1. Clustering in Urban Environments: Virtual forces applied to vehicles
Leandros A. Maglaras (presentation)
Dimitrios Katsaros
University of Thessaly, Greece 1
IEEE ICC, Budapest (Hungary), 13th June 2013 1

2. Presentation outline Earlier work and background knowledge
Virtual Forces applied to vehicles - VFVC
Direction matters – Parameter Qd(t)
Lane detection
Performance evaluation
Conclusions
IEEE ICC, Budapest (Hungary), 13th June 2013 2

3. Clustering in Urban Envirnments
Characteristics
Intersections
Mobility patterns
More complex metrics
Methods
Affinity propagation APROVE
Density based clustering
Distributed group mobility adaptive clustering (DGMA)
Spring Clustering
IEEE ICC, Budapest (Hungary), 13th June 2013 3

4. Contributions
• Virtual Forces Vehicle Clustering (VFVC) method (incorporation of local geographical characteristics in clusterhead election)
• VFVC incorporates current position in clusterhead election (Lane detection)
• Evaluation of the performance of VFVC under different network characteristics (density, velocity, car dimensions).
IEEE ICC, Budapest (Hungary), 13th June 2013

5. Relative force among nodes l,m
The concept of ‘Force directed algorithms’
Relative forces applied to nodes
Forces are assigned as if the edges were springs and the nodes were electrically
charged particles.The entire graph is then simulated as if it were a physical system
Special role of nodes
Relative force among nodes l,m
Distance among nodes
IEEE ICC, Budapest (Hungary), 13th June 2013 5

6. VFVC–characteristics that affect clustering
• Vehicles that follow predefined routes like a bus (Qp)
• Tall vehicles like trucks (Q)
• Vehicles that follow non-turning lanes in a multi lane street (Qd(t)).
• Vehicles that their driver behavior is statistically smooth (Qb).
• Vehicles that mobility can be predicted (Qpr)
Initial charge – same for all vehicles
Total charge of vehicle (i) at time t
IEEE ICC, Budapest (Hungary), 13th June 2013

7. Direction matters Beacon messages
node Identifier (ID), node location, speed vector, total
Force, state, time stamp, vehicle height and lane
Clustering procedure
• If the car follows a non turning lane ► Qd(t)=2
• If the car follows a turning lane ► Qd(t)=1
• If lane is going straight or turns the ► Qd(t)=1.5
IEEE ICC, Budapest (Hungary), 13th June 2013 7

8. Lane detection Lane detection system and a digital street map
GPS + wheel odometer [A.Dean 2009]
Beacon network ► triangulate vehicle position
No GPS - real-time stereo vision [M. Bertozzi 1998]
LIDAR - Light Detection and Ranging [M.Jabbour 2006]
IEEE ICC, Budapest (Hungary), 13th June 2013 8

9. Evaluation setting (1/3)
Virtual Forces Vehicle Clustering (VFVC)
Impact of direction on a main street
Competitors: Sp-Cl IEEE VECON’12], Low-Id,
Mobic
Measured quantities
Average number of clusters (number of CHs)
Average cluster changes / vehicle
Average cluster lifetime
IEEE ICC, Budapest (Hungary), 13th June 2013 9 9

10. Aggregation of roads - intersections
Evaluation setting (2/3)
Urban area of Volos
Aggregation of roads - intersections
Simulated urban area of Volos
Route Distributions
IEEE ICC, Budapest (Hungary), 13th June 2013 10 10

11. Evaluation setting (3/3)
IEEE ICC, Budapest (Hungary), 13th June 2013

12. Impact of Lane detection in mean lifetime
Direction expressed in terms of the street lane occupied by the vehicle is used in order to select the more stable clusterhead
VFVC always performs better that the other methods
IEEE ICC, Budapest (Hungary), 13th June 2013 12 12

13. Total number of clusters
IEEE ICC, Budapest (Hungary), 13th June 2013 13

14. Average cluster changes / vehicle
IEEE ICC, Budapest (Hungary), 13th June 2013 14

15. Summary Mobility in Urban Environments
Lane detection - Direction matters
Virtual Forces Vehicle Clustering (VFVC)
Favors non turning vehicles to become clusterheads (due to staying longer on the road)
Increases cluster lifetime
Special role of vehicles is used to combine different characteristics - parameters of vehicles to improve the clustering performance
IEEE ICC, Budapest (Hungary), 13th June 2013 15 15

16. Thank you! Clustering in Urban Environments:
Virtual forces applied to vehicles
Thank you!
Leandros A. Maglaras (presentation)
Dimitrios Katsaros
University of Thessaly, Greece 16
IEEE ICC, Budapest (Hungary), 13th June 2013 16 16