real time performance of tube and traffic in london
DESCRIPTION
Presented at:Possibilities of Real Time Data WorkshopLondon Data Store City Hall 19 Apr 2010 www.placr.co.ukTRANSCRIPT
Real time performance of tube and traffic in London
Dr David Mountain Placr Ltd
Possibilities of Real Time DataLondon Data StoreCity Hall19 Apr 2010
Travel network sensors
• Real-time travel: requires sensors embedded in the network, including:– Vehicles with GPS;– Dedicated sensors on tube /rail networks.
• Increasingly this information is streamed live:– within organisations for fleet mgmt;– available to public via websites.
• How can we use this to improve the traveller experience?
Network health
• Caching live feeds allows past performance of networks to be analysed.
• Identifies which parts of network are problematic at what times: the where and when of things to avoid.
• Temporal analysis allows travel and wait time estimates based on previous behaviour: manages expectations.
• High spatial resolution:– individual stations;– road sections.
• High temporal resolution:– by hour of day.
Tube departure board archiving
• Scraping tube departure data from TFL board– 180,000 observations per day
• Allows analysis of past behaviour• Predict future behaviour (expected)• Comparing current with expected behaviour
highlights incidents• Following analysis based on 40mn observations
Tube departure boards
Tubes departures: mins between trains vs time of day (avg all lines)
Tubes departures: mins between trains vs time of day (avg by line)
Spatial variation in tube frequency
Spatial variation in tube frequency
Spatial variation in tube frequency
Spatial variation in tube frequency
Tube freq (mins)
Spatio-temporal variation
0600-0700
Tube Freq (mins)
0700-0800
Tube Freq (mins)
0800-0900
Tube Freq (mins)
0900-1000
Tube Freq (mins)
1000-1100
Tube Freq (mins)
1100-1200
Tube Freq (mins)
1200-1300
Tube Freq (mins)
1300-1400
Tube Freq (mins)
1400-1500
Tube Freq (mins)
1500-1600
Tube Freq (mins)
1600-1700
Tube Freq (mins)
1700-1800
Tube Freq (mins)
1800-1900
Tube Freq (mins)
1900-2000
Tube Freq (mins)
2000-2100
Tube Freq (mins)
2100-2200
Tube Freq (mins)
2200-2300
Tube Freq (mins)
2300-0000
Tube Freq (mins)
Recent vs expected for Farringdon
Real-time dashboard of network health
Real-time dashboard of network health
Deviation from expected frequency (mins)
Problems on Circle line
Deviation from expected frequency (mins)
Dashboard over 24hrs
0600-0700
0700-0800
0800-0900
0900-1000
1000-1100
1100-1200
1200-1300
1300-1400
1400-1500
1500-1600
1600-1700
1700-1800
1800-1900
1900-2000
2000-2100
2100-2200
2200-2300
2300-0000
Traffic analysis
• For road network data, 500mn GPS data points from eCourier vehicles (vans, motorbikes, pushbikes)
• Validated to include only purposeful journeys• Cross referenced with road data• Provides unique speed estimates for each road
section, for combinations of:– Mode of transport– Time of day
eCourier floating car data
Vans: 0000-0700
Vans: 0700-1000
Vans: 1000-1500
Vans: 1500-1900
Vans: 1900-0000
Temporally sensitive routing
Marble Arch to Gray’s Inn FieldsAlternative routes for 4-wheel vehicles based on time of day
The morning rush: tubes vs roads
