Mobile Computing, Applications, and Services. 5th International Conference, MobiCASE 2013, Paris, France, November 7-8, 2013, Revised Selected Papers

Research Article

Intelligent Energy-Efficient Triggering of Geolocation Fix Acquisitions Based on Transitions between Activity Recognition States

Download36 downloads
  • @INPROCEEDINGS{10.1007/978-3-319-05452-0_9,
        author={Thomas Phan},
        title={Intelligent Energy-Efficient Triggering of Geolocation Fix Acquisitions Based on Transitions between Activity Recognition States},
        proceedings={Mobile Computing, Applications, and Services. 5th International Conference, MobiCASE 2013, Paris, France, November 7-8, 2013, Revised Selected Papers},
        proceedings_a={MOBICASE},
        year={2014},
        month={6},
        keywords={},
        doi={10.1007/978-3-319-05452-0_9}
    }
    
  • Thomas Phan
    Year: 2014
    Intelligent Energy-Efficient Triggering of Geolocation Fix Acquisitions Based on Transitions between Activity Recognition States
    MOBICASE
    Springer
    DOI: 10.1007/978-3-319-05452-0_9
Thomas Phan1,*
  • 1: Samsung Research America - Silicon Valley
*Contact email: thomas.phan@samsung.com

Abstract

Location-based applications (LBAs) running on smartphones offer features that leverage the user’s geolocation to provide enhanced services. While there exist LBAs that require continuous geolocation tracking, we instead focus on LBAs such as location-based reminders or location-based advertisements that need a geolocation fix only at rare points during the day. Automatically and intelligently triggering geolocation acquisition just as it is needed for these types of applications produces the tangible benefit of increased battery life. To that end, we implemented a scheme to intelligently trigger geolocation fixes only on transitions between specific modes of transportation (such as driving, walking, and running), where these modes are detected on the smartphone using a low-power, high-resolution activity recognition system. Our experiments show that this approach consumes little power (approximately 225 mW for the activity recognition system) and correctly triggers geolocation acquisition at transitional moments with a median delay of 9 seconds from ground-truth observations. Most significantly, our system performs 41x fewer acquisitions than a competitive accelerometer-assisted binary classification scheme and 243x fewer than continuous tracking over our collected data set.