UMTS Radio Network Planning: Mastering Cell Coupling for Capacity Optimization (eBook)

Dr. Hans-Florian Geerdes is a scientist at Zuse Institute Berlin and at the DFG research center MATHEON: Mathematics for Key Technologies. His research focuses on applications of combinatorial optimization in wireless telecommunications.

Abstract 6
Acknowledgments 7
Contents 8
1 Introduction 10
2 Radio network modeling and performance evaluation for UMTS 15
2.1 Cellular wireless communication networks 16
2.2 The UMTS radio interface 19
2.3 Methodology of performance evaluation 26
2.4 The classical static model 27
2.5 Performance evaluation with static simulation 35
3 Interference-coupl ing complementarity systems 42
3.1 linear interference-coupling equation systems 44
3.2 Perfect load control and complementarity systems 48
3.3 Generalized pole equations 59
3.4 Performance indicators 63
4 Expected-i nterference-coupl ing estimates for network performance 69
4.1 The reference method: simplified Monte Carlo simulation 70
4.2 Expected interference coupling with medians of attenuation 73
4.3 Refined estimates for the expected grade of service 79
4.4 Computational experiments 83
4.5 Conclusions on system modeling and performance evaluation 94
5 Network performance optimization 98
5.1 Prerequisites: objectives, parameters, and optimization methods 99
5.2 Survey of network planning literature 109
5.3 Optimization models 116
5.4 Computational case studies 125
5.5 Analysis of case study results 144
5.6 Conclusions on performance optimization 152
6 Conclusion 155
Appendices 157

4 Expected-i nterference-coupling estimates for network performance (S. 63-64)

Radio network planning aims at improving the expected network performance, so we are not interested in network performance on a single snapshot, but on the expected performance for random snapshots. The coupling matrices thus have to be considered random variables subject to a probability distribution induced by the distribution on snapshots, and we are interested in the stochastics of the performance indicators. Simulation methods are commonly used for determining mean values of performance indicators, but they are inherently too time consuming for use in heavy-duty optimization tools, therefore, faster approaches are needed.

While Monte Carlo methods can yield an arbitrary accuracy if sufficient time is granted, high (absolute) precision is dispensable for taking intermediate planning decisions. For a successful optimization campaign, the ability to quickly discriminate between design alternatives is paramount. The right decision can be made in short time, if accuracy is sacrificed in a controlled fashion. The practical relevance of quick estimation techniques is apparent from the fact that many commercial software tools advertise fast proprietary evaluation methods besides Monte Carlo simulation (Aircom International ltd, 2007, Cosiro GmbH, 2006, Ericsson AB, 2006, Lustig et al., 2004).

In this chapter, we develop methods for estimating the expected network performance with little computational effort. The basic idea is to calculate approximations to the mean values of capacity-related performance indicators based on the mean coupling matrix. The scheme depends on suitable choices of the performance model and of the random model. With the interference coupling complementarity systems, we have a detailed model that reflects the relations between cells. We restrict the random model on snapshots to exclude shadow fading, calculating with the medians of attenuation (the deterministic path loss component) instead.

The resulting method of expected interferencecoupling with medians of attenuation is tailored to the common representation of planning data in computer software. We complement it with a specialized method that calculates better estimates of the grade of service using second-order moments. Besides the method itself, this chapter contributes the thorough analysis of the expected coupling method and its validation as a suitable tool for network planning. Our investigations comprise analytical and empirical studies.

On the analytical side, we use the new generalized pole equations in a simplified setting, the results explain how the service mix determines the variance of the coupling matrix and thereby the quality of the estimates. In our computational studies, we essentially demonstrate that the method is sufficiently informative for typical applications in network planning. The remainder of this chapter is structured as follows: We introduce the accurate reference method of Monte Carlo simulation in Sec.4.1. We define the expected coupling estimates and analyze their accuracy in Sec. 4.2. Refined estimates for the grade of service are developed in Sec. 4.3. In Sec. 4.4, we conduct extensive computational experiments to analyze the accuracy of our new estimates and assess the validity of perfect load control in realistic settings. We draw conclusions on network modeling and performance evaluation in Sec. 4.5. Related work. Random quantities are often represented by their mean in a first-order approximation, in so far the expected coupling approach is a canonical choice.