Research Of The Key Technology Of Indoor Location Based On Wifi Used In Fire Rescue

In order to reduce the casualties while fire happens, researchers havebeen studying about how to increase the fire rescue efficiency. With thefast development of indoor location, it is possible for us to use indoorlocation technology in fire rescue. This article mainly applies the researchof indoor location based on WIFI in fire rescue.There are two main points in this research. One is the modelimprovement, parameter correction and some experiment related in thepositioning process based on propagation model. The other is theimprovement of positioning algorithm and related experiment in thepositioning process based on fingerprint matching.The improvement of the positioning process based on propagationmodel contains3aspects. First, we compare several kinds of distanceestimated models by analyzing the advantages and disadvantages andfinally pick the Keenan-Motley model, which is the most widely used onein indoor location, as the one we will use in the following research andexperiment. Then, we do some research in reducing the calculation error ofthe coordinate estimation algorithm and propose the improved trilaterationand the improved maximum likelihood estimation. The improvedalgorithms avoid the error being amplified too much during the calculatingprocess to some extent. Finally, considering the particularity of fire rescue,which may come in the result of access point loss, we suggest adding anew coefficient into the calculating process, which is called reliabilitycoefficient. Different access points come with different reliabilitycoefficients, which are related to the RSSI, distance between the receiving end and the AP and some other factors. While the number of access pointsis not enough for us to calculate with the normal method correctly, it isquite necessary to use the reliability coefficient as a reference to increasethe positioning accuracy.As to the positioning process based on fingerprint matching, theimprovement mainly contains2aspects. First, how to deal with the data.The numbers we get from sampling for200times inevitably contain somegross errors and some random errors. So it is necessary for us to deal withthese errors before we choose a value to restore into the database. Here weuse Pauta criterion to pick out the gross errors and Kalman filter to dealwith the random ones. By doing this, the values we use as the matchingparameters are more accurate. Another improvement is similar as the onewe use in model based positioning process. For each training point, wegive different reliability coefficients to different access point, then, werestore the coefficient together with the other information of the accesspoint such as RSSI, MAC and position information. When we want tomatching the testing data with the ones in the database, the value ofreliability coefficient should be taken into consideration.Experimental results show that, there is little difference between theoriginal algorithms and the improved ones while there are enough accesspoints. But once the number of access point reduced to a certain value, theadvantage of improved algorithm is obvious. So, taking reliabilitycoefficient into consideration has great meaning in the positioning researchwith only a few APs. However, while there are only1to2access pointsremaining, both of the two improved algorithm are defective. It’s true thatpositioning in this extreme situation is so difficult. In order to solve thisproblem, we give the suggestion of installing some fire emergency APs asfire emergency lights.