Visual Traffic Jam Analysis Based on Trajectory Data Zuchao Wang, Min Lu, Xiaoru Yuan, Peking University Junping Zhang, Fudan University Huub van de Wetering, Technische Universiteit Eindhoven Introduction Many cities are suffering from traffic jams Beijing Melbourne Atlanta 2/47 1
Introduction Current traffic jam monitoring technologies Real time road condition from Google Map 3/47 Introduction Complexities of traffic jams Road conditions change with time Different roads have different congestion patterns Congestions propagate along the road network 4/47 2
Introduction Complexities of traffic jams Road conditions change with time Different roads have different congestion patterns Congestions propagate along the road network 5/47 Introduction Complexities of traffic jams Road conditions change with time Different roads have different congestion patterns Congestions propagate along the road network 6/47 3
Introduction Complexities of traffic jams Road conditions change with time Different roads have different congestion patterns Congestions propagate along the road network A visual analytic system to help people study these complexities 7/47 Related Work Traffic Modeling Outlier tree [Liu et al. 2011] Study the propagation of traffic outlier events between regions Probabilistic Graph Model [Piatkowski et al. 2012] powerful but complex 8/47 4
Related Work Traffic Event Visualization [Andrienko et al. 2011] Individual events extraction Incident Cluster Explorer [Pack et al. 2011] Individual events filtering 9/47 Data Description Beijing taxi GPS (from DataTang.com) Size: 34.5 GB #Taxi: 28,519 (7% of total traffic flow volume) #Sampling point: 379,107,927 Time range: 24 days (Mar.2 nd ~25 th, 2009, missing 18 th ) Sampling interval: 30 seconds (60% missing) 10/47 5
Data Description Beijing taxi GPS (from DataTang.com) Size: 34.5 GB #Taxi: 28,519 (7% of total traffic flow volume) #Sampling point: 379,107,927 Time range: 24 days (Mar.2 nd ~25 th, 2009, missing 18 th ) Sampling interval: 30 seconds (60% missing) Beijing road network (from OpenStreetMap) Size: 40.9 MB 169,171 nodes and 35,422 ways 11/47 Overview GPS Trajectories Preprocessing (based on a traffic jam model) Traffic Jams Visual exploration (based on a visual interface) Traffic Jam Knowledge 12/47 6
Overview GPS Trajectories Preprocessing (based on a traffic jam model) Traffic Jams Visual exploration (based on a visual interface) Traffic Jam Knowledge 13/47 Design Requirements Traffic jam model Visual interface Complete Include location, time, speed Structured Study propagations of traffic jams Road Bound Traffic jams happen on roads Informative Show location, time, speed, propagation path Multilevel Explore from city level to road segment level Filter Enabled Select and study a subset of propagations 14/47 7
Overview GPS Trajectories Preprocessing (based on a traffic jam model) Traffic Jams Visual exploration (based on a visual interface) Traffic Jam Knowledge 15/47 Overview GPS Trajectories Preprocessing (based on a traffic jam model) Traffic Jams Visual exploration (based on a visual interface) Traffic Jam Knowledge 16/47 8
Preprocessing Input Raw Taxi Raw Road Road Speed Data Traffic jam Traffic Jam Event Data Traffic Jam Propagation Graphs 17/47 Preprocessing Input Raw Taxi Raw Road 18/47 9
Preprocessing Input Cleaning Raw Taxi Cleaned Raw Road Road Cleaning Cleaned Road 19/47 Preprocessing Input Raw Taxi Raw Road Cleaned Cleaned Road Map Matching GPS Trajectories Matched to the Road 20/47 10
Preprocessing Input Raw Taxi Raw Road Road speed: for each road at each time bin (10 min) Cleaned Cleaned Road a b GPS Trajectories Matched to the Road Road Speed Calculation Road Speed Data 9:10 am 50 km/h 9:20 am 45 km/h 9:30 am 12 km/h 9:40 am 15 km/h 9:10 am 55 km/h 9:20 am 10 km/h 9:30 am 12 km/h 9:40 am 45 km/h 21/47 Preprocessing Input Raw Taxi Raw Road Traffic jam events: road, start/end time bin Cleaned Cleaned Road a b GPS Trajectories Matched to the Road Road Speed Data Traffic Jam Detection Traffic Jam Event Data 9:10 am 50 km/h 9:10 am 55 km/h 9:20 am 45 km/h e 2 e 1 9:30 am 12 km/h 9:20 am 10 km/h 9:30 am 12 km/h 9:40 am 15 km/h 9:40 am 45 km/h 22/47 11
Preprocessing Input Raw Taxi Raw Road Defining propagation based on spatial/temporal relationship: Cleaned Cleaned Road e 2 a e 1 b e 2 happens after e 1, and on a road following e 1 GPS Trajectories Matched to the Road Road Speed Data Traffic Jam Event Data Propagation Graph Construction Traffic Jam Propagation Graphs 9:10 am 50 km/h 9:20 am 45 km/h e 2 e 1 9:30 am 12 km/h 9:40 am 15 km/h 23/47 9:10 am 55 km/h 9:20 am 10 km/h 9:30 am 12 km/h 9:40 am 45 km/h Preprocessing Input Raw Taxi Raw Road A real propagation graph: Start points Cleaned Cleaned Road GPS Trajectories Matched to the Road Road Speed Data Traffic Jam Event Data Propagation Graph Construction Traffic Jam Propagation Graphs End points 24/47 12
Preprocessing Input Raw Taxi Raw Road Road Speed Data Traffic jam Traffic Jam Event Data Traffic Jam Propagation Graphs 25/47 Visual Interface Three levels of exploration Road Segment Level Exploration and Analysis Road of Interest est One Propagation Graph Propagation Graph Level Exploration Propagation Graph List Traffic jam Road Speed Data Traffic Jam Event Data Propagation Graphs of Interest Spatial Density Time and Size Distribution Topological Clustering Custe Traffic Jam Propagation Graphs Spatial Filter Temporal & Size Filter Topological Filter Dynamic Query 26/47 13
Visual Interface: City Level Propagation Graph List Road Speed Data Propagation Graphs of Interest Traffic jam Traffic Jam Event Data Traffic Jam Propagation Graphs 27/47 Visual Interface: City Level Graph list view: show propagation graphs as icons Time range Spatial path: color for congestion time on each road Size: #events, duration, distance 28/47 14
Visual Interface: City Level Propagation Graph List Traffic jam Road Speed Data Traffic Jam Event Data Propagation Graphs of Interest Spatial Density Time and Size Distribution Topological Clustering Custe Traffic Jam Propagation Graphs Spatial Filter Temporal & Size Filter Topological Filter Dynamic Query 29/47 Visual Interface: City Level Graph projection view: topological aggregation All propagation graphs are grouped Each group is represented by a point Graphs in the same group have similar topologies Distance between points represent the topological distance between graph groups 30/47 15
Visual Interface: Single Propagation Level One Propagation Graph Propagation Graph Level Exploration Road Speed Data Propagation Graphs of Interest Traffic jam Traffic Jam Event Data Traffic Jam Propagation Graphs 31/47 Visual Interface: Single Road Level Road Segment Level Exploration and Analysis Road of Interest est One Propagation Graph Road Speed Data Propagation Graphs of Interest Traffic jam Traffic Jam Event Data Traffic Jam Propagation Graphs 32/47 16
Visual Interface: Single Road Level Table-like pixel-based visualization Each cell represents 10 min of a day Each column represents 10 min Each row represents one day Color encodes speed 33/47 Visual Interface: Single Road Level Table-like pixel-based visualization Make non-jam cells smaller to highlight traffic jams 34/47 17
Case Studies Road level exploration Propagation graph exploration Propagation trend exploration 35/47 Case Study: Road Level Exploration a b a b c 7:30~10:00 13:30~18:30 18:30 7:30~8:00 d c d 730 7:30~9:00 900 16:30~18:30 36/47 18
Case Study: Road Level Exploration e e 8:00~17:30 03/04~07 03/19~22 f f 03/11,15:40~16:50 g g 0:00~4:30 21:30~24:00 37/47 Case Study: Propagation Graph Exploration Filter propagation graphs Observe propagation path Check road speed Check propagation time Watch animation 38/47 19
Case Study: Propagation Graph Exploration Filter propagation graphs Observe propagation path Check road speed Check propagation time Watch animation Major end points Major start point 39/47 Case Study: Propagation Graph Exploration Filter propagation graphs Observe propagation path Check road speed Check propagation time Watch animation 40/47 20
Case Study: Propagation Graph Exploration Filter propagation graphs Observe propagation path Check road speed Check propagation time Watch animation E 18:50~19:00 19 00 18:10~19:10 D C 14:20~15:00 17:40~19:30 13:50~19:50 13:30~20:00 B A 41/47 Case Study: Propagation Graph Exploration Filter propagation graphs Observe propagation path Check road speed Check propagation time Watch animation 42/47 21
Case Study: Propagation Trend Exploration One fixed region Different mornings 43/47 Showing to the public International Information Design Exhibition 09/26/2013~10/13/2013 Simplified version Only road level exploration 44/47 22
Conclusion A visual analytic system to study traffic jams An automatic process to extract traffic jams from GPS trajectories A visual interface to support multilevel exploration of traffic jams GPS Trajectories preprocess Traffic Jams exploration Traffic Jam Knowledge 45/47 Future Work Improve the traffic jam model (e.g. add prediction functions) Support more analysis task (e.g. spatial/temporal clustering) Try better visual design of propagation graphs Collaboration with the domain experts 46/47 23
Acknowledgement Funding: NSFC Project No. 61170204 NSFC Key Project No. 61232012 Data: Datatang.com OpenStreetMap Anonymous reviewers for valuable comments Our website: http://vis.pku.edu.cn/trajectoryvispku 47/47 24