Accurate WLAN indoor localization based on RSS, fluctuations modeling

The essential challenge in indoor WLAN localization is the severe fluctuations of the received signal strength (RSS) even for a stationary client. This paper provides a novel approach to develop a location system less sensitive to the time varying RSS. Two components are considered: multipath and noise. First, we provide a convolution model to characterize the multipath effects. Using our model, the multipath can be viewed as an additive random variable by transforming the temporal RSS trajectory into a log-spectrum domain. A subsequent log-spectrum average is utilized instead of a traditional time moving average to reduce the influence of multipath. Second, we perform the discrete Karhunen-Loeve (KL) transform to minimize the residual noisy components. This way, the impact of RSS fluctuationss on a location system can be efficiently mitigated. Our positioning system is developed in a real-world WLAN environment, where the realistic RSS measurements were collected. The proposed approach demonstrates significant improvements in the experiments. The numerical results show that the mean error is reduced by 18.38% on average, as compared to the existing methods.