Fujiyama workshop presentation
- 1. Data Analytics for Metro Performance Management Taku Fujiyama Associate Professor University College London Railway Research Group Department of Civil, Environmental and Geomatic Engineering
- 2. About myself • 5.5 years of work experience at East Japan Railway Company as a railway civil engineer, before joining UCL • Designed stations, including Shinjyuku, Ikebukuro, etc. • Main research topics • Passenger/pedestrian movements within railway systems • Railway traffic management 2
- 3. UCL and Railways • Richard Trevithick (1771 – 1833) • Mining Engineer, Pioneer of high- pressure steam • First person to invent steam locomotives • Developed first passenger-carrying train (‘Catch me who can’) in 1808 3 At the location of UCL Civil Engineering Building
- 4. OVERVIEW 4
- 5. Data and Performance evaluation • Data can be an enabler • We could extra valuable information that • Generally improve our understanding • Identify problems and issues • Etc • However, it is not all simple • There is no perfect data • It could find/address some issues, but usually it does not go further. 5 unless you have frameworks to systematically exploit it
- 6. Framework? Example • Railway tunnel • Traditionally, visual inspection by people (cracks, water, sound, etc) • New measurement systems: Strain gauge, video images, fibre optic cables, etc • A big question. How to see the data? 6 You need to have Right evaluation frameworks (or models) that are built on • Understanding of the nature of the problem of your interest • Right data collection and processing methods • Right evaluation criteria
- 7. Framework 7 Collection Processing Materialising Action in practice Knowledge, service, product • A framework directs Materialisation and Action • It also suggests how data be collected and processed In other words, this creates a value chain, where better outcomes of each element leads to better final outcomes
- 8. Different types of Framework 8 Collection Processing Materialising Action in practice Example 1 • Use an existing framework. Use big data for precise and better information • Use smartcard data for departure time distribution Example 2 • Exploiting (existing) data for your different question • e.g. smart card data for dynamic train control Value creation by data
- 9. Data usability for value creation 9 AFC Tikcet Barrier Train load Footfall counter CCTV Wi-fi GPS + app Web-use Shopping Coverage • • • • • • • • • • • • • Accuracy • • • • • • • • • • • Exploitability • • • • • • • • • • • • • • • Availability (Privacy sensitivity) • • • • • • Technical maturity • • • • • • • • • • • • • • Observation MobileRail-related User-activity • There is no perfect dataset that directly answers any important question • Current trend is to combine different datasets
- 10. Case Study 1: Project Evaluation • Framework: Generalised Journey Time • Based on much theoretical and implementation work in academia and at London Underground • Assumption • Journey time is the factor that determines passenger choice/behaviour • Quality of time is not equal between different actions • Sitting on a seat on train 15 minutes = standing 10 minutes • Big data is essential to calibrate the models • (UCL/London Underground can offer this as a course) 10
- 11. Case Study 2: Journey Planner • Transport for London makes journey planning data available to businesses • Some companies create new business opportunities by providing private journey planners with additional functions 11
- 12. UNDERSTANDING PASSENGER CHOICE OF CAR 12
- 13. • Passenger choice of boarding cars affects station platform congestion • Aim of the project: to understand how passengers choose the cars to board e.g. which car do people choose to board? Background 13
- 14. Why could the outcome be useful? • Uneven crowding across the cars of the whole train means ‘capacity not fully utilised’. • If there were factors that 1) influence passenger choice and 2) are controllable by metro operators, then by changing the factors we can improve capacity utilisation. 14 Our ambition is that controlling occupancy of each car (not each train) becomes a new norm for metro operations
- 15. Loadweigh data 15 § London Underground S stock (2010-) can record the weight of each car after the departure at each station § The weight can be converted to the number of passengers § 7-car trains (for H&C, Circle lines)
- 16. Hypotheses for passenger choice of boarding cars • Main factors 1. Some passengers choose cars close to the exits of their destination (Destination-based choice) 2. Other passengers choose the car next to the entrance (Origin-based choice) • Other potential factors • Seat availability, crowding on train, service headway, etc 16
- 17. Example: 17 § Hammersmith (HMS) and Goldhawk Road (GHR) End of Hammersmith&City, and Circle Lines Data: July&August, 2016 Weekday AM Peak (7-10am)
- 18. • % of passengers by destinations (at Hammersmith) based on an OD survey Destination-based Choice (1) 18 4.0% 3.8% 9.6% 2.9% 7.1% 3.2% 4.6% 14.3% 4.0% 7.2% 5.6% 4.7% 5.7% 6.4% 3.7% 8.0% 4.1% 0.8% 0.4% 0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% 16.0% Hammersmith GoldhawkRoad Shepherd'sBush… WoodLane LatimerRoad LadbrokeGrove WestbournePark RoyalOak Paddington EdgwareRoad BakerStreet reatPortlandStreet EustonSquare King'sCrossSt.… Farringdon Barbican Moorgate LiverpoolStreet AldgateEast Whitechapel StepneyGreen MileEnd BowRoad Bromley-by-Bow WestHam Plaistow UptonPark EastHam Barking
- 19. • Location of exits at destinations Destination-based Choice (2) 19 Car 7 Car 6 Car 5 Car 4 Car 3 Car 2 Car 1 Shepherds Bush Market Car 7 Car 6 Car 5 Car 4 Car 3 Car 2 Car 1 Wood Lane ··· • Location of the exit is next to Car 6 • ‘Destination-based choice’ passengers travelling to this station are assigned to Car 6
- 20. • Location of entrance at the origin Origin-based Choice 20 Car 7 Car 6 Car 5 Car 4 Car 3 Car 2 Car 1Hammersmith • Location of the entrance is next to Car 7 • Assumption that the further the car, the fewer people would takeAllocation of ‘origin-based choice’ passengers
- 21. This presentation 21 • Assuming that • trains are not crowded (have empty seats), • α % of people choose their boarding cars based on their Destination (Destination-based choice) • β % of people choose their boarding cars based on their Origin (Origin-based choice) • α + β = 100% • What would be α (as β=100- α) ? Finding α that suits the real Loadweigh data best
- 22. Result 1: Hammersmith (HMS) 22 α%=55.5% Average GEH=0.29
- 23. Result 2: Goldhawk Road (GHR) 23 α%=55.4% Average GEH=0.17
- 24. Summary of this research • Model development/calibration is in progress. • A summary of investigation: • If we assume • 100- α % of people make Destination-based choice • α % of people make Origin-based choice • Result: • Hammersmith: α = 55.5% • Goldhawk Road: α = 55.4% 24
- 25. Other UCL research • Dwell time • Relationship between passenger loading and dwell time • Impact of door and carriage designs • Passenger movements within stations • Railway Traffic Control • Railway Business management 25
- 26. CONCLUSION 26
- 27. As Concluding Marks 27 • No data is perfect, but our understanding can be better • Active data collection (and more collaboration between data collection technique developers and data users) rather than passively waiting for data! • New value could be created through collaboration between data collectors, processors, service/product developers, and users.