Tasks in MN Site

As shown in Figure 8, the first task on each positioning is that MN scans wireless channels and measures the Signal Strength of each detectable AP. After collecting the AP Signal Strength Set, positioning process can be started.

3.3.1 Predicting by Potential Locations

This is the first stage to predict MN’s location. Considering the moving behavior, most of them may follow the same tracking. That is, if someone traveled to location A and went to location B, the next one who comes to A may follow this track and go to B. As discussed in Location Table data, each location records its 5 next locations.

These data will be updated in recent positioning (Machine Learning). These 5 locations and their related information (AP Signal Strength Set and Similarity Thresholds) will be sent to MN at each succeed position. So, when an MN starts to perform positioning, it checks whether any potential location has been sent by last positioning first. If so, it compares MN’s current measured AP Signal Strength Set with these locations’ AP Signal Strength Set and selects the one with maximum similarity. If the maximum similarity is larger than the selected location’s Similarity Threshold, we can consider this selected location as its current location. MN sends an informing message to Location Server to update the Location Tables by its measured AP Signal Strength Set. Location Server updates this location information on Location Tables and re-calculates the Similarity Threshold of this location and its neighbors with 2 meter distance. Then AP replies an ACK with next potential location list to MN for later usage. The Location Server needn’t to position or confirm MN’s location in this stage.

If this is not the previous position or there is no any potential location sent by Location Server, or predicting by potential locations was failed, it will go to next stage: Predicting by Uniform Linear Moving.

3.3.2 Prediction based on Uniform Linear Moving

behavior. If this behavior is detected, it can be used for location prediction. As we discussed in Section 3.1, MN keeps four history locations, these locations can make 3 vectors. In our design, if these 3 moving vectors are similar, we can use it to predict MN’s current location. Please refer to Figure 9.

Figure 9: Prediction based on Moving Vector

When Uniform Linear Moving happened and was detected, MN sends a Location Query with measured AP Signal Strength Set and the predicted location to Location Server. Location Server confirms this predicted location by the location information (AP Signal Strength Set and Similarity Threshold) recorded in Location Table. If it succeeded, that is, the similarity of AP Signal Strength Set between “MN measured”

and “location table recorded” is larger than the recorded Similarity Threshold, the Location Server updates this predicted location’s information in Location Tables and re-calculated the Similarity Threshold of this location and its 2-meter neighbors. Then the Location Server replies an ACK with next potential location list to MN for next positioning.

If the confirmation was failed, Location Server performs Evaluating by AP Signal Set of MN to get MN’s current location, and updates location table then replies an ACK with next potential location list to MN for next positioning, too.

If the condition in this stage does not exist, it will go to next stage: Predicting by Signal Variation Ratio.

3.3.3 Predict by Signal Variation Ratio

Before an MN sends a query to Location Server to perform positioning on Location Server, it can predict its own location by its signal variation from previous location. That’s why we keep previous location information (Location and AP Signal Strength Set) in MN. That is, if an MN is not doing the first time positioning in this area, it can predict its location based on signal variation ratio first. This predicted location also can be added into position query to be sent to Location Server for confirmation. If confirmation is successful, complement positioning process is not needed on Location Server. This can reduce large amount of loading for Location Server especially when number of entries in the Location Table is large.

3.3.4 Location Query without MN Predicting

If all of above stages cannot be satisfied, the MN will send location query to Location Server with AP Signal Strength Set only. Location Server will evaluate the MN current location based on AP Signal Strength Set, and response to MN.

3.3.5 Algorithm on MN Site

Pot_Lo: potential locations;

Pot_Lo_SminTh: Similarity Threshold of each Pot_Lo;

Pot_Lo_APSS: AP Signal Set of each Pot_Lo;

TRK_Lo: 4 MN last locations, PAPSS: MN previous AP Signal Set;

Prd_Lo: Predicted Location;

Prd_Inf: MN predicted informing message. It asks Location Server to update Location Tables;

Lo_Query: MN Location Query message. Query for MN current location Simi(APSS1,APSS2): Similarity between APSS1 (AP Signal Strength Set)

and APSS2

Input: CAPSS: MN’s current AP Signal Set;

Output: MN’s location;

Begin:

Scan all Wireless Channels to get AP Signal Set(CAPSS);

if (Pot_Lo exists && Max[Simi(CAPSS, Pot_Lo_APSS)] ≧ Pot_Lo_SminTh) {

send Prd_Inf(Pot_Lo(with Max Similarity),CAPSS);

return (Pot_Lo(with Max Similarity));

}

else if (there are 4 TRK_Lo && all moving vectors are similar) {

Prd_Lo(variation between PAPSS and CAPSS);

send Lo_Query (Prd_Lo & CAPSS);

receive ACK or response from Location Server (positioned location, Pot_Lo) {

renew Pot_Lo for next positioning;

push positioned location into TRK_Lo queue;

save CAPSS to PAPSS;

}

return (positioned location);

End

Figure 10: Algorithm on MN