A User-Flocksourced Bus Intelligence System in Dhaka

the first

the world ----------

---

by albert ching May 7, 2012

Master’s Thesis Defense

In collaboration with Stephen J. Kennedy and Muntasir Mamun Advised by Chris Zegras with the gracious help of Zia Wadud, Paul Barter, and Eran Ben-Joseph

Inspired by the Kewkradong team in Dhaka as well as all the entrepreneurs promoting sustainable transport in developing Asia

research question(s)

While smartphones can be designed to collect vast swaths of data, can flocks of people be organized and incentivized to collect data for a targeted period of time and place?

A!

Yes, in a big way.

research question(s)

If not all data in a city can be collected by flocks, can a sampled set be useful, especially if certain behaviors are predictable?

B!

Yes, less data can become big data.

ITERATIVE CITY

1

MOBILE MOBILITY

2

FLOCK- SOURCING

3 4

URBAN LUNCHPAD

theory context experiment results future

1000 SURVEYS

ITERATIVE CITY

1 theory

ITERATIVE CITY

1 theory

1

The future of cities is no longer held in one big plan but in a thousand little, measured strokes.

1 Cheap measurement (spatial + temporal)

1 Masterplanà Simulation à Iteration

WHICH CITIES WILL BENEFIT?

1

MOBILE MOBILITY

2

context

18 Million People 100,000 Cars

<1%

DHAKA

9 Million People 9 Million Two-Wheelers

3 Million Cars >100%

JAKARTA

(20.0)

-

20.0

40.0

60.0

80.0

100.0

100 1,000 10,000 100,000

Car

s, t

ruck

s an

d p

erso

n p

er 1

00 p

erso

ns

Income per person (GDP per capita, $USD, inflation adjusted)

United States

Indonesia

China India Bangladesh

Hong Kong

Singapore S

andr

a an

d A

rcha

ya (2

007)

mot

oriz

atio

n

infl

ecti

on o

f $5,

000

per

capi

ta G

DP

Barter “lock-in” line of 10% car ownership

Japan

Asia

Rest of the World

Income

Mot

oriz

atio

n

Mobile rickshaw wallah in India

Can Owning a Cell Phone Reduce the Desire to Own a Car?

Marketing 1

(Real-Time) Operator Services

3

Users

Information can improve accessibility

to, comfort and safety of shared vehicles

Information can help monitor and evaluate city

performance in a more precise and timely manner

than ever before

Regulators Operators

Information can improve efficiency, management

and profitability of shared fleets

Cars = aspiration

(Real-Time) User Services

2 Responsive

City Planning

4

GO-Jek Dial-a-Motorcycle Transport in Jakarta, August 2011

Fazilka Dial-a-Rickshaw in Punjab, August 2011

Are these business sustainable + scalable?

entrepreneurs

Constellation of Mobile-Driven Mobility Experiments

Sustainable Unsustainable

Navigation

Congestion

Tracking

Vehicle-Security

On-Demand

Safety Alerts

On-Demand

Fare-Tracking

On-Demand Real Time

Arrival Info

Real Time

Arrival Info

Bus Delays

On-Demand Bicycle

Sharing

Singapore

August 2011

Jakarta

Delhi

Bangalore

Fazilka

Kuala

Lumpur

Bangkok

Dhaka

Can an outside institution accelerate experimentation?

FLOCK- SOURCING

3

experiment

Guided crowdsourcing

UBIQUITOUS SENSING

All the data, all the time

Sensors

Privacy Closed

Expensive Data processing

Only objective metrics

Real-time urban data collection techniques

CROWDSOURCING

Some data for lots of disparate times and places

Crowds + Sensors

Gathering sufficient and relevant data

Predictability of mobility (Song, Qu, Blumm, Barabasi 2010)

Lots of data for a specific time and place

Flocks + Sensors

Organizing the flock Flock bias

Real-time urban data collection techniques

FLOCKSOURCING

Sensors

Hardware

Platform

Connectivity

Data storage

Data verification & analysis

Incentivized Volunteers

Unsmartphones

None

Bluetooth

Organized Flock

Organized Vehicles

Involuntary Tracking

Smartphones Tablets PC

Cell network Mobile data Wi-Fi

Excel

Android iPhone Web

Local Cloud

Statistical Packages

Visualization

Software / App MIT App Inventor

Machine learning

Visualiza- tion APIs

Flocksourcing Workflow

main bottlenecks

“Launch and iterate” co-development

Bus Details

Passenger Count

Survey

$10-$15 per person per

day

$175 and rapidly

declining Free $4

per 1 GB Free

Sensors Hardware Connectivity Data storage Software / App

Cost Structure

Flocksourcing

Parallel Experiments

Flock size & nature

8 paid volunteers ($10 per person per day)

Organized by Kewkradong Bangladesh

Target buses

36 & 27 Lines (10 km each)

Data collection target

100 surveys 120 one-way rides

Flock size & nature 3-8 unpaid volunteers

($30 per data plan)

Target buses lines

Any

Data collection target

None

Crowdsourcing the world’s first

experiment

Dhaka Boston

Experimental Design

Bus Details Bus Number Bus Destination Bus Company No. of Seats Speed Location Time Crowding Passenger Count Female Passenger Count

Survey Gender Age Home Location Work Location One-Way Commute Income Phone Ownership Rider Satisfaction Biggest Complaint Riding Frequency

Metrics

*Survey data linked to bus data

Quantitative Qualitative

4

results

1000 SURVEYS

Data Collection

Dash

Kb16

Kb10 Kb20 Kb7 Kb14

Kb13

Kb2

Kb8

Individual Flock Traces

research question(s)

While smartphones can be designed to collect vast swaths of data, can flocks of people be organized and incentivized to collect data for a targeted period of time and place?

A!

research question(s)

If not all data in a city can be collected by flocks, can a sampled set be useful, especially if certain behaviors are predictable?

B!

Dimensions of Data Itself

Predictability

Data Value

Need Less Data

Need More Data

High

Low

High

Low

Ubiquitous Sensing

Crowdsourcing

Dimensions of Data

Collection

1 BUS

CROWDING

2 BUS TRAVEL

TIMES

3 BUS ROUTES

#36

1!

2!

3!

4!

5!

6!

7!

8!

9!

Average Sample Size %Std Dev Min Max

passenger count

Std Dev

variability

32

32

34

64

47

64

62

85

15

15

16

12

14

11

12

11

9

15

64%

68%

39%

41%

27%

32%

35%

25%

58%

2

5

9

9

14

11

11

11

5

51

51

47

54

49

52

50

50

52

24

23

30

33

41

38

32

36

27

BUS CROWDING

BUS CROWDING

1!

2!

3!

4!

5!

6!

7!

8!

9!

8! 9! 10! 11! 12! 1! 2! 3! 4! 5! 6!

empty seats

7!

am

pm

Average #36

+16!

+17!

+10!

+7!

(0)!

+2!

+8!

+4!

+13!BUS CROWDING

#36

one-way commute

OVERALL

Inbound

Outbound

9!1!12.4 km

Weekday

Weekend

1:01 0:52 0:59 0:49 1:22

0:54 1:22

0:54 1:01 1:32

Average 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00

1:03 0:52 0:59 0:49 1:22

0:55 1:22

0:55 1:01 1:32

Average 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00

0:53 0:52 0:52

Average 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00

1:02 1:42

0:46 0:59 0:56 1:01 1:32

Average 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00

0:59 0:46 0:58 0:49 1:22

0:41

Average 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00

BUS TRAVEL TIMES

BUS ROUTING

Ubiquitous Sensing

Crowdsourcing

BUS TRAVEL TIMES BUS

CROWDING

BUS ROUTING

Predictability

Data Value

High

Low

High

Low

+ Machine Learning

Self-organizing flock

BUS TRAVEL TIMES BUS

CROWDING

BUS ROUTING

BUS RIDERSHIP

BUS SATIS-

FACTION

URBAN LUNCHPAD

future

launchpad ----------

The Urban Launchpad is a social-mission driven company launched to generate big data insights in places, and on problems where there is less data.

PUBLIC INFOSTRUCTURE

BEST BUS MAP IN THE WORLD

public

public

public public

public

30 buses (position, speed)

50 buses (position,

speed) flock of 25, 10 days

(satisfaction)

flock of 15, 5 days (crowding)

flock of 30, 15 days (counts)

Who will build?

OUR FIRST PRODUCT

the cheapest and easiest

the world --- -------- A BUS INTELLIGENCE SERVICE IN DHAKA

CUSTOMERS

TECHNOLOGY +

YOUR FLEET

1!

Ongoing data collection

TECHNOLOGY +

OUR FLOCKS

2

One-time data collection

Private bus and mini-bus operators, Paratransit (taxis, auto-rickshaws

cycle rickshaws)

City government, non-profits, academic institutions, new

mobility startups, citizen groups

PRICING

$50* per seat per month

$50* per flock member per day

*50% discount if data is made open to public for mash-up

Bus tracking hardware retails in US for $8-$20K per bus

Retails to less than $3 per survey using pilot results

Is there a viable business model?

Collaborators Stephen Kennedy, MIT DUSP Muntasir Mamun, Kewkradong Tonmoy Saad Bin Hussain, Kewkradong Xitu Masuk Ahmed, Kewkradong Swapon, Kewkradong Chonchol Morshed Alam, Kewkradong Raian Md. Shakhawat Chowdhury, Kewkradong Mamun Bhai, Kewkradong Share My Bus Dhaka & Boston Volunteers Principal Advisors Chris Zegras, MIT Asst. Prof. of Urban Studies and Planning Zia Wadud, BUET Prof of Civil Engineering Paul Barter, NUS Asst. Prof. at LKY School of Public Policy Eran Ben-Joseph, MIT Prof. of Urban Studies and Planning Entrepreneurs Navdeep Asija, Fazilka Eco-Cabs Ravee Aahluwalia, Patiala Eco-Cabs Sundara Raman, Ideophone Anenth Guru, Ideophone Sandeep Bhaskar, Ideophone Sanjeev Garg, Delhi Cycles Atul Jain, Delhi Cycle HR Murali, Namma Cycle Anthony Tan, My Teksi Hooi Ling Tan, My Teksi Nadiem Makarim, GO-Jek Arup Chakti, NITS

Leading Thinkers Apiwat Ratanwahara, Chulalongkorn University Sorawit Narupiti, Chulalongkorn University Charisma Chowdhury, BUET Moshahida Sultana, University of Dhaka Geetam Tewari, IIT-Delhi Anvita Arora, IIT-Delhi Rajinder Ravi, cycle rickshaw expert Tri Tjahjono, Univesiti Indonesia Jamillah Mohamad, University of Malaya Advocates Debra Efroymson, Work for a Better Bangladesh Maruf Rahman, Work for a Better Bangladesh Akshay Mani, EMBARQ Madhav Pai, EMBARQ Chhavi Dhingra, GTZ-India Eric Zusman, IGES Yoga Adiwinarto, ITDP Indonesia Restiti Sekartini, ITDP Indonesia Government Anisur Rahman, Dhaka Transport and Coordination Board Rajendar Kumar, Indian Dept of Information Technology Anil Sethi, Mayor of Fazilka Prodyut Dutt, ADB India Penny Lukito, BAPPENAS Indonesia Firdaus Ali, Jakarta Water Provision Industry RD Sharma, HI-BIRD Bicycles Comfort Cab Malaysia Jacob Yeoh, Yes! 4G Mobile Internet Malaysia Pornthip Konghun, Googlers Thailand James McClure, Google Singapore Kapil Goswami, Google India

Mahalo!

Collaborators Stephen Kennedy, MIT DUSP Muntasir Mamun, Kewkradong Tonmoy Saad Bin Hussain, Kewkradong Xitu Masuk Ahmed, Kewkradong Swapon, Kewkradong Chonchol Morshed Alam, Kewkradong Raian Md. Shakhawat Chowdhury, Kewkradong Mamun Bhai, Kewkradong Share My Bus Dhaka & Boston Volunteers Principal Advisors Chris Zegras, MIT Asst. Prof. of Urban Studies and Planning Zia Wadud, BUET Prof of Civil Engineering Paul Barter, NUS Asst. Prof. at LKY School of Public Policy Eran Ben-Joseph, MIT Prof. of Urban Studies and Planning Entrepreneurs Navdeep Asija, Fazilka Eco-Cabs Ravee Aahluwalia, Patiala Eco-Cabs Sundara Raman, Ideophone Anenth Guru, Ideophone Sandeep Bhaskar, Ideophone Sanjeev Garg, Delhi Cycles Atul Jain, Delhi Cycle HR Murali, Namma Cycle Anthony Tan, My Teksi Hooi Ling Tan, My Teksi Nadiem Makarim, GO-Jek Arup Chakti, NITS

Leading Thinkers Apiwat Ratanwahara, Chulalongkorn University Sorawit Narupiti, Chulalongkorn University Charisma Chowdhury, BUET Moshahida Sultana, University of Dhaka Geetam Tewari, IIT-Delhi Anvita Arora, IIT-Delhi Rajinder Ravi, cycle rickshaw expert Tri Tjahjono, Univesiti Indonesia Jamillah Mohamad, University of Malaya Advocates Debra Efroymson, Work for a Better Bangladesh Maruf Rahman, Work for a Better Bangladesh Akshay Mani, EMBARQ Madhav Pai, EMBARQ Chhavi Dhingra, GTZ-India Eric Zusman, IGES Yoga Adiwinarto, ITDP Indonesia Restiti Sekartini, ITDP Indonesia Government Anisur Rahman, Dhaka Transport and Coordination Board Rajendar Kumar, Indian Dept of Information Technology Anil Sethi, Mayor of Fazilka Prodyut Dutt, ADB India Penny Lukito, BAPPENAS Indonesia Firdaus Ali, Jakarta Water Provision Industry RD Sharma, HI-BIRD Bicycles Comfort Cab Malaysia Jacob Yeoh, Yes! 4G Mobile Internet Malaysia Pornthip Konghun, Googlers Thailand James McClure, Google Singapore Kapil Goswami, Google India

A

appendix

REVENUE POTENTIAL (FLEET ONLY)

$50 per seat per month

9,000 buses in Dhaka

5% 10% 25% 50% 75%

100%

$270K $540K $1.4M $2.7M $4.1M $5.4M

penetration rate

x

annual revenue

Current Bus Riders in Dhaka

Young, Male, Captive, Mobile, Hates Crowding

85% surveyed btwn 24-34

years

16% female (of those counted)

57% ride at least 5 times a

week

100% with a mobile phone (18% with

smartphone, 50% with internet-enabled multimedia phone)

Most common complaint about buses (23%)

Long waits (21%) and Too few buses (20%) were also common

* Potential flock bias

2.7

Happiness

Crowding and Happiness

y = 0.0493x + 3.1012 R² = 0.21825

y = 0.0514x + 2.0214 R² = 0.52836

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

(20) (15) (10) (5) - 5 10 15 20

Happiness

Empty Seats

Empty Full

Significant correlation between crowding and

happiness

Crowded

#27

#36

Determinants of Happiness

Rider Happiness

crowding

slowness

#36 Average Sample

Size %Std Dev Min Max

one-way commute

Std Dev

variability

24

15

9

20

4

0:18

0:16

0:21

0:18

0:12

30%

26%

36%

29%

23%

0:30

0:46

0:30

0:30

0:39

1:42

1:42

1:39

1:42

1:10

1:01

1:02

0:59

1:03

0:53

OVERALL

Inbound

Outbound

9!1!12.4 km

Weekday

Weekend

BUS TRAVEL TIMES

27

36

32

38

41 33

30

23

24

1!2!

3!4!

5!

6!

7!

8!

9!

Home Economics College, Azimpur

Dhaka College, New Market

New Model Degree College, Dhanmondi

Asad Gate, Jatiya Sangsad Bhaban

ASAUB, Agargaon

Agargaon High School, Agargaon

Shewrapara Bus Stand, Shewrapara

Purobi Bus Stand, Section 11

Pallabi Model School, Pallabi

Avg Bus Size

40

0.6 km

2.5 km

3.2 km

5.1 km

6.5 km

8.0 km

11.4 km

12.4 km #36

wi-fi bus stops

BUS CROWDING

BUS ROUTING

Qualitative + Quantitative

Quantitative Only

High

Low

Data Value Data

Collection Dash

Predictable

Unpredictable

Real-Time

Slow-Time

All the Data

Sampled

Ubiquitous Sensing

Flocksourcing

Crowdsourcing

Analog

Dimensions of Data

Collection

Dimensions of Data Itself

Bus Survey

Transport survey on the pedestrian bridge in Mirpur 1, Jan 2012

Marketing 1

Bus Travel Times

#27 Uttara

20 km

1:25 Average

1:47

1:04

*Data based on 42 Rides in March 2012

Bad day 2:07

0:43 Good day

8 am 10 am 6 pm

1:50

Weekend Weekday

(Real-Time) User Services

2

Bus Speed Map Live Bus Location Map

(Real-Time) Operator Services

3

Updated March 2012 Dhaka Bus Dashboard

Responsive City

Planning

4

Bus health Indicators

Rider Happiness

Current Ridership

crowding

marketing slowness

operator profitability

Future Ridership

Affordability of alternatives

1

2

Accessibility

New Market

Uttara

Dhanmondi

Pallabi

Slowness #36

#27

1.3 hours Average one-way

commute time

Azimpur

Uttara

Banani

Dhanmondi

#27 Gazipur 2.5 hours

Accessibility

Most popular commutes

Most painful commute

Happiness by bus company

#27 #36

BRTC 3.6

Suchona 2.8

2.3 VIP 2.3 2.5 Bikolpa

Safety

crowding 3.6

2.8

2.3

BRTC 52 seats per bus

Suchona 48 seats per bus

VIP 39 seats per bus

#27 Bigger buses = happier passengers and more women!

Qualitative + Quantitative

(vs. Only Quantitative)

Real-Time (vs. Slow)

All the Data (vs. Sampled)

Urban data collection techniques

Analog Ubiquitous

Sensing

Crowd Sourcing

Flocksourcing