we derive for the characteristic function (CF) and the probability density function (PDF) of each user's signal-to-interference-plus-noise ratios (SINRs). Numerical examples highlight the accuracy of the analysis and compare the performance of centralized and partially decentralized processing under different cluster scenarios.