Full On-Device Stay Points Detection in Smartphones for Location-Based Mobile Applications

Area of Science:

• Computer Science
• Mobile Computing
• Ubiquitous Computing

Background:

• Location-aware mobile applications rely on detecting frequently visited places (stay points) for personalized services.
• Current methods often use a Mobile Cloud Computing (MCC) approach, transferring GPS data to servers for analysis, which can be energy-intensive.
• On-device stay point detection has been considered computationally prohibitive for smartphones.

Purpose of the Study:

• To propose and validate an event-driven, on-device mechanism for stay point detection.
• To demonstrate higher energy savings compared to MCC solutions by avoiding communication overhead.
• To develop a sensing middleware for Android smartphones to facilitate on-device trajectory analysis.

Main Methods:

• Developed an Android sensing middleware for background GPS data collection and duty cycling.
• Implemented an event-driven paradigm for incremental, on-device analysis of location updates.
• Conducted real-world experiments under various stress levels to evaluate performance.

Main Results:

• The proposed on-device, event-driven mechanism for stay point detection is feasible.
• The system demonstrated significant power efficiency compared to MCC-based solutions.
• The sensing middleware effectively supports background data collection and incremental analysis.

Conclusions:

• On-device stay point detection is achievable and offers superior energy savings.
• The developed event-driven middleware provides an efficient alternative to MCC for location-aware services.
• This approach enhances smartphone battery life by reducing communication costs for trajectory analysis.