Energy-Accuracy Trade-off for Continuous Mobile DeviceLocationKaisen LinUniversity of California San DiegoLa Jolla, CA, [email protected] Kansal, Dimitrios Lymberopoulos,and Feng ZhaoMicrosoft ResearchRedmond, WA, USA{kansal,dlymper,zhao}@microsoft.comABSTRACTMobile applications often need location data, to update locally rel-evant information and adapt the device context. While most smart-phones do include a GPS receiver, its frequent use is restricted dueto high battery drain. We design and prototype an adaptive locationservice for mobile devices, a-Loc, that helps reduce this batterydrain. Our design is based on the observation that the required lo-cation accuracy varies with location, and hence lower energy andlower accuracy localization methods, such as those based on WiFiand cell-tower triangulation, can sometimes be used. Our methodautomatically determines the dynamic accuracy requirement formobile search-based applications. As the user moves, both theaccuracy requirements and the location sensor errors change. A-Loc continually tunes the energy expenditure to meet the chang-ing accuracy requirements using the available sensors. A Bayesianestimation framework is used to model user location and sensorerrors. Experiments are performed with Android G1 and AT&TTilt phones, on paths that include outdoor and indoor locations, us-ing war-driving data from Google and Microsoft. The experimentsshow that a-Loc not only provides significant energy savings, butalso improves the accuracy achieved, because it uses multiple sen-sors.Categories and Subject DescriptorsG.3 [Probability and Statistics]: Probabilistic Algorithms; C.2.4[Computer-Communication Networks]: Distributed Systems—Distributed ApplicationsGeneral TermsAlgorithms, Design, Measurement, Experimentation1. INTRODUCTIONMobile applications often need location information and a largenumber of methods for mobile device localization have been devel-oped [23]. With GPS receivers becoming increasingly common-place in mobile phones and the widespread availability of WiFiand cell-tower signature based location services from Google [9]Permission to make digital or hard copies of all or part of this work forpersonal or classroom use is granted without fee provided that copies arenot made or distributed for profit or commercial advantage and that copiesbear this notice and the full citation on the first page. To copy otherwise, torepublish, to post on servers or to redistribute to lists, requires prior specificpermission and/or a fee.MobiSys’10, June 15–18, 2010, San Francisco, California, USA.Copyright 2010 ACM 978-1-60558-985-5/10/06 ...$10.00.and other providers, such location information is now becominga reality. However, mobile applications still cannot assume con-tinuous and ubiquitous location access in their design because ofthe high energy expense of using the location sensors such as GPSreceivers [12]. The variability in accuracy provided by various lo-cation sensors and the limits on their coverage areas pose addi-tional challenges for application developers. Using multiple loca-tion sensors simultaneously to make up for this variability in accu-racy would further increase energy use.Our goal is to develop location as a system service that automat-ically manages location sensor availability, accuracy, and energy.From an application developer perspective, this simplifies the useof the multiple existing, and potentially forthcoming, location tech-nologies with varying characteristics. From a mobile user experi-ence perspective, this allows the system to optimize battery life byintelligently managing the location energy and accuracy trade-offsbased on available sensor capabilities. This is beneficial for mo-bile platforms that allow several third party applications to run onthe platform, but at the same time must ensure long battery life foracceptable user experience.To realize the above goal, we develop an approach based ontwo observations. First, location applications do not always needthe highest available accuracy, such as that provided by GPS inopen sky view locations. The accuracy needs vary as the usermoves and we can exploit the slack in required accuracy to saveenergy. Second, a phone has multiple modalities to sense locationaside from the GPS: WiFi triangulation [16, 3], cell-tower trian-gulation [23], Bluetooth vicinity, audio-visual sensing [2], amongothers [4]. The availability and accuracy of these modalities varyas the user moves, and appropriate modalities can be selected toefficiently meet the location needs at lower energy costs.As an example scenario, consider a mobile search applicationthat wishes to display nearby “pizza” stores on the phone screen (orwithin its application tile) with latest coupons during meal times.This application must determine the nearest pizza stores to display.If the user is in a densely populated area with multiple pizza stores,a high accuracy is needed to correctly determine the nearest en-tries. However, if the user is in a remote area with few pizza stores,knowing the location to the nearest mile may suffice to determinethe correct entry. Clearly, in the latter case, a low energy locationmodality, simply based on cell tower association, could be used.Figure 1 illustrates the accuracy required at different locations inPortland, if the application was searching for the nearest five pizzastores. Variable accuracy requirements apply to most search basedscenarios where the mobile application is interested in significantentities around a user. In many of these applications, the user doesnot initiate a search but the application displays information proac-tively, such as show times for nearest movies or updates on other285Figure 1: Contour plot of accuracy requirements for findingnearest five pizza stores in Portland region (darker shades rep-resent higher accuracy requirement).location based activities that the user has subscribed to. Anotherscenario may involve the mobile device adapting its role based onthe user context where this context is resolved through distinguish-ing among locations with distinct activities, such as home, office,shopping mall, beach etc. Again, the required location accuracy de-pends on the geographical separation among these places. Variableaccuracy needs also emerge for idle screen advertising and socialnetworking applications [15]. For example, a social networkingapplication may be interested in alerting the user when friends arenearby,
View Full Document
Unlocking...