FREIGHT DATA COLLECTION insights from US and Singapore projects Moshe Ben-Akiva, Tomer Toledo, Jorge Santos, Nathanael Cox, YinJin Lee, Fang Zhao, Vittorio Marzano ABJ40 Freight Survey Subcommittee 1
Outline Background and motivation US truck drivers survey - RP/SP data collection - some results Urban freight in Singapore - modeling architecture - data collection effort 2
Outline Background and motivation US truck drivers survey - RP/SP data collection - some results Urban freight in Singapore - modeling architecture - data collection effort 3
Problem and Opportunity Increasingly difficult to conduct reliable surveys via traditional approaches - Response rates are low - Non-representative samples based on convenience and intercept sampling - Short respondent attention span and limited ability to accurately recall information Smartphones, GPS loggers and RFID tags unobtrusively collect a wealth of valuable information - Better quality and quantity of data, especially when combined with information from shippers and carriers 4
Background Next-generation freight data collection: - leverage pervasive smartphone and GPS loggers, advanced sensing and communication technologies and machine learning architecture - deliver previously unobtainable range of data reflecting what shippers and carriers do, not what they say they did 5
Solution Future Mobility Survey (FMS): technology developed to innovate travel behavior surveys Machine learning algorithms combined with web-based user input to extend and validate smartphone sensing data 6
FMS Underlying technology has been developed, tested and proven effective Yields more detailed and varied data than traditional travel survey approaches Freight application of FMS in the following two studies 7
Outline Background and motivation US truck drivers survey - RP/SP data collection - some results Urban freight in Singapore - modeling architecture - data collection effort 8
survey of truck route choice This study data collected directly from drivers two phases: - Phase I Driver questionnaires with route choice stated preferences (SP) - Phase II GPS-based revealed preferences (RP) data 9
Phase I Key findings Wide variability in preferences towards toll roads and tolls Route choices depend on multiple factors - Travel time, tolls, delays - Toll bearing terms - Driver compensation method - Shipment characteristics For more details: Moshe Ben Akiva, Hilde Meersman, Eddy van de Voorde (eds), Freight Transport Modelling, Emerald Books, May 2013 10
Phase II RP data collection (adaptation of FMS) 11
GPS logger Trucks equipped with off-the-shelf loggers (SANAV CT-24) - Monitor all trips continuously - Transmit data in real-time to server Collects: - Location data - Speed - Timestamp Report Intervals - Time intervals - Minimum distance 12
Backend Algorithms Applied to the data received by the backend (MIT server): - Trace creation (FMS) - Stop detection (FMS) - Map Matching (Open Street Map) - Toll detection (Open Street Map) 13
Web interface Validate and correct movement information Collect additional information - Pick-up & delivery schedules - Cargo type - Tolls, methods of paying Exit survey - Personal information - Context specific SP 14
Web interface 15
Exit Survey 16
Data collection process Over the phone using lists of trucking companies At truck stops and rest areas - Indiana - Massachusetts - Texas - Ontario 17
89 completed participants 2991 days of data - 2067 fully validated 16437 stops detected - 11316 validated 3750 toll crossings Data collected so far 18
Driver type: Long tour Day 50 40 Time Space Diagram User 59 Load Unload Rest Toll Other Day 50 40 30 20 10 0 0 0 500 1000 1500 2000 2500 3000 3500 Distance from Home Location (km) 30 20 10 19
Driver type: Short tour Time Space Diagram User 66 Day 50 Load Unload Rest Toll Other Day 50 40 30 40 30 20 20 10 10 0 0 200 400 600 800 1000 Distance from Home Location (km) 0 20
Driver type: Gypsy Time Space Diagram User 141 Day Load Unload Rest Toll Other Day 30 20 30 20 10 0 0 0 500 1000 1500 2000 Distance from Home Location (km) 10 21
Same driver, different route 22
Same driver, different route 23
Same day, different route 24
Conclusions Combined GPS and web-based data collection is feasible and effective. Significant variability in: - Driver tour patterns - Driver route choice behavior 25
Outline Background and motivation US truck drivers survey - RP/SP data collection - some results Urban freight in Singapore - modeling architecture - data collection effort 26
The SMART-FM project SMART: Singapore-MIT Alliance for Research and Technology FM: Future Urban Mobility project SimMobility Integrated platform Collaborative research laboratory Development and evaluation of mobility portfolios 27
Urban freight data collection end consumer establishment carrier driver tours and routing carbon footprint and fuel cons. pickup & delivery operations logistics choices agents' behaviour o-d freight flows estimation ** *** *** ** *** *** ** *** ** *** * *** *** * *** * * * * 28
Data collection: challenges Adaptation of the US truck driver survey (inter-city) to urban freight environment - Track a variety of commercial vehicles - More frequent stops - More diverse activities Challenges - Lower GPS data quality - Shorter stop durations - Denser road network - Large number of origins and destinations 29
Data collection: solutions Enhance questionnaires augment pre-survey with frequent stops, routes, trips, activity types etc. to capture repetitive behavior modify stop questions to include more diverse activities Improvements to stop/activity detection algorithms extend period of observation enhance machine learning algorithms to include user history and Points of Interest (POI) data 30
Data collection: progress Driver/vehicle survey: pilot testing in progress less expensive (120 US$), programmable GPS logger (SANAV CT-58) also: replace commercial GPS loggers with specially designed OBD devices to estimate carbon footprint Establishment survey: pilot of a tablet-based retailers survey 31
Data collection: progress 32
Conclusion Underlying technology has been developed, tested and proven effective for an Intercity Truck Survey - Yields detailed and varied data about route choice behavior - Adaptation and enhancement for Urban Freight Survey in process 33
Thank you! 34