GPS Calibrated Ad-hoc Localization for

Geosocial Networking

Dexter H. Hu

Cho-Li WangYinfeng Wang

{hyhu,clwang,yfwang}@cs.hku.hk

Outline

• Introduction– Mobile Twitter for Geosocial Networking

• Related Work– MCL and Amorphous

• MobiAmorph algorithm

• Performance Evaluation and Analysis

• Discussion

• Conclusion

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/212

Geosocial Networking• From social networking -> mobile social

networking -> geosocial networking– A new type of social networking in which 

geographic services and capabilities such as geocoding and geotagging are used to enable additional social dynamics.

• Application Example– Location-planning– Social Shopping– Trip tracking

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/213

23/4/214

Mobile Twitter1. Practical for real life usage and encourage ad-hoc

information sharing, – Mobile social applications will be more meaningful

and location-aware – Twit social events with location information attached.

• Car accident, Taxi call, Voting, Disaster/rescue

2. Localization is possible without the deployment of large infrastructure– Help of GPS-enabled mobile users

3. Under certain mobility model of pedestrians in typical urban environment, accurate GPS information can quickly propagate to non-GPS users

GPS Calibrated Ad-hoc Localization for Geosocial Networking

Usage Scenario and Components

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/215

Usage Scenario and Components (cont'd)

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/216

Localization with Historical Data and Moving Velocity

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/217

Figure 2

possible area

Related Work

• Range-free Localization– Monte Carlo Localization (MCL)

• Posterior distribution of a node’s possible locations using a set of weighted samples

– Amorphous• Similar variant DV-HOP, pop-counting technique

which is similar to distance vector routing. • Each seed broadcasts its location to neighbors and

other nodes try to estimate their distance to seeds

23/4/218 GPS Calibrated Ad-hoc Localization for Geosocial Networking

Outline

• Introduction– Mobile Twitter: Geosocial Networking

• Related Work– MCL and Amorphous

• MobiAmorph algorithm

• Performance Evaluation and Analysis

• Discussion

• Conclusion

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/219

Common Notations

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2110

MobiAmorph Algorithm

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2111

Receive enough fresh information

Localization with Historical Data and Moving Velocity

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2112

Figure 2

possible area

MobiAmorph Algorithm• Relaxed Trilateration:

– Multilateration of Amorphous needs at least 3 reference points. – Location estimating with overlapping circles can still have a

decent estimation even there are only two reference points available.

• Increased coverage.

• Historical Data – Last estimated location to increase accuracy and coverage.

• With relaxed trilateration, only one reference information is need

– Two hop count packet• Increased coverage and accuracy

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2113

Outline

• Introduction– Mobile Twitter: Geosocial Networking

• Related Work– MCL and Amorphous

• MobiAmorph algorithm

• Performance Evaluation and Analysis

• Discussion

• Conclusion

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2114

Performance Evaluation and Analysis

• MobiReal Simulator

• Evaluation Goals:1. Coverage and Accuracy of MobiAmorph with

MCL and Amorphous

2. MobiAmorph under various settings for recommended configuration in real deployment

3. Mobile Twitter’s power/memory consumption by MobiAmorph

GPS Calibrated Ad-hoc Localization for Geosocial Networking

Evaluation Scenarios

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2116

Open Area (100m x 100m)

Street Building(500m x 500m)

Effect of Packet Interval and Seed Ratio in Street and Open Area

Parameter Value

Node Speed (m/s) 1.5, 3, 5

Radio Range (m) 10

Seed Ratio 0.2, 0.3, 0.4, 0.5

Packet Interval 5, 15, 30, 60, 90

Density 30

GPS Calibrated Ad-hoc Localization for Geosocial Networking

Effect of Packet Interval and Seed Ratio in Street and Open Area (cont'd)

18

Street Open

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2119

OpenStreet

MobiAmorph Performance on Street Scenario

Parameter Value

Node Speed (m/s) 1.5, 3, 5

Radio Range (m) 10

Seed Ratio 0.2, 0.3, 0.4, 0.5

Packet Interval 5, 15, 30, 60, 90

Density 10, 20, 30, 40

GPS Calibrated Ad-hoc Localization for Geosocial Networking

MobiAmorph Performance on

Street Scenario (cont'd)

21

Mobile Twitter Deployment Evaluationon Android phone

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2122

Outline

• Introduction– Mobile Twitter: Geosocial Networking

• Related Work– MCL and Amorphous

• MobiAmorph algorithm

• Performance Evaluation and Analysis

• Discussion

• Conclusion

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2123

Discussion• Resolution Limitation

– Theoretical limitation for using only connectivity information

• Privacy and Security for Adoption– Malicious seeds– Corrupted relay nodes– Application Message encrypted

• Pedestrian Mobility Model– Urban Pedestrian Flows (UPF)

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2124

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2125

Conclusion• Ad hoc localization with the help of GPS

information in urban environment with pedestrians

• We compared MobiAmorph with other two distributed range-free localization algorithms.

• The Mobile Twitter application is developed with the MobiAmorph algorithm on the Android to boost adoption of geosocial networking.

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2126

Thank you! 謝謝！

GPS Calibrated Ad-hoc Localization for Geosocial Networking23/4/2127