Bayesian filtering methods for target tracking in mixed indoor/outdoor environments Katrin Achutegui 1 , Javier Rodas 2 , Carlos J. Escudero 2 and Joaqu´ ın M´ ıguez 1 Department of Signal Theory and Communications, Universidad Carlos III de Madrid, Spain, {kachutegui,jmiguez}@tsc.uc3m.es, Department of Electronics and Systems, Universidade da Coru˜ na, Spain, {jrodas,escudero}@udc.es Abstract. We propose a stochastic filtering algorithm capable of inte- grating radio signal strength (RSS) data coming from a wireless sensor network (WSN) and location data coming from the global positioning system (GPS) in order to provide seamless tracking of a target that moves over mixed indoor and outdoor scenarios. We adopt the sequen- tial Monte Carlo (SMC) methodology (also known as particle filtering) as a general framework, but also exploit the conventional Kalman filter in order to reduce the variance of the Monte Carlo estimates and to design an efficient importance sampling scheme when GPS data are available. The superior performance of the proposed technique, when compared to outdoor GPS-only trackers, is demonstrated using experimental data. Synthetic observations are also generated in order to study, by way of simulations, the performance in mixed indoor/outdoor environments. Key words: Bayesian filtering; indoor/outdoor tracking; Kalman filter; particle filter; switching models 1 Introduction The existing outdoor and indoor systems for target positioning and/or tracking have evolved in rather different ways. The global positioning system (GPS) is the most common technology in outdoor scenarios. It provides broad coverage, essentially ubiquitous except for a few “tough” environments, such as urban canyons [11], yet it has a poor accuracy, in the order of 10 meters [8, 17]. Po- sitioning based on cellular networks yields a similar precision and the coverage, even if not global [16], can include urban areas where GPS fails. Combinations of both technologies [16] are attractive but do not resolve the accuracy prob- lem. During recent years, localization systems based on wireless sensor networks (WSNs) have gained momentum, specially for indoor applications [16, 13]. In outdoors environments, WSNs providing radio signal strength (RSS), time of arrival (ToA) or angle of arrival (AoA) data can potentially beat GPS and cel- lular networks in terms of accuracy, but they are ad hoc systems to be deployed only in small areas [16].