[MyResearch] Wireless positionning: 2 solutions
According to Johnny Shih (thesis, pdf file), it seems that there are two solutions for localizing a device thanks to a 802.11b network, both use the signal strength:
- empirical model: based on storing pre-recorded measurements in a database (a series of selected points for which the locations (coordinates) of these points are known, signal strength value from the APs are also collected). When a device with unknown position requests positioning, the signal information from all the APs are collected and sent to the database for comparison. An empirical model program created then compares and finds the closest match in the entries of the known locations in the database against the entry of the known point. The comparison uses the K-nearest neighbor algorithm. disadvantages to the systems built using the empirical model. The systems will always require a considerable amount of manual efforts on radio map and database constructions, when they are to be used in a new environment. The other disadvantage the system can lose some accuracy when the current environment condition is different the condition when the radio map was constructed. - propagation model: the distance from a wireless device to an access point can then be calculated given the received signal strength loss value. By having the distances to three and more access points, triangulation method can be applied to determine the location of the device. Difficulties in the actual implementations. In order to achieve accurate results, the signal pathloss values that the positioning system received from the access points have to be precise. many existing systems have shown that the accuracy of positioning will decrease as the distance between the positioning device and the affiliated access point increases. To eliminate the aforementioned problems and increase the positioning accuracy, the system can be designed to give the primary access point (which has the least pathloss value) a higher weight in the triangulation algorithm.