Trafikanalyser baseret på GPS data
Kristian Torp
Department of Computer Science Aalborg University, Denmark
torp@cs.aau.dk
Daisy Partners/Collaboration
2019: Trip Count, Denmark
2019: Trip Count, Copenhagen
Travel-Time: Limfjordstunnel
0 50 100 150 200 250 300
00:00 01:20 02:45 05:30 06:50 08:10 09:30 10:50 12:10 13:30 14:50 16:10 17:30 18:50 20:10 21:30 22:50
• 2019
• Workdays
• 5 minutes resolution
Langebro-Rådhuspladsen
0 50 100 150 200 250
00:00 01:20 02:40 05:25 06:45 08:05 09:25 10:45 12:05 13:25 14:45 16:05 17:25 18:45 20:05 21:25 22:45
• 2019/workdays
• 5 minutes resolution
Svendborg
0 10 20 30 40 50 60
00:00 00:40 01:35 02:30 05:45 06:20 06:55 07:30 08:05 08:40 09:15 09:50 10:25 11:00 11:35 12:10 12:45 13:20 13:55 14:30 15:05 15:40 16:15 16:50 17:25 18:00 18:35 19:10 19:45 20:20 20:55 21:30 22:05 22:40 23:15 23:50
Odense April vs. July
• Route on motorway
• Compare months
• Actual trips!
Point = one trip
50 100 150 200 250 300
00:00 02:24 04:48 07:12 09:36 12:00 14:24 16:48 19:12 21:36 00:00
Travel Time Apil (first week) Travel-Time July
Accident Data
Accident/GPS/Weather Data in DK
Accident Køgebugt Motorway
The Other Direction on Motorway
Sampling GPS data and estimating travel times..
How hard can it be?
1 Second GPS Data
1 second/sample
1 Second Trajectory
1 second/sample
5 Seconds GPS Data/Trajectory
5 second/sample
1 vs 5 Seconds – Computed Speeds
1 second/sample 5 second/sample
Km/h
1 vs 5 Seconds – Computed Speeds
1 second/sample 5 second/sample
Km/h
1 vs 5 Second – Computed Speeds
1 second/sample 5 second/sample
Km/h
120 Seconds GPS Data
120 second/sample
1 vs 120 Second - Wrong Trajectory
120 second/sample
Map-Matching Methods
• Point based map matching
Nearest neighbor road segment search
Pro: Still works with high sampling period
Con: No trajectory restoration
• Trajectory based map matching
Hidden Markov Map Matching Through Noise and Sparseness (Newson and Krumm, SIGSPATIAL ‘09)
Pro: Restores trajectory, even with “holes” in data
Con: Hard to reconstruct correct trajectories at high sampling periods
Using strict path queries techniques for determining reliability
Path-based Queries on Trajectory Data (Krogh et.al, SIGSPATIAL ’14)
Point Based Trajectory Based
• Point method coverage quickly drops
• Trajectory method worse using 120 second data
Road-Segment Coverage
Point Based Trajectory Based
• Point method up to 3% off using 5 sec data
• Trajectory method has problems with 120 sec data
Travel Time
Summary
• GPS data rich source of information about traffic
Remember GDPR can be very time consuming!
• Querying billions of rows is today done interactively!
• Integration of GPS data with other data sources
Spatial/temporal integration point
Weather, accidents, height, houses along each road, …
• Sampling period of GPS data is important for travel-time
Huge differences 1 versus 120 seconds data
Recommendation 15 seconds or lower
• New topics for Daisy Research Group
Real-time processing, fuel-consumption, EVs, bicycles, …
More advanced queries (including data-mining based)
“Trajectory weaving” for long routes