Abstract Continuous determination of position on modern
smartphones, in particular permanent utilization of GPS, considerably reduces operating times of batteries. However, incessant exploitation of power for determining the current location is avoidable, e.g., while the device is not being moved. This basic idea of considering the kinetic context of a device for localization enables the development of advanced localization techniques positively impacting energy consumption. This paper outlines the fundamental concept and the architecture for context-based energy saving strategies for continuous location determination on mobile devices, practically implemented on iOS. It proves their potentials in terms of energy consumption by a series of conducted tests, reflects on accuracy issues, and discusses reutilization for other platforms. 1. Introduction
Today, a great deal of mobile location-based services (LBS) is at consumers’ disposal available as apps for various types of modern smartphones. Built-in positioning technology for location determination in the mobile devices utilizing GPS, cellular network triangulation, or WLAN SSID mapping, is considered state-of-the-art and supports its consumers to find a path from A to B, recognize things around one’s own position, record one’s training routes, or exchange current whereabouts with friends.
Although, mobile LBS are considered an engaged paradigm in mobile computing environments, continuous utilization of LBS still faces a considerable drawback in terms of batteries’ operating times. An active localization module may significantly reduce the uptime of a smartphone compared to the operating times in standby mode. In addition, the devices generate substantial heat, which is not always perceived as pleasing. In most cases, permanent operation of GPS is mainly in charge for draining the battery. Abdesslem et al. [1] have measured a decline
from 170.6 hours in standby mode on a Nokia N95 8GB down to 7.1 hours (i.e., only 4.1%) when GPS was used outdoors. Oshin et al. [14] recorded a divergence of 284 to 12 hours uptime in standby mode and when GPS was active on an HTC Desire HD (i.e., only 4.2%). Barbeau et al. [3] have measured the energy-related impact of the GPS module on a Sanyo SCP-7050 with discrete operating intervals for calculating GPS fixes and examined that battery life was shortened from 33 hours uptime with an operating interval of 5 minutes down to 14 hours with a 60 seconds interval. Bareth and Kupper [4] once more identified the GPS module as the main energy guzzler in modern smartphones (in respect to the positioning technologies available, spending 6.6Ws for an active A-GPS module compared to 2.8Ws for WiFi and 1.0Ws for cellular-based localization).
Naturally, the transmission of data via the network also impacts operating times, e.g., when the device is negotiating for handovers to neighbored antennas or has to increase transmission power at low signal strength [5, 10]. Energy reduction for data transmission is not subject of this survey, though.
Thus, our main research issue for mobile LBS was to find a way to extend the batteries’ operating times of smartphones for continuous active operation, i.e., both when the user actively uses it (i.e., in foreground mode) and when the device is locked and put aside (i.e., in background mode). Continued operation in background mode is mandatory for a lot of mobile LBS applications as they actively alert users at selected locations or provide location services for other applications (e.g., when user positions are mutually exchanged). These requirements define a problem category that can be found in a manifold of commercially available mobile LBS apps, all facing the same problem of battery draining.
Of course, there cannot be a general answer or method to this issue because LBS strongly differ in terms of required accuracy and/or latency of reaction times. If a service e.g., claims permanent use of GPS due to exact location-based examinations (e.g., navigation systems), there is hardly any chance to
2014 47th Hawaii International Conference on System Science
develop a strategy for energy saving. Thus, we define a family of LBS applications and constraints to be fulfilled (related to the requirements given above) in order to be suitable to apply the energy saving strategies, which we propose in this paper. The constraints are defined as follows:
(i) The service is continuously operational, i.e., the position of the device must be determined both in foreground and in background mode. (ii) When in foreground mode, the best possible position of one’s own device is required. (iii) The service dynamically offers location-based points of interest (POIs), which have to be triggered, i.e., their content displayed, an alarm set off, an external service initiated, etc., when the device reaches spatial proximity to them. (iv) An observation mechanism is included within a network of participating clients, i.e., the service shows the current whereabouts of other floating devices. (v) When in observation mode (presuming that there is no permanent monitoring station), the best possible positions of the monitored devices are required.
This means, that within a distributed LBS environment an exact position is needed when the application is in foreground mode, observed by another client or close to POIs. For the remaining time (which represents the major part during operation in many cases, e.g., when the device is in background mode, not moving, not observed, or generally in a static position in the office or at night times, etc.), less energy consuming positioning methods (e.g., cellular network triangulation) could be used in order to extend the batteries’ operating times. Hence, we propose to avoid utilization of GPS as often as possible – a strategy already applied by earlier research, e.g., [1, 11, 16], however with a different approach and technology as the following sections will show. In order to prove our concept, we have implemented this paradigm for iOS devices, not only showing the potentials of the strategies but also revealing the drawbacks of less accurate localization in background mode and latency for the reactivation of GPS when required. The strategies are encapsulated in a social-web app named “Spotnick”, which is available in the App Store and on Google Play, and provides the measuring results included in this paper. For further details on Spotnick, please refer to the following sections.
The paper is structured as follows: Section 2 deals with selected points of state-of-the-art methods and technology. Section 3 gives an insight into the proposed energy saving strategies. Section 4 sketches the test scenarios and measuring metrics. Section 5 provides figures and results and finally, Section 6 concludes the paper and prospects future work.
2. Related Work
Investigations concerning energy consumption for location-based services on mobile devices have been the focal point of research in various scientific and industry labs [2, 3, 6, 19].
Generally, there are three ways to face the draining problem: (1) directly on a hardware-related level, where ameliorations are achieved e.g., by considering Kalman filters [18], (2) by several software-based and application-related high-level strategies utilizing SDK functions and frameworks [7, 8, 9, 21] and (3) with hybrid approaches using additional (less energy consuming) sensor technology (e.g., accelerometers and gyros) and context-based movement detection algorithms as a substitute or an extension of functionality [15, 17, 20].
The strategy described in this paper can be classified to approach (2) and regarded as an extension to existing similar methodologies: Kjaergaard et al. [11] e.g., have developed a context-based model predicting system conditions and movements in order to calculate a schedule for utilization of GPS. In a way, our approach can be compared to that, trying to schedule different positioning techniques due to system states (e.g., device is not moving). Although, the authors have succeeded in significantly saving energy using a Nokia N95, we give a more up-to-date approach using less complex algorithms and utilizing enhanced positioning techniques on an iOS device.
Oshin et al. [14] present a method to evaluate the energy-efficiency of GPS, WLAN, and GSM location sensing technologies of smart phones. Their model considers sensor usage, user context and location accuracy and utilizes the embedded accelerometer for managing activation and deactivation of the location sensors. It is therefore considered a hybrid approach (3). The authors state energy-savings of up to 57%.
Bareth and Küpper [4] claim ameliorations up to 90% compared to a “naive GPS approach” (please note, that authors use different relative reference values) using a hierarchical positioning algorithm dynamically deactivating positioning technologies and activating them with the least energy consumption on demand. Their approach (implemented for Android devices) and also the results are in some aspects comparable to the Spotnick strategy. iOS development, however (as the main platform for our survey), faces some restrictions in terms of direct access to different localization techniques and SDK availability and prevents direct translation of the theoretical thought models proposed by Bareth and Küpper. This paper reveals the workarounds.
946
3. Energy Saving Strategies
The energy saving strategies for our approach are settled on a high application-oriented level, variantly using, not using and switching between system functions and operations provided by SDKs. They are not based on hardware-related optimization attempts.
In principle, the strategy aims at avoiding usage of GPS whenever applicable. Instead, it uses cellular- or WiFi-based positioning or region monitoring (a technique which utilizes the WLAN and Cell-Id handover functions of the smartphone in order to detect a potential movement) and continuously tries to switch between those modes according to particular system events (context). Thus, the architecture principally results in a Finite State Machine (FSM) and has to take decisions for state transitions between those states built upon the following contextual properties and its characteristics: 1. Application mode: the app is either in foreground
mode (f) (the user looks at it) or in background mode (b) (the device is locked).
2. Movement: the user/device is either moving (m) or stationary (s) (not moving).
3. Observation mode: the device is either observed (o) or unobserved (u).
4. POIs nearby: points of interest for a device/user are either near (n) (i.e., within a distinct radius) or distant (d).
Thus, there are four properties having two values
each, creating 16 different states, strictly speaking (see Table 1). Most of them can be combined, though, because it is irrelevant for the strategy if the device is moving, observed or POIs are near or not when the app is in foreground mode (i.e., states 1 to 8 can be combined to one). In this state, the most accurate position is required. Also states 9 to 11 can be combined to one state, in which GPS is likely to be turned on, because the device is either being observed while in background mode (states 9 and 10) or a POI is near, the approach to it must be exactly determined (state 11). From a technical point of view, those two combinations of states 1 to 8 and 9 to 11 can be merged, once more, for they require the best possible position as a common characteristic (see state A in Figure 1). State 12 is unique: the app is in background mode, the device is moving, but not observed by any client, nor in the vicinity of any POI. This means that nobody takes notice of the app at that moment. Positioning must continue, though (by an inaccurate technique, see state B in Figure 1), in order to reactivate GPS at spatial proximity to a POI (which exactly requires to be triggered, transferred, etc.)
within the inaccuracy range of the used positioning method. Finally, states 13 to 16 can again be combined to one state where region monitoring is used in order to recover from an idle mode where the app is in the background and the device is not moving (observation and POIs are irrelevant at that state, see state C in Figure 1).
Table 1. Transition parameters of the FSM
No. App. mode
Move- ment
Obs. mode
POIs near
Consolidated States
1. f m o n
A
2. f m o d 3. f m u n 4. f m u d 5. f s o n 6. f s o d 7. f s u n 8. f s u d 9. b m o n
10. b m o d 11. b m u n 12. b m u d B 13. b s o n
C 14. b s o d15. b s u n 16. b s u d
Note that states B and C can only be reached when
the app is in background mode, while state A is accessible in both foreground and background mode. The consolidation of states reduces the complexity of the FSM to three states (A, B and C, see Figure 1).
foreground | observed | POI near |
Location Accuracy
Kilometer
Background Mode
Start
background & not moving &
Δt > 2min &
moving nd | d | r |
foregrounobservedPOI near
background & moving &
unobserved & no POIs &
Δt > 3min &
background & not moving
B
1
2
3
4
5
Location Accuracy
Best
Region Monitoring
A
mo5
C
BackgrounB k
background & not moving &Δt > 2min &
3
oving
Figure 1. State transitions of energy saving strategy (simplified excerpt)
947
Its state transitions are principally extractable from Table 1, which is a simplified depiction of the concept, though; the actual implementation considers several additional transition constraints (e.g., time) and works as follows: The strategy always starts in foreground mode and therefore with active GPS (state A, Figure 1). Please note that there is no SDK function for iOS to selectively activate or passivate GPS, cellular-, or WLAN-based localization. This must be done indirectly by specifying the desired accuracy. iOS provides the following six different accuracy values:
1. Best For Navigation 2. Best 3. Nearest Ten Meters 4. Hundred Meters 5. Kilometer 6. Three Kilometers Values 1 and 2 presumably, but not compulsory
turn on the GPS module. If a WiFi-based localization already provides the desired accuracy, iOS keeps GPS turned off. For the other values it is not transparent which positioning technique will be used by iOS. Values 5 and 6, however, are likely to activate cellular-based localization and turn off all other sensors. For our implementation we have used value 2 (“Best”) for state A and value 5 (“Kilometer”) for state B.
Starting from state A, the FSM transits to state B when the app goes to background mode, the user keeps moving for at least 3 minutes, but is not observed by another client nor in the vicinity of a POI (see transition 1 in Figure 1 and Formula 1).
A x ((b & m & u & d) & Δt > 3min) � B (1) The FSM transits back to state A when either the
app gets to foreground mode or the device is observed by at least one client or when a POI is near (no further explanation on that due to space limitations). Transition 2 in Figure 1 and Formula 2 explain.
B x (f | o | n) � A (2) When the device is not moving, anymore, in
background mode, then we switch to region monitoring (state C), i.e., the app is completely idle until a system signal wakes it up again due to a network handover event among WLAN access points or cellular antennas. In order to detect a non-moving device, we have defined a distance threshold ε, which may not be crossed, and a time interval Δt, which must be exceeded (here: 2 minutes) until we assume a stationary state where the device is not moving. See transition 3 in Figure 1 and Formula 3.
A x ((b & s) & Δt > 2min) � C (3) Almost the same is true for transiting from state B
to state C, except the time constraint. See transition 4 in Figure 1 and Formula 4.
B x (b & s) � C (4) Finally, when the app is in state C (i.e., in region
monitoring mode), there is only a way back to state A (best accuracy) when the system detects a network handover (i.e., a possible movement of the device), not to state B. See transition 5 in Figure 1 and Formula 5.
C x m � A (5) The FSM in Figure 1 is a simplified excerpt of the
actual implementation and omits further optimization details. Just to give one example: When transiting to a state requiring high accuracy, we do not immediately turn on GPS. We first check the accuracy value of the WiFi network. If this value is good (i.e., just a few meters inaccuracy, which indicates the presence of a registered WLAN network), then there is no need to activate GPS, e.g., when residing within a building where GPS is unavailable, anyway. Only a high inaccuracy value triggers the GPS module.
Figure 2. Spotnick user interface The energy saving strategy as described above has
been practically implemented in the core of a social-web application named Spotnick. This application has not only been developed in order to conduct investigations on energy consumption, it is embedded into a larger research issue on social behavior in distributed mobile LBS (for more information, please refer to e.g., [13]). Spotnick comprises a friend finder component for a user’s Facebook friends with the opportunity of mutually perceiving their whereabouts
948
with just a few seconds delay. When looking at the app, users are able to view their own position and those of their observed friends (see Figure 2, left screenshot) with maximum accuracy (depending on the devices’ positioning capabilities). In addition, a user can trigger so called spots, i.e., self-set POIs that actively announce the arrival of a user on his Facebook wall when he reaches spatial proximity to it (see Figure 2, right screenshot). Thus, Spotnick requires best possible accuracy when the app is in foreground mode, when observing friends or at regional closeness to spots. For the remaining time of usage (which we presume represents the major part of a day for a human being, e.g., when he/she is in his/her office or at night times), less accurate positioning methods comply in order to keep the app functional but not wasting energy. 4. Metrics and Test Scenarios
In order to benchmark the energy saving qualities of our proposed methods using our app Spotnick, we have repetitively conducted a series of tests, which were organized as follows: All tests have been carried out sequentially on the same device, a new iPhone 4S with iOS 6.01 as the operating system. All tests have used the same mini-app, which has recorded time stamps, WGS84 coordinates and the battery status via the iOS SDK in regular time intervals.
Although, we have not been able to provide exact equal test conditions for the particular tests due to cellular network fluctuations, arbitrary WLAN availability, or the speed or duration of our trips while moving through the city traffic, we have at least tried to provide similar test conditions with the following technical equipment and parameters (see Table 2):
Table 2. Test Parameters
Device iPhone 4S, iOS 6.01
(always same device, approx. 1 month old) Settings 3G enabled
WLAN always on Bluetooth always off
Usage no app running except test app no Internet surfing no phone calls or messages
Actions 20km driving 2x a day (always same route from home to office) app is generally in background mode 6x app in foreground mode for 3 min each (2x while driving, 2x in office, 2x at home) min. 8x observation by other clients (for each relevant state of the FSM)
Naturally, we have always used the same phone settings (i.e., WLAN on, Bluetooth off, 3G preferred but EDGE allowed, no phone calls, no messages, no Internet surfing, no other apps running) and have performed daily recurring procedures, i.e., driven the same route from home to the office, which is about 20km distance or approximately half an hour of driving time (depending on the traffic situation) twice a day (bidirectional). At home, in the office, and even while driving the app has been shortly used in foreground mode to see where other clients are residing (twice for each location and approx. 3 minutes each – simulating an assumed behavior of people using such an app). In addition, the device has been observed by 1 to 3 other clients (also for about 3 minutes each) in rather irregular time intervals, such that every possible status permutation (i.e., 8 different states given by the FSM, see Table 1) has been created at least once (app in foreground, in background, device moving, not moving, observed, not observed, while driving, when stationary).
With these prerequisites, four test series have been conducted (using the mini-app for data recording):
1. Standby Mode Test: This test should examine the
practical operating time when no app is running. It serves as a reference.
2. Full GPS Test: This test should examine the practical operating time when GPS is operational all the time.
3. Full GPS and Data Transmission Test: Same as 2, including the transmission of a simple HTTP data package every minute (simulating the transmission of position data for monitoring mode).
4. Spotnick Test: This test executed the app Spotnick, which contained the energy saving strategies described in the previous section.
5. Results
According to the four test scenarios the measured results are as follows: Table 3 lists the parameters and typical results for one of the conducted Standby Mode Tests. The test was started on January 29, 2013 and lasted for almost exact 6 days until the battery was drained completely. All location services were turned off (i.e., no GPS and no cellular or WLAN-based positioning, WLAN was turned on) and there was no data transmission during the test phase (i.e., no browsing, no mail traffic, no messaging, etc.). We measured an average decline of 0,7% battery power per hour and took these values as a reference for the subsequent tests (i.e., the operating times are referred to as 100%).
949
Table 3. Measurement results standby mode
iPhone 4S, iOS 6.01 Start Date Jan 29, 2013 Location Services off Data Transmission none, WLAN on Duration 6d Average Decline 0,7 % p.h. Relative Duration 100,0 %
Figure 3 illustrates the decline during the test
phase. Please note, that the tests were not started at midnight – it is just the scale that starts at 0:00 for an easier comparison. We recognize an almost clear linear decline of power. The slight variations are likely to depend on varying environmental conditions, e.g., attempts to connect to WLAN. However, we have not investigated these divergences in detail.
Figure 3. Battery consumption standby mode Test number 2 was the Full GPS Test. Table 4 lists
the parameters and results for a typical test run carried out on February 12, 2013. Continuous use of GPS and network-based localization (i.e., cellular triangulation and WLAN mappings) reduce operating times down to 11 hours and 40 minutes (which is only 8,1% of the time compared to the Standby Mode Test) with an average decline of 8,6% battery power per hour.
Table 4. Measurement results constant GPS
iPhone 4S, iOS 6.01 Start Date Feb 12, 2013 Location Services GPS+Network Data Transmission none, WLAN on Duration 11h 40min Average Decline 8,6 % p.h. Relative Duration 8,1 %
Figure 4 illustrates this drastic decline of battery operating times. For a better impression of the difference the original scale in this graph is the same as in Figure 3.
Figure 4. Battery consumption constant GPS However, it still goes worse. If we additionally
transmit small HTTP packets in regular time intervals (i.e., every minute in this example) containing an identification number, a timestamp, the current position and accuracy of the device’s positioning technique such that other observing clients may perceive the current whereabouts of this device (Full GPS and Data Transmission Test, see Table 5 and Figure 5) then we measure operating times of only 11 hours and 24 minutes.
Table 5. Measurement results for continuous GPS usage and HTTP requests every minute
iPhone 4S, iOS 6.01 Start Date Feb 14, 2013 Location Services GPS+Network Data Transmission Every Minute Duration 11h 24min Average Decline 8,8 % p.h. Relative Duration 7,9 %
Surprisingly, these results are marginally worse
compared to the Full GPS Test. Initially, we expected a more significant decline due to repetitively executed data transfer every minute (which makes 684 packages sent and the same amount of establishing a connection and closing it again). The effective operating time is now 7,9% compared to the Standby Mode Test, which is significant and reflects the measurements achieved by [1] (4,1% using a Nokia N95) and [14] (4,2% using an HTC Desire HD). The battery loses power by 8,8% per hour.
950
Figure 5. Battery consumption for continuous GPS usage and HTTP requests every minute
The measured results of tests 2 and 3 tell their own
tale in terms of energy consumption. It is obvious that continuous utilization of GPS (and beyond that coupled with regular data transmission) decisively decrease batteries’ operating times. Hence, the results for our proposed context-based energy saving strategies presented in section 3 are vital (regarding the Spotnick Test). They are consolidated in Table 6. The test lasted a bit longer than 3 days, which is 50,5% compared to the Standby Mode Test with an average decline of 1,4% of battery power per hour.
Table 6. Measurement results Spotnick
iPhone 4S, iOS 6.01 Start Date Feb 25, 2013 Location Services Spotnick Strategy Data Transmission On Position Updates Duration 3d 46min Average Decline 1,4 % p.h. Relative Duration 50,5 %
Figure 6 shows the corresponding graph.
Considering the descent of the curve it is clearly recognizable when GPS was operational (state A, e.g., after the first 24-hour vertical line), but for this snapshot it is not distinguishable whether the app was in foreground mode, or the device has been observed, or POIs have been near. The flattest parts of the curve represent the region monitoring states (state C), where the device has been stationary (e.g., at night times or in the office). If we e.g., take a look at the descent after the second 24-hour vertical line, we recognize a movement in background mode (state B) while not being observed and no POIs near. The length of the line (i.e., the duration of the movement) seems too long for a trip to the office, though. During our tests we have recognized that the transition from state B to C
(see Formula 4) is tricky in some cases due to ambiguous calculations regarding the detection of minimal or no movement (using variable thresholds ε for varying accuracy values of different positioning techniques). As a result, the FSM in some cases remains in state B longer than desired.
Figure 6. Battery consumption Spotnick Considering the accuracy of the calculated
positions, we refer to the successive two maps in Figure 7. The above picture (red line) shows a route detail driven in a car with GPS turned on (test no. 2: Full GPS Test). The picture below (blue line) shows the same route recorded using the Spotnick strategy.
Figure 7. Accuracy comparison GPS (red) vs.
Spotnick (blue)
951
The comparison clearly demonstrates the thought principle of the strategy: Less localization points that are sometimes less accurate need less power for determination, but are still enough for recognizing the covered route in order to trigger POIs and to turn on GPS when needed.
Figure 8 (top picture) shows a clipping of the daily way from our office at the University of Linz (starting beyond the top right corner of the map) to a home location (beyond the left edge). The red line marks the path measured by GPS. It represents the most accurate path we can get from the iPhone. The blue line marks the path from the Spotnick app including all energy saving mechanisms presented in section 3.
Figure 8. Accuracy comparison GPS (red) vs.
Spotnick (blue)
In urban areas (right half of the map) the recorded positions are clearly distinguishable from the GPS points, however, for this use-case still accurate enough to recognize an unambiguous movement along a particular city district. In rural areas (left half of the map) the recorded points were less in amount (because farther apart) and less accurate (because of imprecise cellular positioning). Nevertheless, an inaccurate point is only recognized for a short period of time until the moving device is informed about being monitored and therefore turns on GPS with a slight latency.
The picture in the middle of Figure 8 shows a switch from GPS to network positioning (see the blue line, clipped part out of the top picture). Coming from the top right corner, GPS is turned on due to an observing client and it has been turned off right before the first junction in this map. When GPS is off, the picture shows a deviation in the recorded points.
Finally, the bottom picture in Figure 8 shows Spotnick’s behavior outside city boundaries when in background mode. We recognize deviations to the GPS recorded route with more or less accurate values, however, also for rural areas, the strategy immediately provides an approximate position for potential sudden observers, who may perceive a friend’s position in the middle of a river for just a few seconds until the software recognizes being observed and turns on GPS.
Figure 9. Spotnick long-term strategy For short trips from home to our office and back
(about 20km or half an hour driving time) the strategy worked well, as for the major part of the day the device was idle in state C (see the results in Figure 6). In order to also prove the validity of the strategy for long trips, we have conducted several tests with 200km or 2,5 hours driving time for each direction (i.e., 400km a day
952
or 5 hours driving time in total). The results were worse than for the office trips, as expected, because of shorter idle states C. Nevertheless, we have managed to keep the batteries’ operational times significantly above 24 hours. Unfortunately, we cannot give a more precise figure due to the low number of tests conducted for long-term evaluations. Figure 9 gives an impression on the accuracy: We recognize single heaps of measured points along the route (e.g., when observed or in city areas with a greater antenna density). We also recognize long distances (several kilometers) without a measured point, which is likely to be due to a low antenna density in rural areas.
6. Summary, Comparison and Prospect
Continuous utilization of GPS on smartphones is draining batteries. Enhanced energy saving strategies considering the kinetic context of a device extend operating times. Figure 10 summarizes the results of our conducted tests with a theoretical optimum when the device is in standby (green line) and a worst result with GPS engaged and regular data transmission (red line; n.b. orange line nearly congruent with red line). In between is our energy saving strategy with Spotnick (blue line), which indicates a clear improvement compared to the exhaustive utilization of GPS.
Figure 10. Battery consumption iPhone 4S
These strategies also apply to further devices and systems. We have implemented them for Android devices, as well (see Narzt [12]). The technical details are different, because Android allows direct access to hardware-related functions. However, the basic concept is similar, and the results are even better compared iOS. Figure 11 summarizes the outcome of the tests for Android. Please note, that the Standby Mode Test only lasted for four days for this device (an HTC Desire HD), which has been taken as a reference value. The Spotnick Tests achieved an uptime of 2½ days, which is 65% compared to standby.
Figure 11. Battery consumption HTC Desire Table 7 compares the quality of our results to other
approaches. We recognize promising figures that are equal (for Bareth and Küpper [4]) or significantly better than alternative energy saving strategies.
Table 7. Comparison with other approaches
Approch, Device, OS Full
GPS [hours]
Stra-tegy
[hours]
Improvement [times]
Sa-vings
[%] Spotnick iOS iPhone 4S, iOS 6.01
11,4 72,7 6,4 84,3
Spotnick Android HTC Desire HD, Android 2.3.5
7,0 61,3 8,8 88,6
Abdesslem et al. [1] Nokia N95, Symbian
9,2 22,2 2,4 58,6
Oshin et al. [14] HTC Desire HD, Android 2.3.3
11,0 25,0 2,3 56,0
Bareth and Küpper [4] Motorola Milestone, Android 2.2
7,6 66,6 8,8 88,6
From the architectural perspective, all approaches
have been developed on application layer. For proper provision to other LBS apps facing similar problem categories as described initially, the strategies are supposed to be settled on a lower level (embedded into system functions). So, the next challenges in this field will be the consolidation of the various strategies developed and the integration of the most effective ideas into system libraries, abstracting from proprietary sensor systems and hardware features, always offering the most appropriate energy saving mechanisms independent from hardware or operating systems. 10. References [1] F. B. Abdesslem, A. Phillips, T. Henderson, “Less is More: Energy-Efficient Mobile Sensing with SenseLess”, ACM SIGCOMM Workshop on Networking, Systems, Applications on Mobile Handhelds (MobiHeld'09), Poster Session, Barcelona (Spain), August 2009.
953
[2] N. Balasubramanian, A. Balasubramanian, and A. Venkataramani, “Energy consumption in mobile phones: a measurement study and implications for network applica-tions,” Proceedings of the 9th ACM Internet measurement conference (IMC 2009), NY, USA, 2009, pp. 280-293. [3] S. J. Barbeau, M. A. Labrador, A. Perez, P. Winters, N. Georggi, D. Aguilar, and R. Perez, “Dynamic Management of Real-Time Location Data on GPS-enabled Mobile Phones,” Proceedings of the 2nd International Conference on Mobile Ubiquitous Computing, Systems, Services, and Technologies, (UbiComm 2008), Spain, 2008, pp. 343 – 348. [4] U. Bareth and A. Küpper, “Energy-Efficient Position Tracking in Proactive Location-Based Services for Smartphone Environments,” 35th Annual Computer Software and Applications Conference (COMPSAC), ISBN: 978-0-7695-4439-7, Munich, Germany, 2011, pp. 516-521. [5] C. Barrows, S. Blumsack, and R. Bent, “Using Network Metrics to Achieve Computationally Efficient Optimal Transmission Switching,” Proceedings of the 46th Hawaii International Conference on System Sciences (HICSS), Wailea, Maui, HI, USA, 2013, pp. 2187-2196. [6] A. Carroll and G. Heiser, “An Analysis of Power Consumption in a Smartphone,” Proceedings of the 2010 USENIX Annual Technical Conference, Boston, MA, USA, USENIX Association Berkeley, CA, USA, 2010, pp. 21-21. [7] I. Constandache, S. Gaonkar, M. Sayler, R. R. Choudhury, and L. Cox, “EnLoc: Energy-efficient localization for mobile phones,“ Proceedings of the IEEE INFOCOM Conference, ISSN: 0743-166X, E-ISBN: 978-1-4244-3513-5 Rio de Janeiro, 2009, pp. 2716-2720. [8] T. Farrell, R. Lange, and K. Rothermel, “Energy-efficient Tracking of Mobile Objects with Early Distance-based Reporting,” Proceedings of the 4th International Conference on Mobile and Ubiquitous Systems (MobiQuitous 2007), 2007, pp. 1-8. [9] S. Gobriel, C. Maciocco, C. Tai, and A. Min, “A EE-AOC: Energy Efficient Always-On-Connectivity Architec-ture,” Proceedings of the 8th International IARIA Conference of Wireless and Mobile Communications, ICWMC 2012, ISBN: 978-1-61208-203-5, Italy 2012, pp. 110-117. [10] J. Kellokoski, J. Koskinen, R. Nyrhinen, and T. Hamalainen, “Efficient Handovers for Machine-to-Machine Communications between IEEE 802.11 and 3GPP Evolved Packed Core Networks,” Int. Conf. on Green Computing and Communications (GreenCom), 2012, pp. 722-725. [11] M. Kjaergaard, J. Langdal, T. Godsk, and T. Toftkjaer, “EnTracked: energy-efficient robust position tracking for mobile devices,” Proceedings of the 7th Int. Conference on Mobile Systems, Applications, and Services, MobiSys 2009, New York, USA, ISBN: 978-1-60558-566-6, pp. 221-234.
[12] W. Narzt, “High-Level Energy Saving Strategies for Mobile Location-Based Services on Android Devices,” Proceedings of the 9th International IARIA Conference on Wireless and Mobile Communications, ICWMC 2013, Nice, France, July 21 - 26, 2013, in press. [13] W. Narzt and G. Pomberger: “Digital Graffiti – a Smart Information and Collaboration System”, Proceeding of the 2013 International Conference on Electronic Engineering and Computer Science, EECS 2013, Elsevier IERI Procedia (ISSN: 2212-6678), Beijing, China, 2013, in press. [14] T. O. Oshin, S. Poslad, and A. Ma, “A Method to Evaluate the Energy-Efficiency of Wide-Area Location Determination Techniques Used by Smartphones,” 15th International Conference on Computational Science and Engineering (CSE), Cyprus, 2012, pp. 326-333. [15] T. O. Oshin, S. Poslad, and A. Ma, “Improving the Energy-Efficiency of GPS Based Location Sensing Smart-phone Applications,” 11th International IEEE Conference on Trust, Security and Privacy in Computing and Communications (TRUSTCOM), UK, 2012, pp. 1698-1705. [16] J. Paek, J. Kim, and R. Govindan, “Energy-Efficient Rate-Adaptive GPS-based Positioning for Smartphones,” Proceedings of the 8th Annual International Conference on Mobile Systems, Applications, and Services (MobiSys 2010), San Francisco, CA, USA, 2010, pp. 299-314. [17] A. Rahmati and L. Zhong, “Context-for-wireless: context-sensitive energy-efficient wireless data transfer,” Proceedings of the 5th international conference on Mobile systems, applications and services, MobiSys 2007, New York, USA, ISBN: 978-1-59593-614-1, pp. 165-178. [18] I. M. Taylor and M. A. Labrador, “Improving the Energy Consumption in Mobile Phones by Filtering Noisy GPS Fixes with Modified Kalman Filters,” Proceedings of IEEE Wireless Communications and Networking Conference, March 28-31, 2011, pp. 2006-2011. [19] L. Wang and J. Manner, “Energy Consumption Analysis of WLAN, 2G and 3G interfaces,” Green Computing and Communications (GreenCom), 2010 IEEE/ACM Int'l Conference on Int'l Conference on Cyber, Physical and Social Computing (CPSCom), 2010, pp. 300-307. [20] M. Youssef, M. Yosef, and M. El-Derini, “GAC: Energy-Efficient Hybrid GPS-Accelerometer-Compass GSM Localization,” IEEE Global Telecommunications Conference (GLOBECOM 2010), ISSN: 1930-529X, Print ISBN: 978-1-4244-5636-92010, pp. 1-5. [21] Z. Zhuang, D. T. R, and L. Altos, “Improving Energy Efficiency of Location Sensing on Smartphones,” in Proceedings of the 8th international conference on Mobile systems, applications, and services, ISBN: 978-1-60558-985-5, 2010, pp. 315–330.
