Energy-Efficient Positioning for Smartphones using Cell

Report
Energy-Efficient Positioning for
Smartphones using Cell-ID Sequence
Matching
Jeongyeup Paek, Kyu-Han Kim,
Jatinder P. Singh, Ramesh Govindan
MobiSys 2011
Motivation
• Localization:
– Cell-tower: less accurate, but energy-efficient
• Error range up to several Km
– GPS: accurate, but energy-hungry
• Quickly drains the battery (<12hours)
• Extra information:
– Sensor data: accelerometer, Wireless data, etc.
– User behavior: predictable mobility patterns
• Goal: design a lightweight positioning system that
exploits the predictable user mobility pattern
– E.g., commuting via buses (Nature paper)
Cell-tower Localization
• Invoked network-based localization for every 2s, and
recorded the position in the map
• Results: less accurate, not frequent enough samples
Cell-tower Localization
GPS Energy Consumption
• Energy-hungry GPS
• Plus, periodic or adaptive duty-cycling may not achieve
significant energy savings under all conditions
– E.g., due to on/off delay (cf., NSDI paper like back-of-theenvelope calculation is feasible)
CAPS Design
• User’s spatio-temporal mobility pattern + CellID transition history
CAPS Design
• Spatial GPS sampling at the cell-ID transition
points (and location estimation based on these
inputs)
– Assuming constant speed (and further map data can
be used)
CAPS Design: Overview
CAPS Design: Cell-ID Sequencing
• Sequence learning:
– Records <cell-ID, GPS position, timestamp>
– Segments records with threshold size (e.g., 10, 20)
• Sequence matching:
– Smith-Waterman algorithm used for gene sequence matching in
bioinformatics
– Score function: gap, match, mismatch
– Constraint: last cell-ID must match (CID 6)
• Sequence selection:
– Highest scored item
– Tie-breaking: time-of-day > longer sequence > etc..
– If no valid match is found, GPS is turned on; GPS is turned off if
its error range falls below threshold (e.g., 80m)
CAPS Design: Cell-ID Sequencing
(-0.5 for gap, mismatch)
CAPS Design: Cell-ID Sequencing
• Trade-offs
– Match sequence length
– Penalty scores
Evaluation
Evaluation
USC-HSC route
Discussion
• Map matching can be used..
• Cell-tower signal strength?
– Commodity smartphones do not export celltower
locations and signal strength from multiple
celltowers to the application programmer.
• Managing the sequence database may be an
important issue for CAPS

similar documents