Multiple-Access Interference Processes Are Self-Similar in Multimedia CDMA Cellular Networks
J. Zhang, Takis Konstantopoulos · IEEE Transactions on Information Theory · 2005
We consider bursty data communications in code-division multiple-access (CDMA) cellular networks. The significant fluctuation of the cochannel multiple-access interference (MAI) in such systems makes it very challenging to carry out radio resource management. A main goal of this paper is to obtain a fundamental understanding of the temporal correlation structure of the MAI, which plays a crucial role in effective resource allocation. To this end, we take a cross-layer design approach, and characterize the stochastic MAI process while taking into account both the burstiness of data traffic and time-varying channel conditions. Our main results reveal that under standard assumptions on ON/OFF traffic flows and fading channels, the MAI process exhibits scale-invariant burstiness and is "self-similar" (with Hurst parameter 1/2<H<1), in both the uplink and the downlink cases. The MAI self-similarity indicates that the MAI levels are long-range dependent and therefore there exists a nontrivial predictive MAI structure across multiple time scales. The predictive MAI structure can be utilized for effective interference management through dynamic resource allocation. We illustrate this via a rate control scheme based on the MAI prediction, and our results show that the performance gain is substantial. The exploitation of the MAI temporal correlation structure for resource allocation parallels and complements multiuser detection which utilizes the MAI snapshot structure at the symbol level.