A Deployment Optimization Model for Wimax Base-Stations

Abstract

Telecommunication technologies are growing rapidly due to the highly demanding markets for different telecom services and applications. Wireless solutions are essential in places where cable infrastructure does not exist, like in remote and developing areas. To satisfy the demands of business and residential customers for high quality voice, video and data services, it is required to deploy a reliable broadband wireless network taking into consideration the possible options to maximize the overall profits and customer satisfaction. This work aims at developing an optimization deployment model for base-stations of fixed wireless access technologies, in particular the Worldwide Interoperability for Microwave Access (WiMAX). The base-stations deployment problem is formulated as a mixed integer linear programming (MILP). The model generates the optimum deployment configurations of the base-stations including the selection of best locations to install the base-stations among a number of candidate sites, number of base-stations to be installed in the selected sites and the optimum transmission power per site. Both technical and economic feasibilities are considered by maximizing the received signal strength, data traffic utilization, number of served nodes and the net annual profit. A sensitivity analysis is conducted to assess the impact of different parameters on the network performance. The results show that the most critical parameters are the weighting coefficients associated with number of served nodes and annual profit in addition to the maximum net throughput and base-station cost. A bin packing (BP) algorithm is proposed as a heuristic procedure to solve large size problems within a reasonable computational time. A MATLAB program is developed and the algorithm is applied to the same scenarios solved by the proposed MILP model. The results of the BP model are compared to the optimum solutions in order to benchmark the algorithm performance. The performance of the algorithm is reasonable since the deviation in objective value varies between 5% and 20% in most of the cases.