6.8 Algorithm results – Detailed analysis and discussion
6.8.1 Scenario 1 – Example with accurate results achieved
In step 1, the algorithm compares the received signal strength values from all four sensors for a client against a threshold for each sensor as explained in section 4.5 to 4.7. One of the experiments conducted for location 8 (figure 5-14) will be discussed in the ensuing paragraphs. Table 6-4 provides the initial data as given by the sensors for a client. The algorithm checks these RSS values against the threshold for each room where a sensor is placed and decides weather the client is inside the room or beyond a wall (row 4 table 6-4). The threshold calculated for the rooms are given in row three of the table. From this data it is quiet clear that the client is beyond the wall with respect to each sensor. The same can be verified by looking at figure 6-19. The solid square dot represents the client and is not in any of the rooms which contain sensors. The next row of the table shows the translated distances of the signal strengths reported by the sensors, using equation 3-13. The reported distances are large figures. Sensor 2 and sensor 3 are reporting the client placed at 103 and 61 m distance respectively. Section 4.7 demonstrates that the maximum distance reported by any sensor for a client cannot exceed 29m for the example test area.
Table 6-4 Received signal strength values for a client and translated distance without applied correction
Sensor 1 Sensor 2 Sensor 3 Sensor 4
RSS (dBm) 60.90 83.15 76.07 63.30
Threshold
(dBm) 57.814 53.749 55.739 53.913
Beyond single
wall yes yes yes yes
Translated Distance
without
In step 2, the raw RSS values in row 1 of table 6-4 are checked for the client’s position beyond 2 walls. Again the RSS values are compared with calculated thresholds for two walls determined by the algorithm. Same are shown in table 6-5.
Table 6-5 RSS values checked for 2 walls and corrections applied by algorithm Sensor 1 Sensor 2 Sensor 3 Sensor 4
RSS (dBm) 60.90 83.15 76.07 63.30
Threshold two
walls (dBm) 69.73 68.07 70.39 67.69
Beyond two
walls No yes yes no
Corrected
Distance (m) 6.10 24 13.87 8.02
Table 6-5 provides the translated distances of the client from each sensor in the last row. The last rows of the two tables when compared tell us that the corrections are identified by the algorithm and applied to arrive at certain readings. This can be seen in appendix A (A.8.2) where algorithm numerical results are shown and the last ‘NOTT’ gives the status of client with respect to each sensor. Digit ‘1’ means the client is beyond single wall and digit ‘2’ means the client is beyond two walls. The same can now be triangulated to ascertain any overlapping areas. See figure 6-19.
In step 3, the raw RSS values that are identified to be beyond two walls are now tested against the absolute threshold value calculated for the test area. It is noted that circle 2 represents 22.18m of corrected distance which has been reduced from an exaggerated figure of 139.95m as given in table 6-4. The raw reported strength by sensor 2 was 83.15 dBm. Since it was detected to be beyond two walls, which is evident in figure 6- 17 when the client is observed with respect to sensor 2, 10 dB’s were subtracted. The resulting 73.15 dBm is now compared to the final absolute threshold of the sensor 2 calculated as 68 dBm (section 4.7). The algorithm finally assigned the value of 68 dBm to sensor 2 in respect of the client as based on the logic described in section 4.7. This 68 dBm is translated to ~24m as shown in the last row of table 6-5. Similarly all values of reported strength of a client that are identified as beyond two walls pass through this test of respective absolute threshold.
Table 6-5, in the last row, gives the corrected RSS values which when translated into circles gives the picture of overlapped and distinct circles as shown in figure 6-19. A solid square dot represents the client position. We can deduce by looking at the figure that circle 1, 2, 3 and 4 are reasonably close to passing through the clients position except for sensor2. To achieve final location estimation, the four sensors are combined in sets of 3s to produce four positions, as explained in section 4.10. Figure 6-20 provides the triangulation result for sensors 1, 2 and 4.
The distance error is 0.59 m, which is less than the 3m designed error. In the original result of figure 6-17 circle 4 is not overlapping with circle 2. This means, in figure 6-20 circle 2 has contracted, while circle 4 has expanded so both meet at a new position with circle 1 remaining unchanged. This can be clearly observed when the two figures (6-19 & 6-20) are compared. This process has brought the estimated position closer to the client. Similarly considering figure 6-21, instead of sensor 1, sensor 3 is introduced with sensor 2 and 4 to obtain a triangulation result from a combination of sensors 2, 3 and 4.
Again comparing this figure to 6-17, sensor 4 is not overlapping with either of the circles 2 or 3. Here in figure 6-19, circle 3 has expanded to meet circle 2 and circle 4. The error reported after triangulation is 3.9m.
Figure 6-21 Sensors 2,3 and 4 reported position of Client
Similarly combination of sensors 1,2 and 3 reports an error of 1.32m and sensors 1,3,4 reports an error of 1.13m. These results are shown in figures 6-22 and 6-23
Figure 6-22 Triangulation result by sensors 1,2 and 3.
Figure 6-23 Triangulation result for sensors 1,3 and 4
Estimated locations from all four set of sensors in figure 6-20 to 6-23, are averaged (4.10) and the final position is estimated which is given in figure 6-24. Although one set of circles (triangulation), figure 6-19, reported error of almost 3.9m averaging from all 4
Figure 6-24 Final error based on triangulation results from four set of sensors
We have observed through analysis of the previous figures that the algorithm accurately detects one or two walls between the client and the sensors and applies a correction accordingly. The forced overlapping of circles also reduces the location error.
A client at location 6 in figure 5-14 is discussed in figure 6-25. The outcome, if two of the four sensors report exaggerated values and how this affects the final result, will be examined for one of the worst case location estimates.