Optimal setup of interference threshold in a multi-cell WCDMA environment
Introduction
The cell capacity of a CDMA system is mainly determined by the sum of two types of interferences, the one generated within the cell and the other from neighbouring cells. The former one, called the own-cell interference (OWNI), is controlled by the cell itself while the latter one, called the other-cell interference (OTHERI), is usually assumed given in the literature. The size of OWNI can be controlled by call admission policies (CAPs), among which the most popular one is threshold-type. Taking the threshold-type CAP, we are interested in finding the policy's threshold value (TV), i.e. the upper bound on the volume of OWNI generated by active calls at a cell, beyond which further call admission at the cell is rejected. But for that, the OTHERI at the cell should be correctly figured out, the size of which is in turn determined by the OWNIs of neighbouring cells. Due to traffic variation and geographic diversification, the sizes of OTHERIs vary significantly among cells in practice. However in all existing studies concerning the CDMA capacity analysis, the size of OTHERI is assumed given fixed over all cells by a uniform exogenous factor (see Ref. [1] pp. 164). The assumption is indeed too simplified to reflect the aforementioned reality. This motivates us to focus on obtaining the optimal OWNI TVs concurrently for all constituent cells.
The traffic volume of multimedia Internet services is characterized by the downlink dominance. With attention paid to the downlink traffic only, we aim at maximizing the expected system throughput while satisfying the interference constraint at each cell.
Section snippets
Admissible region and expected throughput
An admissible region of a cell is defined as the set of cell states within the pre-specified OWNI TV. A cell state is represented by vector where nk denotes the number of active class-k calls. Given the OWNI TV of cell l as il, the admissible region of the cell is represented byHere referred to as the downlink load factor of class-k, denotes the downlink load generated by a single class-k connection, which is obtained by
Performance maximization model (PMP)
Given M cells in the service area, the Performance maximization model (PMP) is formally stated as follows:
(PMP)s.t.The objective function (4) is the expected system throughput taking the form of the sum of expected cell-throughputs. The OWNI TV of cell l(il) is divided into a series of ij,ls (j=1,…,J) according to the increasing rate of expected cell-throughput. The variable ij,l refers to the amount
Numerical example
For a comprehensive test, we consider a 17-cell WCDMA system with R=1 as in Fig. 3.Each cell of the entire system has one of three demand-levels. The arrival rate of each demand-level and other input data, taken from the studies [4], [5], [6] are listed in Table 1. In Table 2, we show the OWNI TV and demand-level for each cell along with the associated objective value when 2 Note that, OWNI level is divided into 10
Conclusion
The solution of the formulated linear program shows how the system throughput can be maximized. As expected, the TV of a boundary cell is greater than the one in dense area. The contribution of this study is in quantifying the average gain in system throughput when allowing differences in OWNI TVs among cells.
References (6)
- et al.
WCDMA for UMTS
(2000) - et al.
The Stochastic Knapsack Problem
IEEE Trans. On Commun.
(1989) - et al.(1990)