Indoor coverage estimation from unreliable measurements using spatial statistics

Abstract

Estimating the coverage of a wireless network is one of the key problems in network planning and management. In outdoor environments this is usually done using modern network planning tools combined with intensive drive tests. However, in indoor environments the problem is much more difficult. Solutions based on propagation modeling require precise building information for accuracy, and even then their performance is highly varying. Refining such predictions using measurements from mobile terminals is a promising possibility, but is not straightforward due to the noisy and unreliable measurement quality. In this paper we study the performance of spatial statistics techniques for coverage prediction in indoor environments. Using data collected in an indoor testbed with 60 low cost radio receivers, we show that such techniques can yield accurate coverage predictions provided suitable preprocessing and filtering of the data is performed. Further, a simple optimization approach enables high prediction accuracy to be achieved using only a small subset of the available measurement devices. These results are also highly relevant to the minimization of drive tests (MDT) approach currently being developed in 3GPP to enable mobile terminals carry out coverage measurements for wireless networks.