Big data in Thessaloniki Josep Maria Salanova Grau Center for Research and Technology Hellas Hellenic Institute of Transport Email: jose@certh.gr - emit@certh.gr Web: www.hit.certh.gr
Big data in Thessaloniki Traditional sensors (loops, radars and cameras) Stationary sensors network: Point to point tracking of MAC ids along the network through Bluetooth detectors (more than 40). Dynamic sensors fleet: Floating Car Data provided in real time by a professional fleet (more than 1.000 vehicles). Cooperative technologies (COMPASS4D and COGISTICS): RSU is a static sensor and OBU is a dynamic sensors (CAM message). Social networks content
How do we use Big Data? Travel time estimation (average values and distributions) Mobility patterns identification OD matrices generation and validation Traffic flow estimations Route choice models development and calibration Macroscopic traffic models calibration Microscopic simulation models calibration Road Hazards Detection Personalized services for drivers (BMs) Other activities
Traditional sensors The surveillance system of the Peripheral Ring Road (more than 100.000 vehicles per day in both directions) The Thessaloniki s Urban Mobility Management System installed in the city center (more than 50.000 vehicles per day) The traffic lights management system of the whole city
Point to point BT network 10 detectors installed in 2012 (www.mobithess.gr) 2.3 million detections per week (peak period) 25.000 unique devices detected per day (one intersection) 600.000 tracked trips per week 20.000 tracked trips per day (one path) 33 detectors installed in 2014 (SEE-ITS & EASYTRIP) 1.5 million detections per week (off-peak period) 13.000 detections per day (one intersection) More detectors installed in other cities and in Bulgaria (SEE-ITS & EASYTRIP)
Point to point BT network 2012 2014 2014 Number of detections Intersection Monday Tuesday Wednesday Thursday Saturday Identity Friday 07/12/12 Sunday 09/12/12 03/12/12 04/12/12 05/12/12 06/12/12 08/12/12 A 9.217 10.017 9.877 10.481 10.713 9.457 7.390 B 11.803 11.923 11.414 11.968 12.790 9.793 8.019 C 14.623 15.282 15.434 16.683 17.281 14.253 11.849 D 20.620 20.500 19.473 21.291 22.826 17.952 14.086 E 10.481 10.753 10.364 10.678 10.841 8.398 6.476 F 21.360 21.901 21.223 17.872 23.027 17.655 14.635 G 23.768 23.583 21.681 21.538 23.506 16.525 12.002 H 11.540 11.796 11.136 12.513 8.801 9.385 7.552 I 8.138 7.973 7.900 8.098 8.769 5.643 4.522 J 6.655 6.839 7.040 6.642 7.338 5.012 3.863
Point to point BT network Valid trips Path identity Monday Tuesday Wednesday Thursday Friday Saturday Sunday 1 1.977 2.360 3.037 3.673 4.266 4.645 4.589 2 3.358 3.686 3.573 2.744 3.556 3.060 2.558 3 3.916 4.236 4.038 4.158 4.305 4.117 2.904 4 1.347 1.333 1.395 1.306 1.075 1.170 940 5 1.350 1.379 1.345 1.342 1.459 1.000 833 6 3.589 3.683 3.742 3.300 4.099 4.026 3.447 7 16.977 17.340 16.683 16.860 19.332 14.759 12.479 8 8.799 8.838 8.535 7.022 10.001 8.142 6.880 9 5.930 5.702 5.766 4.335 5.776 5.035 4.358 10 1.018 1.075 1.070 1.066 1.119 624 433 11 300 316 353 280 294 261 293 12 5.083 5.366 5.203 5.024 4.198 4.851 4.030 13 8.338 8.474 8.348 9.294 9.587 8.157 6.502 14 683 616 670 925 671 570 440 15 3.520 3.523 3.610 2.623 4.003 2.907 2.568 16 3.995 4.165 3.746 3.370 3.326 3.650 3.063 17 7.735 7.953 7.269 6.172 8.233 6.907 6.051 18 4.188 4.260 3.982 3.706 4.100 3.772 3.083 19 2.624 2.802 2.604 2.119 2.959 2.036 1.657 20 3.090 3.091 2.978 3.520 3.359 2.624 2.389 21 1.570 1.616 1.446 1.466 1.658 1.392 1.012 22 2.681 2.474 2.671 2.482 2.819 1.864 2.012
Point to point BT network Data filtering and processing for real time travel time estimation Invalid cases Long stop Other mode (pedestrian) Short stop Possible detour Other mode (bus) Valid cases Detector B Detector A Time (Trip End)
Point to point BT network Data filtering and processing for real time travel time estimation
Point to point BT network Real time travel time provision to drivers (VMS, internet, smart device)
Point to point BT network Traffic flow estimation from detections (20%)
Point to point BT network Without filter 15 filter 5 filter 60 filter Without filter Filtering each 5 min Filtering each 15 min Filtering each 60 min Ratio 0,3412 0,2179 0,1972 0,0442 R 2 0,9166 0,9193 0,9337 0,8594 Largest differences -401 / 623-26% / 75% -410 / 437-23% / 61% -336 / 389-22% / 57% -536 / 767-35% / 79%
Floating Car Data More than 1.000 vehicles (taxi fleet) Circulating 16-24 hours per day Pulse generated each 100 meters (10-12 seconds without congestion) or 60 seconds 500-2.500 pulses per minute More than 1.500 trucks (two fleet operators)
Floating Car Data Device ID GPS position (X, Y, Z) Orientation (degrees) Speed (km/h) Timestamp Zone
Floating Car Data Map matching Low cost GPS receiver (grey) High quality GPS receiver without CDGPS (green) The same receiver with CDGPS (black) The black circle has a 10-metre radius
Floating Car Data Detailed and updated maps can be generated from FCD National and international network Regional and interurban network Urban network
Floating Car Data Road direction One way Two ways
Floating Car Data Maximum, average and minimum speed profiles along Tsimiski (2 kilometers)
Combination: FCD & BT detectors Virtual static detectors (free, no maintenance, easy to change position)
Social content (Facebook check in) Spatial coverage of the city center locations every 5-15 minutes.
Thessaloniki Big Data framework Real time data processes and services provision in Thessaloniki Real world Traffic flows Filter Filter Travel times Historical traffic data Traffic flow estimation Travel time provision services Off-line processes Macroscopic traffic model Short term forecast Real time processes Routing services Real time services Traffic scenaria Data expansion Traffic congestion provision services
Thessaloniki Big Data framework Real time traffic conditions information based on a combination of traffic modeling and real time measurements (traffic flow and travel time)
Other applications Impacts of high intensity storms on urban transportation: Applying traffic flow control methodologies for quantifying the effects Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 All other locations
Other applications Impacts of high intensity storms on urban transportation: Applying traffic flow control methodologies for quantifying the effects 08/02/2013 15/02/2013 22/02/2013 Average daily difference (%) Duration of event s impact Average difference during the event (%) Maximum difference during the event (%) Zone 1 20 20 18-10% 5:00-8:00-10,7% -34,8 % Zone 2 33 33 27-18% 5:00-12:00-14,1% -47,5 % Zone 3 32 32 28-13% 5:00-12:00-15,6% -49,3 % Zone 4 20 20 18-10% 5:00-11:00-5,5% -37,7 % Zone 5 52 52 33-37% 5:00-19:00-23,1% -67,8 % Athens Region 32 32 28-13% - -8,3% -31,6% Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Athens region 08/02/2013 194 619 981 369 280 27.101 15/02/2013 160 627 1.025 366 315 27.143 22/02/2013 144 491 747 307 181 20.530
Other applications Taxi observatory in Barcelona The area of study 18 municipalities 2.650.000 habitants 330 km 2 more than 12.800.000 daily trips The network 1.600 km 8.000 intersections 20.000 links The taxi market 10.482 taxis 11.076 driver licenses 24 dispatching centers 272 taxi stands 225.000 daily trips The taxi trips database 9 years (2004-2012) 1.200.000 valid taxi trips 10.000.000 km (50% occupied) 600.000 hours (33% occupied) 270.000 trips with GPS data
Other applications Taxi observatory in Barcelona Taxi demand density maps O-D matrix estimation from GPS coordinates of customers Origins and Destinations
Other applications Agent based model for taxi services
Other applications Agent based model for taxi services Minimum number of taxis (taxis/hour*km 2 ) 26 First best solution 34 Second best solution 30
Other applications Agent based model for taxi services Access / Waiting time (users) Benefit (drivers)
Other applications Agent based model for taxi services
Other applications Agent based model for taxi services Waiting time (users) Benefit (drivers)
Other applications Agent based model for taxi services
Thank you! Dr. Josep Maria Salanova Grau jose@certh.gr +30 2310 498 433