[Back]

L. Xia, D. Wu, E. Mok, G. Retscher:
"Adaptive Indoor Hybrid Positioning for LBS";
Talk: 8th International Symposium on Location-Based Services LBS 2011, Vienna; 2011-11-21 - 2011-11-23; in: "8th International Symposium on Location-Based Services LBS 2011", (2011), 15 pages.

Following with varieties of network infrastructures, such as the third generation communication base stations, WiFi hotspots, sensors of IOT (Internet of Things), etc. have been deployed more and more widely, location based services (LBS) are becoming necessary and attractive. The provision of accurate, reliable, continuous and real-time positions is critical for LBS. GNSS, related to GPS, GLONASS, Galileo and Compass, can work under most conditions except indoor and blocked regions where are hot areas in our daily lives. So other technologies should be considered to enhance location awareness in these hot areas, and WLAN, cellular communication network, RFID, etc. can be the candidates, among which, WLAN is thought to be the most potential one. As some inherent characteristics, location fingerprinting method is always chosen for calculating the positions in most short-range radio signals-based location aware events. However, traditional fingerprinting methods need long time offline site survey to form matching database which must be updated continuously for changing environments. These shortages limit the usage of location fingerprinting method, as well as the availability and reliability of short-range radio signals based location information.
In our contributions, we proposed an adaptive localization technique that, base stations broadcasted the mapping relationship between their fixed coordinates and received heterogeneous AP signals (including WLAN APs, Zigbee nodes, etc.) from surroundings, and then the terminals would calculate positions based on received radio signals using their own mapping models which were constructed and adjusted with the base stations´ broadcasting information. The difficulty of mapping random high dimension signals vector to 2- or 3-dimension coordinates is one of the inherent limitations of this type location awareness. Fortunately, neural networks can provide the mapping and modeling ability. This paper chose BP (back propagation) neural network to learning and modeling the mapping relationships between radio signals and correlated coordinates. Moreover, the parameters of relationships were broadcasted using an ad hoc network--WSN which is one of the robust, easy-deploy, cheap, and low-power data links. A number of experiments were conducted in Geography department building at Sun Yat-Sen University. The results have shown that our method can produce relative high accuracy results, while require little support of additional positioning database and short training period. What´s more, heterogeneous sensor signals location perceptions perform more accurate, reliable, and robust, compared with WLAN signals standalone. As a result, more sensors, such as RFID, Infrared should be integrated to provide much more nice location service.

LBS, Adaptive, Indoor hybrid localization, BP Neural network, WLAN, WSN