Contextaware battery management for mobile phones N Ravi
Context-aware battery management for mobile phones N. Ravi et al. , Conf. on IEEE International Pervasive Computing and Communications, pp. 224 -233, 2008. 이시혁 theshy@sclab. yonsei. ac. kr
Contents • Background • Proposed system • CABMAN system design – Overview – Sysem-specific components – Charging opportunity predictor(cont. ) – Call time predictor – Battery life time predictor – Viceroy and user interface • Evaluation – Environment – Charging opportunity predictor – Call time predictor – Battery-lifetime predictor • Discussion & Conclusions S FT COMPUTING @ YONSEI UNIV. KOREA 16
Background • Improving rapidly for mobile device(such as smartphone) – Processing power – Storage capacities – Graphics – High-speed connectivity • Faced battery capacities – Not experiencing the exponential growth curve as other technologies – Remaining the key bottleneck for mobile devices in the near future 2 S FT COMPUTING @ YONSEI UNIV. KOREA 16
Background • Current solutions – “Battery low” audio signal – A remaing time estimate at current power consumption – User interface : unchanged for a number of years • The reasons for need to be changed – Convergence makes more multi-functional computing devices – WLAN interfaces are relatively hungry consumers of energy – Pervasive computing applications have provided reasons for mobile devices to be executing always-on background applications 3 S FT COMPUTING @ YONSEI UNIV. KOREA 16
Proposed system • CABMAN (Context-Aware Battery MANagement architecture for mobile device) • Battery management architecture – Crucial applications to users should not be compromised by noncrucial applications. – The opportunities for charging should be predicted instead of using absolute battery level as the guide – Context can be used to predict charging opportunities • Goal for system – the next charging opportunity – the call time requirements of the user over a period of time (assuming that telephony is the most critical application) – the “discharge speedup factor” of the set of non-crucial applications running. 4 S FT COMPUTING @ YONSEI UNIV. KOREA 16
CABMAN system design Overview • 3 -categories – System-specific monitors – Predictors – The viceroy/UI • Consist of 8 -components The viceroy/UI Predictors Monitors 5 S FT COMPUTING @ YONSEI UNIV. KOREA 16
CABMAN system design Sysem-specific components • Context monitor : sensing and storing context information • Call monitor : log communication – incoming/outcoming calls – Incoming/outgoing SMSs • Process monitor : tracks the processes running on the device • Battery monitor : probe and enquire about remaining charge and voltage level 6 S FT COMPUTING @ YONSEI UNIV. KOREA 16
CABMAN system design Charging opportunity predictor • Determine the charging opportunity for crucial application – True, CABMAN should not inconvenience the user with unnecessary warnings or actions – False, If the phone battery if relatively full, CABMAN should warn the user that they risk a dead battery • Location sensing – A way of inferring charging opportunity – Disadvantage of using only location : it does not accommodate for mobile chargers(such as car) – Additional context information • • Time- of-day Speed Presence of other wireless devices Charge-logs 7 S FT COMPUTING @ YONSEI UNIV. KOREA 16
CABMAN system design Charging opportunity predictor(cont. ) • Cell-based charging opportunity predictior – Dectection of other beacon type • Wifi • APs – Direct positioning information • GPS • A-GPS – Detecting the id of the current cell (e. g. those at home or perhaps the work place) • marking the cells in which this normally occurs • if the user often “refuses”, then the cell can be unmarked G • Examples – Currnet samples : ABC – History : DEABCFG F E D A B C 8 S FT COMPUTING @ YONSEI UNIV. KOREA 16
CABMAN system design Call time predictor • To protect the availability of telephony – Crucial application – The call time needs of the user should be predicted • Methods to predict the call time – Static : Ask to the user, set a minimum call time level – Dynamic : find the average of number of minutes of call time (each hour of the day) – Hybrid : Static+Dynamic 9 S FT COMPUTING @ YONSEI UNIV. KOREA 16
CABMAN system design Battery life time predictor • Difficult to predict battery life time – Different chemistry of the battery – Use of the applications with different battery demand • Measure the base curve in idle mode – New laptop – Old laptop – HP i. PAQ Linear Non-linear spiky 10 16 S FT COMPUTING @ YONSEI UNIV. KOREA
CABMAN system design Viceroy and user interface • Viceroy : CABMAN’s central component – Continually monitor whether the bettery lifetime prediction – combined with the battery requirement of the estimated call time requirement from the call time predictor – means that the battery will expire before or after the next charging opportunity • Warning t > r−f(m) t: an estimation of the time interval before the next charging opportunity surfaces r: an estimate of the remaining battery Lifetime m : an estimate of the required calltime f(m): the map from call time to battery lifetime. 11 16 S FT COMPUTING @ YONSEI UNIV. KOREA
Evaluation Environment • CABMAN prototype – Linux – Symbian OS • MIT’s Reality Mining project – charging opportunity predictor – call time predictor – gathered by deploying Nokia 6600 phones – 80 subjects for around nine months 12 16 S FT COMPUTING @ YONSEI UNIV. KOREA
Evaluation Charging opportunity predictor • Settings – Half the subjects : single charging station – Other half : two charging stations 13 16 S FT COMPUTING @ YONSEI UNIV. KOREA
Evaluation Call time predictor weekdays S The length of phone calls FT COMPUTING @ YONSEI UNIV. KOREA weekends The number of calls made during each hour 14 16
Evaluation Battery-lifetime predictor • Base curve together with discharge curves (actual and derived) for the new HP laptop Old Dell laptop Comparing accuracy of our algorithm with ACPIs HP i. PAQ S FT COMPUTING @ YONSEI UNIV. KOREA 15 16
Discussion & Conclusions • Charging-opportunity predictor and call-time predictor perform reasonably well for an average user whose life entropy is not very high. • Unfixed charging place (e. g. car) • Describe three key components of CABMAN: – The use of context information such as location to predict the next charging opportunity – More accurate battery life prediction based on a discharge speedup factor – The notion of crucial applications such as telephony 16 16 S FT COMPUTING @ YONSEI UNIV. KOREA
- Slides: 17