View
367
Download
2
Embed Size (px)
DESCRIPTION
Oplægget blev holdt ved InfinIT-arrangementet Big Data og data-intensive systemer i Danmark, der blev af holdt en 15. januar 2014. Læs mere om arrangementet her: http://infinit.dk/dk/arrangementer/tidligere_arrangementer/big_data_i_danmark.htm
Citation preview
GPS FOR INTERACTIVE TRAFFIC
ANALYSIS
Kristian Torp
Department of Computer Science
Aalborg University, Denmark
Outline
• Technical part
Data foundation
Software architecture
Data-warehouse design
• Traffic-analysis part
Congestion and speed maps
Eco routing (fastest versus most fuel-efficient route)
Path analysis (routes taken between points A and B)
Trip analysis (where from and where to)
• Questions and answers
InfinIT 2014-01-15 2
Overall Idea
ITS
Software
Digital
Map
GPS Data
POI
Travel-time
Map
Congestion Map
Input Output
Path Analysis
Eco Routing CANBus
Data
…
InfinIT 2014-01-15 3
Trip Analysis
…
GPS and CANBus Data
id Veh id Lat Long Time Velocity
1 42 57.0 9.9 12:00:00 80
2 42 57.0 9.9 12:00:01 80
3 42 57.0 9.9 12:00:02 70
4 42 57.0 9.9 12:00:03 75
5 42 57.0 9.9 12:00:04 80
InfinIT 2014-01-15 4
id Veh. id Fuel Breaks Temp. RPM GPS data
1 42 123 0 90 1400 …
2 42 120 0 91 1424 …
3 42 90 40 90 850 …
4 42 12 80 90 850 …
5 42 239 9 90 2299 …
Statistics
• ~2.5 billion GPS observations
+80 % of road network
~1.8 million GPS observations daily
• ~250 million CANBus/OBD/FMS observations
12 % of road network
~0.5 million CANBus observations daily
• +14.000 vehicles in total
~3.000 vehicles daily (roughly half in weekends)
• Data warehouse updated nightly
Used for payment of taxi drivers
InfinIT 2014-01-15 5
• Not really “big data”
Compensate by small hardware budget
Big enough!
Software Architecture
64 bit Operating System
PostgreSQL 9.3/ Postgis 2.1
M-GEMMA Psycopg
Project-generic Code
Python 3.3
pygrametl
Project-specific Code
NetworkX
InfinIT 2014-01-15 6
Data Warehouse Design
Positioning
Data
Vehicle Data
Source
Batch
Load
Date Time Attributes
InfinIT 2014-01-15 7
Overall ETL Data Flow
GPS Data insert trip
Phase 1 Phase 2 Phase 3
point point-fact
trip-fact Map-match
data-fact
InfinIT 2014-01-15 8
Speed and Congestion Maps
• Maps covers all of Denmark
InfinIT 2014-01-15 9
Congestion
Køer i Jylland?
InfinIT 2014-01-15 10
Køer i Jylland?
InfinIT 2014-01-15 11
Randers Vej, Aarhus Limfjordstunnel
Eco-routing: Airport - AAU CPH
InfinIT 2014-01-15
Fuel 13:57 7.44 km. 0.74 l.
Fastest 10:51 9.11 km. 0.83 l.
12
Eco-routing: Airport – Railway Station
InfinIT 2014-01-15
Best 17:08 8.06 km. 0.80 l.
Worse 18:06 13.64 km. 1.23 l.
13
Køge Bugt – Ring 3: 15:00 – 18:00
InfinIT 2014-01-15 14
12 trips
761 secs
18.4 km
125 trips
655 secs
16.7 km
122 trips
669 secs
17.3 km
• 3 different routes
Køge Bugt – Ring 3: 7:00 – 9:00
InfinIT 2014-01-15 15
96 trips
815 secs
16.7 km
18 trips
709 secs
17.3 km
• 11 different routes
JUG?
Helsingørmotorvejen 7:00 – 9:00?
InfinIT 2014-01-15 16
• 855 trips
37 %
37 %
20 %
Køge Bugt: Where to 7:00 – 9:00?
InfinIT 2014-01-15 17
• 893 trips
Køge Bugt: Where from 7:00 – 9:00?
InfinIT 2014-01-15 18
• 893 trips
Summary
• Traffic is a great application for spatio-temporal data
Daisy’s “core business” for many years
• Many applications of GPS data
Congestion and speed maps
Route-choice/path analysis
Eco-routing
Intersection analysis using trips
• Fuel consumption from CANBus data
Interest for fleet owners
Green mobility
• 1NF is not dead
SQL is a pretty good query language
Parallel data processing a necessity
InfinIT 2014-01-15 19
1NF+SQL
Additional Information
InfinIT 2014-01-15
• http://daisy.aau.dk/its
• http://daisy.aau.dk/its/spqdemo
• Benjamin out-front doing demos
20