Academia.eduAcademia.edu

Figure 5 – uploaded by Ahmed BinSahaq

See full PDFdownloadDownload figure
Figure 7 Hybrid approach. GPRS and Wi-Fi communication coverage (see online version for colours)

Figure 7 Hybrid approach. GPRS and Wi-Fi communication coverage (see online version for colours)

Related Figures (19)

Source: Dutta et al. (2009) Figure 1 CommonSense monitoring handheld devices (see online version for colours)
Source: Dutta et al. (2009) Figure 1 CommonSense monitoring handheld devices (see online version for colours)
Figure 2 Sensor board on tram and city CO pollution map from OpenSense system (see online version for colours)
Figure 2 Sensor board on tram and city CO pollution map from OpenSense system (see online version for colours)
Table 1 Comparison of mobile sensing systems periodically collected and sent in a store-and-forward fashion whereby collected data are buffered until the vehicle is within the range of an access point. Then, data are forwarded to a central server for additional analysis and data mining. Access points are located at road intersections and traffic lights. knowledge exists in our case. Table 1 provides a summary of the systems mentioned previously along with a simple explanation of how each works.
Table 1 Comparison of mobile sensing systems periodically collected and sent in a store-and-forward fashion whereby collected data are buffered until the vehicle is within the range of an access point. Then, data are forwarded to a central server for additional analysis and data mining. Access points are located at road intersections and traffic lights. knowledge exists in our case. Table 1 provides a summary of the systems mentioned previously along with a simple explanation of how each works.
Source: Devarakonda et al. (2013) 3 System model In OpenSense (Li et al., 2012; Aberer et al., 2010), the sensors (buses with sensing units) move according to a predetermined time-route table (bus and tram routes) and perform periodic sensing. This system is also the case for Ear-Phone (Rana et al., 2010) in which mobile sensors move along roads. In our work, we assume that sensors move in a random way and that a mobile sensor may not visit the same location more than once. In OpenSense mobile sensor (e.g., tram and buses) move in well-known routes. Therefore, in OpenSense, the possibility that a sensor visits the same location is high which is not the case in our work. We have no control on sensor movement or expectation to which locations the sensors may move. If the sensor passes by a mission’s location many times, it will improve the coverage of that mission over the network lifetime.
Source: Devarakonda et al. (2013) 3 System model In OpenSense (Li et al., 2012; Aberer et al., 2010), the sensors (buses with sensing units) move according to a predetermined time-route table (bus and tram routes) and perform periodic sensing. This system is also the case for Ear-Phone (Rana et al., 2010) in which mobile sensors move along roads. In our work, we assume that sensors move in a random way and that a mobile sensor may not visit the same location more than once. In OpenSense mobile sensor (e.g., tram and buses) move in well-known routes. Therefore, in OpenSense, the possibility that a sensor visits the same location is high which is not the case in our work. We have no control on sensor movement or expectation to which locations the sensors may move. If the sensor passes by a mission’s location many times, it will improve the coverage of that mission over the network lifetime.
Figure 4 Network model. 100 mobile sensors, 10 missions and one base-station (see online version for colours) At each time unit, a single mission /; may receive utility from one or more sensors. The utility received by M;, which we denote as u,, is defined as follows:
Figure 4 Network model. 100 mobile sensors, 10 missions and one base-station (see online version for colours) At each time unit, a single mission /; may receive utility from one or more sensors. The utility received by M;, which we denote as u,, is defined as follows:
Table 2 Sensor node characteristics of the waspmote sensor platform
Table 2 Sensor node characteristics of the waspmote sensor platform
In this section, we evaluate the performance of an MWSN with an event detection (e.g., chemical leakage or fire detection) application as discussed in Section 3. Network sensors use different sensing and data delivery schemes as discussed in Section 4.2. We evaluate these different data delivery schemes to see which is more suitable for MWSNs. We also show results comparing threshold-based sensing to an energy-aware scheme.
In this section, we evaluate the performance of an MWSN with an event detection (e.g., chemical leakage or fire detection) application as discussed in Section 3. Network sensors use different sensing and data delivery schemes as discussed in Section 4.2. We evaluate these different data delivery schemes to see which is more suitable for MWSNs. We also show results comparing threshold-based sensing to an energy-aware scheme.
Table 3. Simulation parameters
Table 3. Simulation parameters
Figure 8 Expected utility contribution (e;;) according to the distance (D;;) between sensor S; and mission M/; We performed a set of experiments to evaluate the network’s performance. The results are based on the network running for a period of 1800 s averaged over 30 runs.
Figure 8 Expected utility contribution (e;;) according to the distance (D;;) between sensor S; and mission M/; We performed a set of experiments to evaluate the network’s performance. The results are based on the network running for a period of 1800 s averaged over 30 runs.
Using a GPRS connection eliminates the effect of the deadline interval (T’) on utility received by missions because all data packets are received with only a small delay. We found that increasing the number of missions had little effect on the average utility each mission received (with the assumption that sensors have a sufficient energy supply). This is due to the omnidirectional operation of sensors. As described earlier, a gas sensor reading can be used for detecting pollution levels in an area within the radius of its sensing range. Hence, all points (e.g., missions in our case) within that area are covered. This is not the case when using sensors that are directional, such as cameras, where the sensor can only contribute to the closest mission and other missions receive zero utility even if they lie within the sensor’s range. Therefore, henceforth, we show results for only 30 missions except where the number of missions makes a difference.
Using a GPRS connection eliminates the effect of the deadline interval (T’) on utility received by missions because all data packets are received with only a small delay. We found that increasing the number of missions had little effect on the average utility each mission received (with the assumption that sensors have a sufficient energy supply). This is due to the omnidirectional operation of sensors. As described earlier, a gas sensor reading can be used for detecting pollution levels in an area within the radius of its sensing range. Hence, all points (e.g., missions in our case) within that area are covered. This is not the case when using sensors that are directional, such as cameras, where the sensor can only contribute to the closest mission and other missions receive zero utility even if they lie within the sensor’s range. Therefore, henceforth, we show results for only 30 missions except where the number of missions makes a difference.
Node grouping inhibited the nodes’ ability to spread across a larger geographic area, which is reflected in the sparse coverage of the missions deployed near border areas. By contrast, the RWP model covers larger areas, and missions are likely to be covered better than in the RPGM model because the sensors can move individually. number of nodes inside each group varies (2, 3, and 5 nodes) within a maximum distance of 40 m from the group’s centre. There was zero probability of a node changing its group to keep node density constant in each group. Figure 10 shows that RWP outperforms RPGM due to node grouping. As the number of group members increases, the possibility that group members all sense the same missions increases, in which case redundant sensory data may be received for a single mission. This increase in group members does improve utility for that mission, but other missions may not receive any utility, which affects the overall performance. This effect becomes obvious when we decrease the number of nodes within each group.
Node grouping inhibited the nodes’ ability to spread across a larger geographic area, which is reflected in the sparse coverage of the missions deployed near border areas. By contrast, the RWP model covers larger areas, and missions are likely to be covered better than in the RPGM model because the sensors can move individually. number of nodes inside each group varies (2, 3, and 5 nodes) within a maximum distance of 40 m from the group’s centre. There was zero probability of a node changing its group to keep node density constant in each group. Figure 10 shows that RWP outperforms RPGM due to node grouping. As the number of group members increases, the possibility that group members all sense the same missions increases, in which case redundant sensory data may be received for a single mission. This increase in group members does improve utility for that mission, but other missions may not receive any utility, which affects the overall performance. This effect becomes obvious when we decrease the number of nodes within each group.
As expected, a hybrid approach behaves similarly toa GPRS connection alone. In the hybrid approach, GPRS is always used in the absence of a base station. Therefore, all data packets are delivered either using a Wi-Fi access point (when available) or using a GPRS connection.
As expected, a hybrid approach behaves similarly toa GPRS connection alone. In the hybrid approach, GPRS is always used in the absence of a base station. Therefore, all data packets are delivered either using a Wi-Fi access point (when available) or using a GPRS connection.
Figure 13. Energy-aware vs. threshold-based sensing schemes for 30 missions, 200 nodes and a GPRS connection: (a) T = 0.50 and (b) t = 0.25 (see online version for colours)
Figure 13. Energy-aware vs. threshold-based sensing schemes for 30 missions, 200 nodes and a GPRS connection: (a) T = 0.50 and (b) t = 0.25 (see online version for colours)
the threshold-based scheme outperforms the energy-aware scheme in average utility received by each mission during the network lifetime. However, the energy-aware scheme preserves the nodes’ energy better than the threshold-sensing scheme ( all energy-aware nodes are still alive at the end of the simulation time when 7 = 25%, but only approximately 30% are alive in the case of the threshold-based scheme as shown in Figure 15(b).). sensors become more conservative and increase thei r sensing thresholds (see equation (3)). Thereafter, they sense only missions with high utility expectations. This strategy reduces energy spent on less useful missions when battery evels are low, which extends the network’s lifetime even when small thresholds are used, as shown in Figure 13(a) and (b). Figure 14 shows the average utility received by each mission throughout the entire network lifetime. It is clear that the energy-aware scheme performs better than pure threshold-based sensing when a low threshold value is used. However, the threshold- based scheme does well with a high threshold; when T =50%, the threshold-based scheme is almost identical to the energy- aware scheme.
the threshold-based scheme outperforms the energy-aware scheme in average utility received by each mission during the network lifetime. However, the energy-aware scheme preserves the nodes’ energy better than the threshold-sensing scheme ( all energy-aware nodes are still alive at the end of the simulation time when 7 = 25%, but only approximately 30% are alive in the case of the threshold-based scheme as shown in Figure 15(b).). sensors become more conservative and increase thei r sensing thresholds (see equation (3)). Thereafter, they sense only missions with high utility expectations. This strategy reduces energy spent on less useful missions when battery evels are low, which extends the network’s lifetime even when small thresholds are used, as shown in Figure 13(a) and (b). Figure 14 shows the average utility received by each mission throughout the entire network lifetime. It is clear that the energy-aware scheme performs better than pure threshold-based sensing when a low threshold value is used. However, the threshold- based scheme does well with a high threshold; when T =50%, the threshold-based scheme is almost identical to the energy- aware scheme.
In Figures 17 and 18, the sensor uses both communication modules (Wi-Fi and GPRS) interchangeably to transmit data packets. The addition of the Wi-Fi module (the hybrid
In Figures 17 and 18, the sensor uses both communication modules (Wi-Fi and GPRS) interchangeably to transmit data packets. The addition of the Wi-Fi module (the hybrid
In Figure 17, at the beginning, the threshold-based scheme performs better than the energy-aware scheme because it does not have to consider energy levels. Consequently, the number of live nodes decreases drastically when their energy is rapidly depleted, as shown in Figure 17(a) and (b). However, the energy-aware scheme expends energy only on valuable missions by increasing the sensing threshold. This strategy helps to extend the network lifetime. At the end of the simulation, approximately 95% of the nodes are still alive, as shown in Figure 17(b). However, this is reflected in the average utility received by missions throughout the network lifetime, as shown in Figure 18. Although the energy-aware scheme saved energy, its performance was somewhat below the threshold method when using a high threshold (50%). Conversely, when using a low threshold (25%), the energy- aware scheme performs better because it expends energy only for higher utility missions. approach) improved the node’s battery life. By the end of the simulation, the percentage of still-alive nodes in the hybrid approach was approximately 30% as shown in Figure 17(a). However, when only a GPRS connection was used, the percentage of still-alive nodes decreased to approximately 13%, as shown in Figure 13(a).
In Figure 17, at the beginning, the threshold-based scheme performs better than the energy-aware scheme because it does not have to consider energy levels. Consequently, the number of live nodes decreases drastically when their energy is rapidly depleted, as shown in Figure 17(a) and (b). However, the energy-aware scheme expends energy only on valuable missions by increasing the sensing threshold. This strategy helps to extend the network lifetime. At the end of the simulation, approximately 95% of the nodes are still alive, as shown in Figure 17(b). However, this is reflected in the average utility received by missions throughout the network lifetime, as shown in Figure 18. Although the energy-aware scheme saved energy, its performance was somewhat below the threshold method when using a high threshold (50%). Conversely, when using a low threshold (25%), the energy- aware scheme performs better because it expends energy only for higher utility missions. approach) improved the node’s battery life. By the end of the simulation, the percentage of still-alive nodes in the hybrid approach was approximately 30% as shown in Figure 17(a). However, when only a GPRS connection was used, the percentage of still-alive nodes decreased to approximately 13%, as shown in Figure 13(a).
Related topics:
Computer ScienceDistributed ComputingGeneral packet radio service (GPRS)
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved