MEILI: a travel diary collection, annotation and automation system
1 like582 views
MEILI is a travel diary collection, annotation and automation system that was created to improve upon the limitations of previous attempts. It implements mobility collectors for both Android and iOS, improves the user interface based on feedback, splits the system into well-defined modules, and changes the data model from point-based to period-based. The presentation evaluates MEILI's performance based on a case study with over 170 users, analyzing the distribution and response times of create, read, update and delete operations. It identifies read operations as a bottleneck and discusses lessons learned regarding user studies, usability, and applying artificial intelligence in transportation research.
MEILI: a travel diary collection, annotation and automation system
- 1. MEILI: a travel diary collection, annotation and automation system A. C. Prelipcean1 , G. Gid´ofalvi1 , Y. Susilo2 1Division of Geoinformatics, Dept. of Urban Planning and Environment 2Dept. of Transportation Science KTH Royal Institute of Technology acpr@kth.se @Adi Prelipcean adrianprelipcean.github.io 01 July 2016
- 2. Outline This presentation will be about: 1. Travel behaviour 2. Travel diaries and travel diary collection methods 3. Towards an automated travel diary collection system – First implementation – Lessons learned 4. MEILI: a travel diary collection, annotation and automation system – Improvements over previous attempt – MEILI architecture and operation flow – A brief case study 5. Summary and conclusions 2
- 3. Travel behaviour How do we use travel behaviour? Some of the main reasons for analyzing travel behaviour are: to investigate the reasons and mechanisms that underlie an individual’s travel decision making process, 3
- 4. Travel behaviour How do we use travel behaviour? Some of the main reasons for analyzing travel behaviour are: to investigate the reasons and mechanisms that underlie an individual’s travel decision making process, to predict the effect of implementing new transportation policies or changing the transportation infrastructure, or 3
- 5. Travel behaviour How do we use travel behaviour? Some of the main reasons for analyzing travel behaviour are: to investigate the reasons and mechanisms that underlie an individual’s travel decision making process, to predict the effect of implementing new transportation policies or changing the transportation infrastructure, or to understand the dynamic of transportation movement within study areas. 3
- 6. (Activity) Travel diaries What are they? A way of summarizing where, why and how a user traveled during a defined time frame by specifying: The destination of a trip Img: http://soarministries.com/hp_wordpress/wp-content/uploads/2011/08/Destinations-Icon.jpg 4
- 7. (Activity) Travel diaries What are they? A way of summarizing where, why and how a user traveled during a defined time frame by specifying: The destination of a trip The trip’s purpose Img: https://cdn2.vox-cdn.com/thumbor/93Yaxs7y3Tb8tzFfppyRsSn_yN8=/1020x0/cdn0.vox-cdn.com/ 4
- 8. (Activity) Travel diaries What are they? A way of summarizing where, why and how a user traveled during a defined time frame by specifying: The destination of a trip The trip’s purpose The means of transportation, i.e., trip legs Img: https://d3ui957tjb5bqd.cloudfront.net/images/screenshots/products/4/42/42990/ 4
- 9. (Activity) Travel diaries How to collect them? Traditionally - Users declare what they have done in a survey, e.g., PP or CATI Img: http://www.schoolsurveyexperts.co.uk/i/photos/paper_survey.jpg 5
- 10. (Activity) Travel diaries How to collect them? Traditionally - Users declare what they have done in a survey, e.g., PP or CATI New methods - E.g., GPS collection + Web and Mobile GIS based interaction 5
- 11. Towards an automated travel diary collection system Considerations on an initial attemp use smartphones to collect GPS data fused with accelerometer readings display the collected data via a web interface and let users annotate their data into travel diaries store the annotations in a database 6
- 12. Towards an automated travel diary collection system Implementation of the initial attempt implemented Mobility Collector on Android (Prelipcean et al. 2014) 7
- 13. Towards an automated travel diary collection system Implementation of the initial attempt implemented Mobility Collector on Android (Prelipcean et al. 2014) implemented a Web Interface for basic user interaction 7
- 14. Towards an automated travel diary collection system Implementation of the initial attempt implemented Mobility Collector on Android (Prelipcean et al. 2014) implemented a Web Interface for basic user interaction stored the annotations in the database in a point-based model 7
- 15. Towards an automated travel diary collection system Implementation of the initial attempt implemented Mobility Collector on Android (Prelipcean et al. 2014) implemented a Web Interface for basic user interaction stored the annotations in the database in a point-based model trialed the system on 30 users (working in transportation) 7
- 16. Initial attempt Lessons learned Mobility Collector is battery efficient Android only implementation restricts the user pool in iOS predominant markets (such as Sweden) users wanted more freedom to interact with their data in the web interface difficult to extract trips and triplegs from a point-based model difficult to improve the system due to the lack of isolated functional components 8
- 17. MEILI Improvements over the previous attempt implemented Mobility Collector for iOS 9
- 18. MEILI Improvements over the previous attempt implemented Mobility Collector for iOS improved the UI / UX based on user feedback and discussions with UI / UX experts 9
- 19. MEILI Improvements over the previous attempt implemented Mobility Collector for iOS improved the UI / UX based on user feedback and discussions with UI / UX experts 9
- 20. MEILI Improvements over the previous attempt implemented Mobility Collector for iOS improved the UI / UX based on user feedback and discussions with UI / UX experts split the system into well defined functional modules 9
- 21. MEILI Improvements over the previous attempt implemented Mobility Collector for iOS improved the UI / UX based on user feedback and discussions with UI / UX experts split the system into well defined functional modules changed the data model from a point-based model into a period-based model 9
- 22. MEILI Improvements over the previous attempt implemented Mobility Collector for iOS improved the UI / UX based on user feedback and discussions with UI / UX experts split the system into well defined functional modules changed the data model from a point-based model into a period-based model complemented the web interface operations: – Create - insertion of trips, triplegs, locations, POIs – Read - pagination operations between consecutive trips – Update - update of trips, triplegs, locations, POIs – Delete - deletion of trips, triplegs, locations, POIs 9
- 23. MEILI Architecture and operation flow 10
- 24. MEILI Implementation details Mobility Collector - module for collecting trajectories fused with accelerometer readings (Android, iOS) Travel Diary - module for each user to annotate her collected data into travel diaries (various HTML libraries, NodeJS, ExpressJS) Database - module for storing the data collected by Mobility Collector and the annotations performed via the Travel Diary (PostgreSQL, PostGIS) API - module that securely connects the Database module to the Travel Diary module (NodeJS, ExpressJS) AI - middleware module that extracts travel diary semantics from trajectories (custom code) 11
- 25. MEILI Case study overview 171 users used MEILI for at least one day 51 users used MEILI for at least one week collected 2142 trips and 5961 triplegs schema of 16 different travel modes schema of 13 different purposes POI set of 21953 entries in the database transportation POI set of 6610 entries in the database 12
- 26. MEILI Overview of the performed CRUD operations the two peaks correspond to emails sent to users 13
- 27. MEILI Overview of the performed CRUD operations the two peaks correspond to emails sent to users the two off-peaks correspond to weekends and lack of data to annotate 13
- 28. MEILI Overview of the performed CRUD operations the two peaks correspond to emails sent to users the two off-peaks correspond to weekends and lack of data to annotate there is little variability in the distribution of CRUD operations 13
- 29. MEILI Distribution of CRUD operations in relation to the time of day 8PM-12AM 4PM-8PM 12PM-4PM 8AM-12PM 4AM-8AM 12AM-4AM M ondayTuesday W ednesdayThursday FridaySaturday Sunday 42% 19% 9% 32% 0% 0% 47% 38% 2% 12% 12% 13% 100% 40% 15% 16% 40% 28% 5% 0% 9% 3% 44% 34% 3% 81% 0% 0% 0% 13% 3% 2% 0% 0% 0% 0% 3% 0% 20% 0% 0% 2% 0 20 40 60 80 100 People interact most with MEILI – Tuesday to Friday during morning and noon – Saturday to Monday during evening and night 14
- 30. MEILI Response time for operations - Create and Read 0 200 400 600 800 1000 1200 1400 1600 Previous(14.7% ) N ext(85.3% ) ExecutionTime(ms) Read Operations in MEILI Operation performed often Execution Time CUD Non-indexed Indexed 0 5 10 15 20 D estination(0.1% ) Location(0.2% )PO I(1.8% ) Purpose(6.9% ) Trip(21.7% ) Transition(34.6% ) Tripleg(34.6% ) ExecutionTime(ms) Create Operations in MEILI Operations performed often Non-indexed Indexed Read operations are expensive Indexing reduces the execution time by an order of magnitude Even after indexing, read operations are a bottleneck 15
- 31. MEILI Response time for operations - Update and Delete 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 Trip(44.5% ) Tripleg(55.5% ) ExecutionTime(ms) Delete Operations in MEILI Both operations are performed often Non-indexed Indexed 0 0.5 1 1.5 2 Location(0.3% ) PO I(1.4% ) Trip(48.6% )Tripleg(49.7% ) ExecutionTime(ms) Update Operations in MEILI Operations performed often Non-indexed Indexed Both types of operations are cheap before indexing Indexing reduces the execution time to nanoseconds 16
- 32. MEILI Lessons learned it is difficult to organize case studies that overlap release dates of different smartphone OSes getting feedback from a specific non-representative user group can be damaging UX is influenced by the ease of information access in the MEILI Travel Diary and by the waiting time for each operation Read operations (e.g., pagination) are a bottleneck significant improvements needed for making MEILI Travel Diary bug-free and intuitive the methodology for Artificial Intelligence applied in Transportation Science is underdeveloped 17
- 33. Summary introduced MEILI, an open source travel diary collection, annotation and automation system. designed MEILI’s architecture in a modular way to isolate the development process to each module trialed MEILI for 9 days in Stockholm and studied the distribution of operations performed by users to identify peak and off-peak periods measured the execution times of the MEILI operations to identify bottlenecks and reduced their impact by applying various indexing structures provided a set of valuable lessons learned during multiple case studies of applying MEILI to collect travel diaries 18
- 34. Acknowledgements and References Acknowledgments This work was partly supported by Trafikverket (Swedish Transport Administration) under Grant “TRV 2014/10422”. References source code for MEILI https://github.com/Badger-MEILI Mobility Collector - Prelipcean, A. C., Gid´ofalvi, G., & Susilo, Y. O. (2014). Mobility collector. Journal of Location Based Services, 8(4), 229-255. a framework for the comparison of travel diary collection systems - Prelipcean, A. C., Gid´ofalvi, G., & Susilo, Y. O. (2015). Comparative framework for activity-travel diary collection systems. In Models and Technologies for Intelligent Transportation Systems (MT-ITS), 2015 International Conference on (pp. 251-258). IEEE. on AI performance measures relevant to travel diaries - Prelipcean, A. C., Gid´ofalvi, G., & Susilo, Y. O. (2016). Measures of transport mode segmentation of trajectories. International Journal of Geographical Information Science, 30(9), 1763-1784. 19
- 35. Thank you for your attention! Questions and Discussions Adrian C. Prelipcean Phd Student Division of Geoinformatics KTH, Royal Institute of Technology http://adrianprelipcean.github.io/ acpr@kth.se @Adi Prelipcean This work is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License.