Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in...

34
Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden Y. Susilo 2 , A. C. Prelipcean 1,2 , G. Gid´ ofalvi 1 , A.Allstr¨om 3 , I. Kristoffersson 3 , J. Widell 3 1 Division of Geoinformatics, KTH Royal Institute of Technology 2 Department of Transportation Science, KTH Royal Institute of Technology 3 Sweco Transport System AB [email protected] @Adi Prelipcean adrianprelipcean.github.io 12 July 2016

Transcript of Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in...

Page 1: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Lessons from a trial of MEILIa smartphone based semi-automatic activity-travel diary

collector, in Stockholm city, Sweden

Y. Susilo2, A. C. Prelipcean1,2, G. Gidofalvi1,A. Allstrom3, I. Kristoffersson 3, J. Widell3

1Division of Geoinformatics, KTH Royal Institute of Technology2Department of Transportation Science, KTH Royal Institute of Technology

3Sweco Transport System AB

[email protected]

@Adi Prelipcean

adrianprelipcean.github.io

12 July 2016

Page 2: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Outline

This presentation will be about:1. Travel diaries

– Definition and uses– Collection methods

2. MEILI: an (activity) travel diary collection, annotationand automation system

– System Overview– Collected Data

3. Case study and Lessons Learned– Experimental setup– Results– Lessons learned: survey schedule– Lessons learned: user experience– Lessons learned: data collection

4. Summary and conclusions2

Page 3: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Travel behaviour

How do we use travel behaviour?

Some of the main reasons for analyzing travel behaviour are:

I to investigate the reasons and mechanisms that underliean individual’s travel decision making process,

I to predict the effect of implementing new transportationpolicies or changing the transportation infrastructure, or

I to understand the dynamic of transportation movementwithin study areas.

3

Page 4: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Travel behaviour

How do we use travel behaviour?

Some of the main reasons for analyzing travel behaviour are:

I to investigate the reasons and mechanisms that underliean individual’s travel decision making process,

I to predict the effect of implementing new transportationpolicies or changing the transportation infrastructure, or

I to understand the dynamic of transportation movementwithin study areas.

3

Page 5: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Travel behaviour

How do we use travel behaviour?

Some of the main reasons for analyzing travel behaviour are:

I to investigate the reasons and mechanisms that underliean individual’s travel decision making process,

I to predict the effect of implementing new transportationpolicies or changing the transportation infrastructure, or

I to understand the dynamic of transportation movementwithin study areas.

3

Page 6: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

(Activity) Travel diaries

What are they?

A way of summarizing where, why and how a user traveledduring a defined time frame by specifying:

I The destination of a trip

I The trip’s purposeI The means of transportation, i.e., trip legs

Img: http://soarministries.com/hp_wordpress/wp-content/uploads/2011/08/Destinations-Icon.jpg 4

Page 7: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

(Activity) Travel diaries

What are they?

A way of summarizing where, why and how a user traveledduring a defined time frame by specifying:

I The destination of a tripI The trip’s purpose

I The means of transportation, i.e., trip legs

Img: https://cdn2.vox-cdn.com/thumbor/93Yaxs7y3Tb8tzFfppyRsSn_yN8=/1020x0/cdn0.vox-cdn.com/uploads/chorus_asset/file/2509782/confused_man.0.jpg

4

Page 8: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

(Activity) Travel diaries

What are they?

A way of summarizing where, why and how a user traveledduring a defined time frame by specifying:

I The destination of a tripI The trip’s purposeI The means of transportation, i.e., trip legs

Img: https://d3ui957tjb5bqd.cloudfront.net/images/screenshots/products/4/42/42990/white-transportation-icons-300x200.jpg

4

Page 9: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

(Activity) Travel diaries

How to collect them?

I Traditionally - Users declare what they have done in asurvey, e.g., PP or CATI

I New methods - E.g., GPS collection + Web and MobileGIS based interaction

Img: http://www.schoolsurveyexperts.co.uk/i/photos/paper_survey.jpg

5

Page 10: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

(Activity) Travel diaries

How to collect them?

I Traditionally - Users declare what they have done in asurvey, e.g., PP or CATI

I New methods - E.g., GPS collection + Web and MobileGIS based interaction

5

Page 11: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

MEILI: an (activity) travel diary collection,

annotation and automation system

What is MEILI?

MEILI-travel diary collection, annotation & automation systemI MEILI Mobility Collector, which is a smartphone app

that collects trajectories fused with accelerometerreadings from users.

I MEILI Travel Diary, which is a web app that allowsusers to annotate trajectories into travel diaries.

I MEILI Database, which is the database that stores boththe collected and annotated data.

I MEILI API, which securely connects the MobilityCollector and the Travel Diary to the Database.

I MEILI AI, which is an Artificial Intelligence module thatautomatically annotates the trajectories collected byusers. 6

Page 12: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

MEILI Workflow

7

Page 13: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Case study

Collection details

I Stockholm, Sweden

I Traditional PP collection: 42 users, 29th September 2014

I Modern MEILI collection: 30 users, 29 Sept - 05 Oct 2014

I 28 users collected with both PP and MEILI

I users work in transportation

I PP collected 94 trips

I MEILI collected 87 trips (on the 29th) and 608 trips(during the whole period)

I trip correspondence between PP and MEILI was foundbased on temporal co-occurance and purpose matching

8

Page 14: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Case study

A brief overview of the collected data

Statistics computed on intersection set are similar. MEILI offers moreinformation, i.e., waiting time and spatial and temporal indicator values.

9

Page 15: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Case study

A brief overview of the collected data

MEILI captures tripleg-level information, and identifies potentialproblems with the collection.

9

Page 16: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Case study

A brief overview of the collected data

The difference between the time a user has to wait for ”personal”transportation modes and public transportation modes.

When waiting for public transportation modes, the time spent waitingamounts for roughly 10-20% of the whole trip duration.

9

Page 17: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

User feedback

34 users provided feedback on:I User Interface (based on map operations)

I MEILI Mobility Collector ease of install (85% reported noproblems)

10

Page 18: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

User feedback

34 users provided feedback on:

I MEILI Mobility Collector ease of install (85% reported noproblems)

10

Page 19: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

User feedback

34 users provided feedback on:

I MEILI Mobility Collector ease of install (85% reported noproblems)

I MEILI Mobility Collector’s effect on battery life (50% reportedlittle effect on battery life, 25% reported a major effect on batterylife)

10

Page 20: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

User feedback

34 users provided feedback on:

I MEILI Mobility Collector ease of install (85% reported noproblems)

I MEILI Mobility Collector’s effect on battery life (50% reportedlittle effect on battery life, 25% reported a major effect on batterylife)

I MEILI Travel Diary trip and tripleg annotation process

10

Page 21: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

User feedback

34 users provided feedback on:

I MEILI Mobility Collector ease of install (85% reported noproblems)

I MEILI Mobility Collector’s effect on battery life (50% reportedlittle effect on battery life, 25% reported a major effect on batterylife)

I MEILI Travel Diary trip and tripleg annotation process

I selecting a trip’s destination from the POI dataset (47% found itdifficult to find a POI and 20% would like more POI alternatives)

10

Page 22: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

User feedback

34 users provided feedback on:

I MEILI Mobility Collector ease of install (85% reported noproblems)

I MEILI Mobility Collector’s effect on battery life (50% reportedlittle effect on battery life, 25% reported a major effect on batterylife)

I MEILI Travel Diary trip and tripleg annotation process

I selecting a trip’s destination from the POI dataset (47% found itdifficult to find a POI and 20% would like more POI alternatives)

I wrongfully detected trips by MEILI (60% reported at least onewrongfully detected trip)

10

Page 23: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

User feedback

34 users provided feedback on:

I MEILI Mobility Collector ease of install (85% reported noproblems)

I MEILI Mobility Collector’s effect on battery life (50% reportedlittle effect on battery life, 25% reported a major effect on batterylife)

I MEILI Travel Diary trip and tripleg annotation process

I selecting a trip’s destination from the POI dataset (47% found itdifficult to find a POI and 20% would like more POI alternatives)

I wrongfully detected trips by MEILI (60% reported at least onewrongfully detected trip)

I collection integrity (67% found smartphone collection moreinvasive)

10

Page 24: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Lessons learned

User Interface for AnnotationsI improved the UI / UX based on user feedback and

discussions with UI / UX experts

I implemented sequential navigation through tripsI implemented the crowdsourcing of public and

transportation POIs (e.g., if one users declares a missingbus stop it is available for the rest)

I 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

11

Page 25: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Lessons learned

User Interface for AnnotationsI improved the UI / UX based on user feedback and

discussions with UI / UX experts

I implemented sequential navigation through tripsI implemented the crowdsourcing of public and

transportation POIs (e.g., if one users declares a missingbus stop it is available for the rest)

I 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

11

Page 26: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Lessons learned

User Interface for Annotations

I improved the UI / UX based on user feedback anddiscussions with UI / UX experts

I implemented sequential navigation through trips

I implemented the crowdsourcing of public andtransportation POIs (e.g., if one users declares a missingbus stop it is available for the rest)

I 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

11

Page 27: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Lessons learned

User Interface for Annotations

I improved the UI / UX based on user feedback anddiscussions with UI / UX experts

I implemented sequential navigation through trips

I implemented the crowdsourcing of public andtransportation POIs (e.g., if one users declares a missingbus stop it is available for the rest)

I 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

11

Page 28: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Lessons learned

User Interface for Annotations

I improved the UI / UX based on user feedback anddiscussions with UI / UX experts

I implemented sequential navigation through trips

I implemented the crowdsourcing of public andtransportation POIs (e.g., if one users declares a missingbus stop it is available for the rest)

I 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

11

Page 29: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

StorageI changed the data model from a point-based model into a

period-based model– trips and triplegs are explicitly modeled in the period-based

model– added the concept of passive period between consecutive trips

that describes the time spent performing an activity– added the concept of passive period between consecutive

triplegs that describes the time spent waiting fortransportation mode

I the new data model has been indexed and optimized (seePrelipcean et al. 2016 - MEILI: a travel diary collection,annotation and automation system, presented at MobileTartu 2016, submitted to Journal of TransportGeography)

12

Page 30: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Inference Methods and Artificial Intelligence

Some words on AI and segmentation tasks

I Trip segmentation - based on heuristics rules obtainsreasonable accuracy (P=96.7%, R=73.8%)

I Tripleg segmentation - changed from implicit to explicitby detecting sequences of GPS points with low deviationin movement characteristics - f(speed, accelerometer)

I Travel mode inference - changed from point-basedRandom Forest to a Nearest-Neighbor point-basedconsensus within a period

I Destination inference - difficult to obtain high precisionwith limited user history since most personal POIs (workand home locations) are missing from the database

I Purpose inference - limited by the performance of thedestination inference

13

Page 31: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

What do we gain from using modern methods?

I More detailed trip and trip-leg level information that ismissing in traditional methods

I The data are already centralized and stored using a datamodel that reduces the complexity of performing traveldiary specific operations

I Can use spatial and temporal indicators to assess thequality of the collected data and propose ground truthcandidates

I MEILI can be reused for data collection in differentcountries if localized

I MEILI is a viable alternative to study travel diaries on abroader scale (not limited to regions / countries) and ona wider time frame (not limited to one day / week)

14

Page 32: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

SummaryI introduced MEILI, an open source travel diary collection,

annotation and automation system.

I presented MEILI’s modular architecture that isolates thedevelopment process to each module

I provided brief discussions on inferences and AI used intravel diaries

I provided a set of valuable lessons learned during casestudies of applying MEILI to collect travel diaries

15

Page 33: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Acknowledgements and References

AcknowledgmentsThis work was partly supported by Trafikverket (SwedishTransport Administration) under Grant “TRV 2014/10422”.

ReferencesI source code for MEILI https://github.com/Badger-MEILII Mobility Collector - Prelipcean, A. C., Gidofalvi, G., & Susilo, Y. O.

(2014). Mobility collector. Journal of Location Based Services,8(4), 229-255.

I a framework for the comparison of travel diary collection systems -Prelipcean, A. C., Gidofalvi, 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.

I on AI performance measures relevant to travel diaries - Prelipcean,A. C., Gidofalvi, G., & Susilo, Y. O. (2016). Measures of transportmode segmentation of trajectories. International Journal ofGeographical Information Science, 30(9), 1763-1784.

16

Page 34: Lessons from a trial of MEILI a smartphone based semi-automatic activity-travel diary collector, in Stockholm city, Sweden

Thank you for your attention!Questions and Discussions

Adrian C. PrelipceanPhd StudentDivision of GeoinformaticsKTH, Royal Institute of Technologyhttp://adrianprelipcean.github.io/[email protected]@Adi Prelipcean

This work is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License.