Download presentation
Presentation is loading. Please wait.
Published byOphelia Palmer Modified over 9 years ago
1
Microsoft Research Faculty Summit 2007
2
John Krumm Microsoft Research Redmond, WA
3
55 GPS receivers 241 subjects 1.97 million points 106,000 miles 171,000 kilometers 13,845 trips Home addresses and demographic data Greater Seattle Seattle DowntownClose-up Garmin Geko 201 $115 10,000 point memory Median recording interval 6 seconds 63 meters
4
Destination Modeling Predestination – Destination prediction Snap-to-Road – Map matching with temporal constraints Personalized Routes Location Privacy
5
Destinations of drivers in our location survey John Krumm and Eric Horvitz, "Driver Destination Models", Eleventh International Conference on User Modeling (UM 2007), June 25-27, 2007, Corfu, Greece.
6
U.S. Geological Survey – Seattle Area What are the most attractive kinds of ground cover?
8
Time of DayDay of Week
9
Rate of Decline versus Demographics Single versus partner – no significant difference Children versus no children – no significant difference Extended family nearby versus not – no significant difference Gender – women decline faster than men Drivers reach steady state after about two weeks
10
Destination Modeling Predestination – Destination prediction Snap-to-Road – Map matching with temporal constraints Personalized Routes Location Privacy
11
John Krumm and Eric Horvitz, "Predestination: Inferring Destinations from Partial Trajectories", Eighth International Conference on Ubiquitous Computing (UbiComp 2006), September 2006.
12
Anticipatory information Location-based advertising Hybrid vehicle efficiency Traffic Warning Destination Safeco Field (54% chance): 15-minute delay at I-405 & I-90. Suggest I-5 instead. Destination Seattle Center (31% chance): Broad St. closed. Suggest Denny Way instead. Going to the airport? Park with us for $8/day!
13
Greater Seattle, ~ 40 km X 40 km1 km grid
14
Ground Cover Prior U.S. Geological Survey – Seattle Area
15
All Possible DestinationsDestinations of One Subject
16
Day 1Day 2Day 3Day 4Day 5Day 6Day 7 Day 8Day 9Day 10Day 11Day 12Day 13Day 14 Personal destinations = visited cells + clustering + sparkling
17
start Current Location Candidate Destination R r ΔtΔt
18
From 2001 U.S. National Household Transportation Survey
19
Efficient driving likelihood: Trip time likelihood: Open-world prior: Final probability: Closed-world prior: Wedding cakes: Ground cover:
20
Half of trips (3667) for training efficiency distributions Remaining half for testing Leave-one-out for personal destinations prior
21
Destination Modeling Predestination – Destination prediction Snap-to-Road – Map matching with temporal constraints Personalized Routes Location Privacy
22
Congestion PricingLocation Based ServicesPay As You Drive (PAYD) Insurance Collaborative Traffic Probes (DASH)Research (London OpenStreetMap) John Krumm, "Inference Attacks on Location Tracks", Fifth International Conference on Pervasive Computing (Pervasive 2007), May 13-16, 2007, Toronto, Ontario, Canada.
23
Pseudonomized GPS tracks Infer home location Reverse white pages for identity
24
Last Destination – median of last destination before 3 a.m. Median error = 60.7 meters
25
Weighted Median – median of all points, weighted by time spent at point (no trip segmentation required) Median error = 66.6 meters
26
Largest Cluster – cluster points, take median of cluster with most points Median error = 66.6 meters
27
Best Time – location at time with maximum probability of being home Median error = 2390.2 meters (!)
28
GPS interval – 6 seconds and 63 meters GPS satellite acquisition – ≈45 seconds on cold start, time to drive 300 meters at 15 mph Covered parking – no GPS signal Distant parking – far from home Covered ParkingDistant Parking
29
Windows Live Search reverse white pages lookup (free API at http://dev.live.com/livesearch/)http://dev.live.com/livesearch/
30
GPS Tracks (172 people) Home Location (61 meters) Home Address (12%) Identity (5%) MapPoint Web Service reverse geocoding Windows Live Search reverse white pages
31
Originalσ= 50 meters noise added Effect of added noise on address- finding rate
32
OriginalSnap to 50 meter grid Effect of discretization on address-finding rate
33
1.Pick a random circle center within “r” meters of home 2.Delete all points in circle with radius “R”
34
© 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.