last mile navigation using smartphonesasian monsoon winds arabs magnetic compass and kamal english...
TRANSCRIPT
321Yuanchao Shu Paris, France September 10, 2015
LAST MILE NAVIGATIONUSING SMARTPHONESYuanchao Shu, Kang G. Shin, Tian He and Jiming Chen
ACM MobiComParis, France
September 2015
322Yuanchao Shu Paris, France September 10, 2015
NAVIGATIONTHE ACT OF FINDING THE WAY TO GET TO A PLACE
323Yuanchao Shu Paris, France September 10, 2015
https://studioartofficial.wordpress.com/2014/08/19/vintage-tuesday-6/
324Yuanchao Shu Paris, France September 10, 2015
• Navigation
http://www.hipi.info/2014/07/7-cool-airplane-twitter-header-1500x500.html
325Yuanchao Shu Paris, France September 10, 2015
HISTORY OF NAVIGATION
PolynesianMotion of stars, waves
HISTORY OF SPECIAL INSTRUMENTS
Portuguese and SpanishMariner’s astrolabe and compassCircumnavigation and mapping
AsianMonsoon winds
ArabsMagnetic compass and Kamal
EnglishSextant and chronometer
AmericanSatellite navigation system
326Yuanchao Shu Paris, France September 10, 2015
MODERN NAVIGATION SYSTEMSWORKING PRINCIPLE
PositioningPositioning
Mapping
Positioning
Mapping
Path Planning
327Yuanchao Shu Paris, France September 10, 2015
MODERN NAVIGATION SYSTEMS
• Positioning
– Outdoor: satellite-based, meter-level positioning accuracy.
– Indoor: WiFi, geomagnetism, IMU, Bluetooth, FM [Youssef’05, Yoon’13, Xiong’13].
• Mapping
– We need a map.
• Satellite mapping, war-driving, floorplan mapping etc.
• Path Planning
– Extensively studied in robotics and mathematics.
STATE-OF-THE-ART
Does this suffice?
328Yuanchao Shu Paris, France September 10, 2015
LACK OF MAP INFORMATIONBOTTLENECK OF NAVIGATION SYSTEMS
CoverageGlobal, rural, indoor
FidelityProvisional paths
ResolutionTrails, parking lots
Maps ?Positioning Path Planning Navigation
329Yuanchao Shu Paris, France September 10, 2015
LACK OF MAP INFORMATIONBOTTLENECK OF NAVIGATION SYSTEMS
CoverageGlobal, rural, indoor
FidelityProvisional paths
ResolutionTrails, parking lots
Maps ?Positioning Path Planning Navigation
3210Yuanchao Shu Paris, France September 10, 2015
LAST MILE NAVIGATION PROBLEMNavigates one to the vicinity of destination tens of miles away,
but fails to find a feasible path from there to final destination
3211Yuanchao Shu Paris, France September 10, 2015
• Last mile navigation
• Plug-and-play
• Lightweight
• Smartphone-based
FollowMe
3212Yuanchao Shu Paris, France September 10, 2015
BASIC IDEAOF FOLLOWME
• Exploits “scents/crumbs” left behind by the previous travelers.
Ice Age: Scrat's Continental Crack-Up
3213Yuanchao Shu Paris, France September 10, 2015
BASIC IDEA
(leader’s)
Trace-collection phase
(follower’s) Navigation phase
OF FOLLOWME
(leader’s)
Trace-collection phase
(follower’s) Navigation phase
(leader’s)
Trace-collection phase
(follower’s) Navigation phase
3214Yuanchao Shu Paris, France September 10, 2015
USE CASESOF FOLLOWME
3215Yuanchao Shu Paris, France September 10, 2015
DESIGNTrace Collection & Real-time Navigation
3216Yuanchao Shu Paris, France September 10, 2015
ARCHITECTUREOF FOLLOWME
Trace Collection Module
3217Yuanchao Shu Paris, France September 10, 2015
TECHNICAL DESIGN
Geo-magnetic Accelerometer Gyroscope Barometer Steps Turns Level changes
REFERENCE TRACE CONSTRUCTIONSensory data Detection results
Time
Sensory data
Steps
Turns
Level changes
Reference trace
Timestamps
3218Yuanchao Shu Paris, France September 10, 2015
ARCHITECTUREOF FOLLOWME
Navigation Module
Trace Collection Module
3219Yuanchao Shu Paris, France September 10, 2015
ARCHITECTUREOF FOLLOWME
Navigation Module
Trace Collection Module
3220Yuanchao Shu Paris, France September 10, 2015
0 20 40 60 80 100 120
10
20
30
40
50
60
70
Time (s)
Magnitude (T
)
Trace 1
Trace 2
0 20 40 60 80 100 120
10
20
30
40
50
60
70
Time (s)
Magnitude (T
)
Trace 1
Trace 2
0 20 40 60 80 100 120
10
20
30
40
50
60
70
Time (s)
Magnitude (T
)
Trace 1
Trace 2
0 20 40 60 80 100 120
10
20
30
40
50
60
70
Time (s)
Magnitude (T
)
Trace 1
Trace 2
0 20 40 60 80 100 120
10
20
30
40
50
60
70
Time (s)
Magnitude (T
)
Trace 1
Trace 2
0 20 40 60 80 100 120
10
20
30
40
50
60
70
Time (s)
Magnitude (T
)
Trace 1
A NAVIGATION EXAMPLE
A
B
C
D
Ea
bc
d
e
3221Yuanchao Shu Paris, France September 10, 2015
TECHNICAL DESIGN
• Step-constrained trace synchronization algorithm
– Filter out high-freq. mag. and utilize differential info. to handle device and usage diversity
– Sync. based on legacy dynamic time warping (DTW)
WALKING PROGRESS ESTIMATION
3222Yuanchao Shu Paris, France September 10, 2015
TECHNICAL DESIGNWALKING PROGRESS ESTIMATION
• Step-constrained trace synchronization algorithm
– Filter out high-freq. mag. and utilize differential info. to handle device and usage diversity
– Sync. based on legacy dynamic time warping (DTW)
• Full knowledge of traces
• Quadratic computational complexity
– Online DTW with linear computation overhead
𝐷 𝑖 𝑗 = min(𝐷 𝑖 − 1 𝑗 − 1 ,𝐷 𝑖 − 1 𝑗 , 𝐷[𝑖][𝑗 − 1]) + 𝑑𝑖𝑠𝑡(𝑖, 𝑗)
(1,1)
User trace
Reference trace
3223Yuanchao Shu Paris, France September 10, 2015
TECHNICAL DESIGN
• Step-constrained trace synchronization algorithm
– Filter out high-freq. mag. and utilize differential info. to handle device and usage diversity
– Sync. based on legacy dynamic time warping (DTW)
• Full knowledge of traces
• Quadratic computational complexity
– Online DTW with linear computation overhead
• Step-constrained search space
• Dynamically changing search band
WALKING PROGRESS ESTIMATION
1 23 4
5 6 7
Step #
1 2 3 4 5 6 7
3224Yuanchao Shu Paris, France September 10, 2015
• Step-constrained trace synchronization algorithm
– Filter out high-freq. mag. and utilize differential info. to handle device and usage diversity
– Sync. based on legacy dynamic time warping (DTW)
• Full knowledge of traces
• Quadratic computational complexity
– Online DTW with linear computation overhead
• Step-constrained search space
• Dynamically changing search band
TECHNICAL DESIGNWALKING PROGRESS ESTIMATION
12
34
56
7
1 23 4
5 6 7
3225Yuanchao Shu Paris, France September 10, 2015
3226Yuanchao Shu Paris, France September 10, 2015
IMPLEMENTATION AND EVALUATION• Implementation
– Android 4.4.2, Samsung Galaxy S5
– Two threads
• Data collection : 50Hz
• Signal processing
– DTW buffer size: 12s (c = 600)
• Evaluation
– Four-story campus building
– 5 participants
– 10 different reference traces
3227Yuanchao Shu Paris, France September 10, 2015
EVALUATIONNAVIGATION ACCURACY
0.0
0.2
0.4
0.6
0.8
1.0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
FollowMe
Geomagnetism, with PF
WiFi, with PF
CDF of spatial error in navigation
Spatial error (m)
3228Yuanchao Shu Paris, France September 10, 2015
EVALUATIONNAVIGATION ACCURACY
0 s
2 s
4 s
6 s
8 s
10 s
User A User B User C User D
Checkpoint A Checkpoint B
Checkpoint C Checkpoint D
Lead time of navigation instructions at different checkpoints
3229Yuanchao Shu Paris, France September 10, 2015
RELATED WORK
RoboticsSpecial hardware-based nav.
[Cho’10, Bonin’08]
Complicated humans’ locomotion;Limited energy buffer of smartphones.
Geo-magneticAnomalies-based local. and nav.
[Glanzer’10, Gozik’11, Chung’11, Grand’12],
Ubiquitous and stable;Localization-based navigation (map?);
Tedious fingerprint collection.
SmartphoneNav. with or w/o infrastructure[Li’12, Chintalapudi,’10, Rai’12, Xiong’’13,
Yang’12, Chen’12]
Accumulative error and usage-dependent;Non-universal (e.g., GPS, WiFi);High bootstrap effort of fingerprinting.
Leader-followerTrace-based nav.[Constandache’10, Riehle’12, Zheng’14]
Customized devices (e.g., robots);Infrastructure-dependent (e.g., WiFi, beacons);Constraints imposed on users.
3230Yuanchao Shu Paris, France September 10, 2015
CONCLUDING REMARKS
INFRASTRUCTURE FREECloud-based or Ad-hoc
No need of floor plans (maps)
WiFi/Bluetooth-independent
HIGH EFFICIENCYLow-power sensors
Low computation
Energy efficient
MINI. USER INVOLVEMENTPlug-and-play
Fast and easy bootstrapping
No action required during NAV
3231Yuanchao Shu Paris, France September 10, 2015
3232Yuanchao Shu Paris, France September 10, 2015
More info. and updates
FollowMe
3233Yuanchao Shu Paris, France September 10, 2015
THANK YOUQ&A