Ðóóóððý Çôøøññð Ìööò×ññøøøö Èðð Blockin Blockinññòø Óö Áòòóóö Ïïööðð×× Óññùòò Blockin Blockinøøóò Ëý×øøñ×

Indoor Wireless Communi ation Systems J. He, A. Verstak, L. T. Watson, Fellow, IEEE, C. A. Stinson, N. Ramakrishnan, C. A. Sha er, T. S. Rappaport, Fellow, IEEE, C. R. Anderson, Student Member, IEEE, K. Bae, Student Member, IEEE, J. Jiang, Student Member, IEEE, and W. H. Tranter, Fellow, IEEE Abstra t|In this paper, a global optimization te hnique is applied to solve the optimal transmitter pla ement problem for indoor wireless systems. An eÆ ient pattern sear h algorithm|DIRECT (DIviding RECTangles) of Jones, Perttunen, and Stu kman (1993)| has been onne ted to a parallel 3D radio propagation ray tra ing modeler running on a 200-node Beowulf luster of Linux workstations. Surrogate fun tions for a parallel WCDMA (wideband ode division multiple a ess) simulator were used to estimate the system performan e for the global optimization algorithm. Power overage and BER (bit error rate) are onsidered as two di erent riteria for optimizing lo ations of a spe i ed number of transmitters a ross the feasible region of the design spa e. This paper brie y des ribes the underlying radio propagation and WCDMA simulations and fo uses on the design issues of the optimization loop. Index Terms|bit error rate, DIRECT algorithm, global optimization, power overage, ray tra ing, surrogate fun tion, transmitter pla ement, wideband ode division multiple a ess. I. Introdu tion Optimal transmitter pla ement provides high spe tral eÆ ien y and system apa ity while redu ing network osts, whi h are the key riteria for wireless network planning. As the omplexity and popularity of modern wireless networks in reases, automati transmitter pla ement provides ost savings when ompared to the traditional human pro ess of site planning. Automati design tools are being developed to o er eÆ ient and optimal planning solutions. The system des ribed here, S4W (Site-Spe i System Simulator for Wireless system design), is among the few known wireless system tools for in-building network design, besides [2℄, [6℄, [8℄, and [12℄. An optimization loop in S4W is proposed to maximize the eÆ ien y of simulated hannel models and surrogate fun tions are proposed to redu e the ost of simulations. Transmitter pla ement optimization is one spe i problem that an be solved by S4W . The key ontributions of the present work are surrogate modeling, appli ation of the DIRECT algorithm, and the simulation results.