Academia.eduAcademia.edu

Figure 9 – uploaded by Alojz Kopáčik

See full PDFdownloadDownload figure
When position X,,Y, of pedestrian is corrected, the direction of pedestrian move- ment has to be taken into the account. It is used for a selection of the ideal part of a path. This condition prevents an incorrect adjustment of a point. It could happen in the case when the adjustment process would be limited only to searc- hing of the nearest point (Fig. 9).

Figure 9 When position X,,Y, of pedestrian is corrected, the direction of pedestrian move- ment has to be taken into the account. It is used for a selection of the ideal part of a path. This condition prevents an incorrect adjustment of a point. It could happen in the case when the adjustment process would be limited only to searc- hing of the nearest point (Fig. 9).

Related Figures (17)

To increase the long-term accuracy in the determination of the pedestrian position the Map Matching method was implemented into the calculation process. This met- hod uses information from maps for verification and correction of the determination of pedestrian position. The process of the map utilization is described in Section 3.3. Scheme of the whole process can be seen in the flowchart (Fig. 1) - Calculation of
To increase the long-term accuracy in the determination of the pedestrian position the Map Matching method was implemented into the calculation process. This met- hod uses information from maps for verification and correction of the determination of pedestrian position. The process of the map utilization is described in Section 3.3. Scheme of the whole process can be seen in the flowchart (Fig. 1) - Calculation of
Ilkoviéova, L.. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24 1, B, y -parameters of an adaptive step length estimation (individual for each user) ° — walking frequency ) — average amplitude of an acceleration norm
Ilkoviéova, L.. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24 1, B, y -parameters of an adaptive step length estimation (individual for each user) ° — walking frequency ) — average amplitude of an acceleration norm
Ilkoviéova, L.. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24
Ilkoviéova, L.. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24
Fig. 4. Change of walking frequency and change of average amplitude of acceleration for different types of walk. The integration of the angular velocity signal cumulates errors whose influence could be described by linear regression model. The initial determination of the model parameters could be determined when the smart phone is stable, staying in the same position without any rotation. These are calculated during the measure- ment from the signal when the smart phone is moved approximately with con- stant azimuth — relative long parts of the trajectory (lines) of min. 6 sec duration.
Fig. 4. Change of walking frequency and change of average amplitude of acceleration for different types of walk. The integration of the angular velocity signal cumulates errors whose influence could be described by linear regression model. The initial determination of the model parameters could be determined when the smart phone is stable, staying in the same position without any rotation. These are calculated during the measure- ment from the signal when the smart phone is moved approximately with con- stant azimuth — relative long parts of the trajectory (lines) of min. 6 sec duration.
Ilkoviéova, L. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24
Ilkoviéova, L. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24
Ilkoviéova, L. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-2¢ Using magnetometers could be determined the absolute orientation of the pede- strian during the whole measurement (Fig. 6) (Tian et al. 2014). The disadvanta- ge of these sensors is the high sensitivity to the magnetic field variations due to the steel structures and electrical equipment situated in a sensor neighbourhood. Because of this fact the magnetometers will be used only for approximate corre- ction of the absolute orientation of a pedestrian.
Ilkoviéova, L. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-2¢ Using magnetometers could be determined the absolute orientation of the pede- strian during the whole measurement (Fig. 6) (Tian et al. 2014). The disadvanta- ge of these sensors is the high sensitivity to the magnetic field variations due to the steel structures and electrical equipment situated in a sensor neighbourhood. Because of this fact the magnetometers will be used only for approximate corre- ction of the absolute orientation of a pedestrian.
Fig. 7. Determination of an orientation with drift elimination.
Fig. 7. Determination of an orientation with drift elimination.
There are many map matching algorithms, which have different approach to use the map information. Most of the applications used the map matching algorithm as a search problem, where find nearest point of the ideal path to the position calculated by IMS (White et al. 2000). In our work is map matching algorithm used only for the situation, where is the trajectory in the collision with floor map. There are used three conditions for finding correct position (current orientation of pedestrian, range of stationing of the corrected position and minimum length of the offset), which limited set of probably solution to one (see below).
There are many map matching algorithms, which have different approach to use the map information. Most of the applications used the map matching algorithm as a search problem, where find nearest point of the ideal path to the position calculated by IMS (White et al. 2000). In our work is map matching algorithm used only for the situation, where is the trajectory in the collision with floor map. There are used three conditions for finding correct position (current orientation of pedestrian, range of stationing of the corrected position and minimum length of the offset), which limited set of probably solution to one (see below).
Ilkoviéova, L.. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24 The trajectory of the pedestrian was calculated by 3 methods: 1) using a constant step length 2) using an adaptive step length 3) using an adaptive step length and the map matching algorith
Ilkoviéova, L.. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24 The trajectory of the pedestrian was calculated by 3 methods: 1) using a constant step length 2) using an adaptive step length 3) using an adaptive step length and the map matching algorith
Fig. 11. Trajectory of a pedestrian with constant step length. Fig. 11 shows a trajectory of a pedestrian calculated with the first method where the constant step length was used. It is possible to see that the worst results were achieved in the short parts of a trajectory (2-3, 3-4, 6-7, 7-8, 9-10, and 10-11) where the orientation of a pedestrian has been changed very often. The reason for decreasing accuracy of a pedestrian position and bad repeatability of a trajectory
Fig. 11. Trajectory of a pedestrian with constant step length. Fig. 11 shows a trajectory of a pedestrian calculated with the first method where the constant step length was used. It is possible to see that the worst results were achieved in the short parts of a trajectory (2-3, 3-4, 6-7, 7-8, 9-10, and 10-11) where the orientation of a pedestrian has been changed very often. The reason for decreasing accuracy of a pedestrian position and bad repeatability of a trajectory
Ilkoviéova, L.. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24
Ilkoviéova, L.. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24
In the second method of processing (Fig. 12) an adaptive step length estimation algorithm was used which improves the accuracy of a pedestrian position at the problematic parts of the trajectory (2-38, 3-4, 6-7, 7-8, 9-10, and 10-11). Improved results can be seen in the repeatability of the trajectory (a pedestrian walked through the predefined trajectory three times).
In the second method of processing (Fig. 12) an adaptive step length estimation algorithm was used which improves the accuracy of a pedestrian position at the problematic parts of the trajectory (2-38, 3-4, 6-7, 7-8, 9-10, and 10-11). Improved results can be seen in the repeatability of the trajectory (a pedestrian walked through the predefined trajectory three times).
The differences between used methods of processing are shown in each details of the trajectory (Fig. 14). The detail shows a part of the trajectory which con- sists from a short section with the different orientation. In this situation a step length was significantly changed. In this part of the trajectory is a high probability of the wrong selection of a part of an ideal path as well. For this rea- son, a condition for finding of the nearest point from an ideal path was designed (Fig. 10).
The differences between used methods of processing are shown in each details of the trajectory (Fig. 14). The detail shows a part of the trajectory which con- sists from a short section with the different orientation. In this situation a step length was significantly changed. In this part of the trajectory is a high probability of the wrong selection of a part of an ideal path as well. For this rea- son, a condition for finding of the nearest point from an ideal path was designed (Fig. 10).
Table 2. Comparison of the coordinate difference on measured points (during the last repetition trajectory). Adaptive step length estimation improved the accuracy of travelled distance. This fact is possible to see in the results in the Table 3. The error of total travelled distance calculated with a constant step length was 2.7 m (1.3% of total travelled distance) but with an adaptive step length estimation error decreased to 1.2 m (0.2% of total travelled distance). This difference in the accuracy of the individual methods is influenced by the variation of a step length which depends on the speed of pedestrians and it is also influenced by environment (entering the room, avoiding the obstacles). Calculations with help of Map Matching are not a part of the comparison (Table 3) because it does not influence the accuracy of travelled distance. The position of a pedestrian is directly corrected.
Table 2. Comparison of the coordinate difference on measured points (during the last repetition trajectory). Adaptive step length estimation improved the accuracy of travelled distance. This fact is possible to see in the results in the Table 3. The error of total travelled distance calculated with a constant step length was 2.7 m (1.3% of total travelled distance) but with an adaptive step length estimation error decreased to 1.2 m (0.2% of total travelled distance). This difference in the accuracy of the individual methods is influenced by the variation of a step length which depends on the speed of pedestrians and it is also influenced by environment (entering the room, avoiding the obstacles). Calculations with help of Map Matching are not a part of the comparison (Table 3) because it does not influence the accuracy of travelled distance. The position of a pedestrian is directly corrected.
Ilkoviéova, L. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24
Ilkoviéova, L. et al.: Pedestrian Indoor Positioning and Tracking using..., Geod. list 2016, 1, 1-24
Fig. 16. Cycle for the height change detection and actualization of floor map. due to the low resolution of the pressure sensor. That is why the process of actu- alization of the height of a pedestrian will be realized only in such situations when the pedestrian will be located on stairs (or on another place where the height is changed) (Tanigawa et al. 2008). The floor map will be updated according to the actual height of a pedestrian (Fig. 16).
Fig. 16. Cycle for the height change detection and actualization of floor map. due to the low resolution of the pressure sensor. That is why the process of actu- alization of the height of a pedestrian will be realized only in such situations when the pedestrian will be located on stairs (or on another place where the height is changed) (Tanigawa et al. 2008). The floor map will be updated according to the actual height of a pedestrian (Fig. 16).
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved