myliu

Transcript

1. Generating Driving Trajectories with Mobile Subscription Data for Mobile Network Simulations Christoph Kieslich, Gerald Ostermayer
2. Page 2 •  A novel approach to generate driving trajectories for mobile subscribers •  Estimating cell site coverage •  Mapping mobile users on the road network Introduction Fußzeile
3. Page 3 Fußzeile Motivation Mobile network simulators rely on moving subscribers in order to simulate a network. But where do you get these information? Statistical Stochastic Real world O-D Matrix Random walk Manhattan grid Mobile subscription data
4. Page 4 Mobile subscription data •  Events that are carried out it in the core network of the network operator •  Events " Call establish / terminate " Hand over " Location area update " paging Fußzeile Time ID Cell LAC Event Latitude Longitude 1328972576 1328972972 1328972975 1328972996 1328973009 121 121 121 121 121 21405 38505 28505 18505 37455 21 21 21 21 21 34 32 32 32 32 Poslat1 Poslat2 Poslat3 Poslat4 Poslat4 Poslong1 Poslong2 Poslong3 Poslong4 Poslong4 Valerio, D.; D'Alconzo, A.; Ricciato, F. & Wiedermann, W. Exploiting Cellular Networks for Road Traffic Estimation: A Survey and a Research Roadmap Vehicular Technology Conference, 2009. VTC Spring 2009. IEEE 69th, 2009, 1-5
5. Page 5 Pipeline Fußzeile Generate Voronoi diagram Filter Call / Handover Estimate start / endpoint Calculate / validate route Timing
6. Page 6 Pipeline | Voronoi diagram •  An estimation of the coverage for each cell site is needed •  Voronoi partitioning has been used •  Fast but not accurate approximation •  No propagation model Fußzeile 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.4 0.5 0.6 0.7 0.8 0.9 1 Aurenhammer, F. Voronoi diagrams - a survey of a fundamental geometric data structure ACM Comput. Surv., ACM, 1991, 23, 345-405 Baert, A.-E. & Seme, D. Voronoi mobile cellular networks: topological properties Parallel and Distributed Computing, 2004, 29-35
7. Page 7 Start End Pipeline | Filter Events •  Only connected user (active call) send hand over events •  Hand over events have the highest geographic accuracy in GSM •  A trajectory is initiated by a call Fußzeile
8. Page 8 Pipeline | Estimate position •  Only serving cell is know •  No information about the position within the cell •  Our approach is based on population density data " People will more likely start, end their travel in high density areas Fußzeile 0 100 200 300 400 0 200 400 x in meter y in meter Voronoi cell Population 240 Population 45 Population 30
9. Page 9 Pipeline | Routing •  Routing between start and end position •  OpenStreetMap road network •  Validation " Timing " Euclidean distance " GPS and Hausdorff distance (only fo •  Rerouting when validation failed " Different start and end position Fußzeile
10. Page 10 Pipeline | Timing •  Each event contains a timestamp •  Timestamp can be used to derive velocity •  Hand over points will be annotated with timestamps •  Interpolation is done between hand over points •  Timing works better for rural than urban areas Fußzeile
11. Page 11 Results •  We calculated trajectories for subscribers in rural and urban areas •  Trajectories have been validate with GPS tracks •  Timing problems were discovered when hand over sequence was week " Voronoi partitioning " Network load Fußzeile
12. Page 12 Conclusion •  Trajectories can be estimated for mobile subscribers •  They are very precise enough for mobile network simulations •  Cover the whole network •  Adapt to seasonal changes •  Timing information can be added •  A better representation of the coverage area is needed Fußzeile
13. Page 13 ANY QUESTIONS? Fußzeile