the first
A User-Flocksourced Bus Intelligence System in Dhaka ---
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)
A!
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? Yes, in a big way.
research question(s)
B!
If not all data in a city can be collected by flocks, can a sampled set be useful, especially if certain behaviors are predictable? Yes, less data can become big data.
theory
context
experiment
results
1
2
3
4
ITERATIVE CITY
MOBILE MOBILITY
FLOCKSOURCING
1000 SURVEYS
future
URBAN LUNCHPAD
theory
1 ITERATIVE CITY
theory
1 ITERATIVE CITY
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
1
WHICH CITIES WILL BENEFIT?
1
context
2 MOBILE MOBILITY
DHAKA
18 Million People 100,000 Cars <1%
JAKARTA
9 Million People 9 Million Two-Wheelers 3 Million Cars >100%
Asia Rest of the World
inflection of $5,000 per capita GDP
60.0
Sandra and Archaya (2007) motorization
80.0
Cars, trucks and person per 100 persons
Motorization
100.0
United States
Japan
40.0
20.0
Singapore
Indonesia Barter â&#x20AC;&#x153;lock-inâ&#x20AC;? line of 10% car ownership
Hong Kong -
(20.0)
Income per person (GDP per capita, $USD, inflation adjusted) 100
Bangladesh
India
1,000
China
10,000
100,000
Income
Can Owning a Cell Phone Reduce the Desire to Own a Car?
Mobile rickshaw wallah in India
Users
1
Operators
2 (Real-Time)
Marketing
User Services
Cars = aspiration
Information can improve accessibility to, comfort and safety of shared vehicles
3
(Real-Time) Operator Services Information can improve efficiency, management and profitability of shared fleets
Regulators
4 Responsive City Planning
Information can help monitor and evaluate city performance in a more precise and timely manner than ever before
GO-Jek Dial-a-Motorcycle Transport in Jakarta, August 2011
Fazilka Dial-a-Rickshaw in Punjab, August 2011
entrepreneurs
Are these business sustainable + scalable?
Constellation of Mobile-Driven Mobility Experiments
Unsustainable Navigation Congestion
On-Demand On-Demand On-Demand Real Time Arrival Info Safety Alerts Fare-Tracking
Tracking Vehicle-Security
Sustainable Real Time On-Demand Arrival Info
Bicycle Sharing
Bus Delays
Singapore Kuala Lumpur
Bangkok
Jakarta Delhi Bangalore Fazilka Dhaka
Can an outside institution accelerate experimentation?
August 2011
experiment
3 FLOCKSOURCING Guided crowdsourcing
Real-time urban data collection techniques
UBIQUITOUS SENSING
CROWDSOURCING
All the data, all the time
Some data for lots of disparate times and places
Sensors
Crowds + Sensors
Privacy Closed Expensive Data processing Only objective metrics
Gathering sufficient and relevant data
Predictability of mobility (Song, Qu, Blumm, Barabasi 2010)
Real-time urban data collection techniques
FLOCKSOURCING
Lots of data for a specific time and place Flocks + Sensors Organizing the flock Flock bias
Flocksourcing Workflow Sensors
Hardware
Platform
Involuntary Tracking
Unsmartphones
None
Incentivized Volunteers
Organized Flock
Organized Vehicles
Smartphones
Tablets
PC
Android
iPhone
Web
Software/App
Connectivity
Bluetooth
Datastorage Dataverification &analysis Visualization
main bottlenecks
MIT App Inventor
Excel
Cell network
Mobile data
Local
Cloud
Statistical Packages
Visualization APIs
Wi-Fi
Machine learning
â&#x20AC;&#x153;Launch and iterateâ&#x20AC;? co-development
Survey
Passenger Count
Bus Details
Cost Structure Sensors
$10-$15
per person per day
Hardware
$175
and rapidly declining
Software/App
Free
Connectivity
$4
per 1 GB
Datastorage
Free
Experimental Design
Dhaka
Boston
Flock size & nature
Flock size & nature
8 paid volunteers
3-8 unpaid volunteers
($10 per person per day) Organized by Kewkradong Bangladesh
($30 per data plan)
Target buses
Target buses lines
36 & 27 Lines (10 km each)
Any
Data collection target
Data collection target
100 surveys 120 one-way rides the worldâ&#x20AC;&#x2122;s first
Flocksourcing
experiment
None
Crowdsourcing
Metrics Quantitative
Qualitative
Bus Details
Survey
Bus Number Bus Destination Bus Company No. of Seats Speed
Location Time Crowding
Passenger Count Female Passenger Count
Gender Age Home Location Work Location One-Way Commute Income Phone Ownership Rider Satisfaction Biggest Complaint Riding Frequency
*Survey data linked to bus data
results
4 1000 SURVEYS
Data Collection Dash
Kb16
Individual Flock Traces
Kb10
Kb2
Kb14
Kb20
Kb7
Kb8
Kb13
research question(s)
A!
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?
research question(s)
B!
If not all data in a city can be collected by flocks, can a sampled set be useful, especially if certain behaviors are predictable?
Ubiquitous Sensing High
Need More Data
Data Value
Need Less Data
Low Low
Crowdsourcing
High Predictability Dimensions of Data Itself
Dimensions of Data Collection
1
BUS CROWDING
2
BUS TRAVEL TIMES
3 BUS ROUTES
passenger count Average
Sample Size
Std Dev
%Std Dev
Min
Max
9!
27
15
15
58%
5
52
8!
36
85
9
25%
11
50
7!
32
62
11
35%
11
50
6!
38
64
12
32%
11
52
5!
41
47
11
27%
14
49
4!
33
64
14
41%
9
54
3!
30
34
12
39%
9
47
2!
23
32
16
68%
5
51
BUS 1! CROWDING
24
32
15
64%
2
51
variability
#36
empty seats Average
9! 8! 7! 6! 5! 4! 3! 2!
+16! +17! +10! +7!
(0)! +2! +8! +4!
BUS +13! 1! CROWDING
7!
8!
9!
10!
am
pm
11!
12!
1!
2!
3!
4!
5!
6!
#36
one-way commute
OVERALL 12.4 km
1!
9!
Inbound
1:01
0:52
0:59
0:49
Average 7:00
8:00
9:00
1:02
Outbound
Weekday
0:46
0:59
8:00
9:00
0:46
0:58
Average 7:00
8:00
1:03
10:00
10:00
11:00
12:00
9:00
0:52
0:59
0:49
Average 7:00
8:00
9:00
10:00 1:22
10:00
13:00
12:00
13:00
0:54
1:01
14:00
15:00
1:32
16:00
8:00
9:00
10:00
17:00
18:00
19:00
20:00
1:32
0:56
1:01
14:00
15:00
16:00
17:00
18:00
19:00
20:00
15:00
16:00
17:00
18:00
19:00
20:00
1:22 0:41 11:00
12:00
1:22
0:55
11:00
13:00
12:00
13:00
0:52
Weekend Average 7:00
11:00
0:49
0:53
BUS TRAVEL TIMES
1:22
0:54
1:42
Average 7:00 0:59
1:22
#36
11:00
14:00
1:32
0:55
1:01
14:00
15:00
16:00
17:00
18:00
19:00
20:00
15:00
16:00
17:00
18:00
19:00
20:00
0:52
12:00
13:00
14:00
BUS ROUTING
Ubiquitous Sensing
High
BUS TRAVEL TIMES
BUS CROWDING + Machine Learning
Data Value
BUS ROUTING
Low
Crowdsourcing
Low
High Predictability
BUS RIDERSHIP BUS TRAVEL TIMES
BUS CROWDING
BUS SATISFACTION
BUS ROUTING
Self-organizing flock
future 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. URBAN ---------LUNCHPAD
launchpad
PUBLIC INFOSTRUCTURE 30 buses (position, speed)
BEST BUS MAP IN THE WORLD flock of 30, 15 days (counts)
public public public public
50 buses (position, speed)
flock of 15, 5 days (crowding)
public
Who will build?
flock of 25, 10 days (satisfaction)
OUR FIRST PRODUCT the cheapest and easiest --A BUS INTELLIGENCE
-------SERVICE IN DHAKA the world
1!
2
TECHNOLOGY + YOUR FLEET
TECHNOLOGY + OUR FLOCKS
Ongoing data collection
One-time data collection
CUSTOMERS
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*
$50*
per seat per month
per flock member per day
Bus tracking hardware retails in US for $8-$20K per bus
Retails to less than $3 per survey using pilot results
*50% discount if data is made open to public for mash-up
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
Mahalo!
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
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
appendix
A
REVENUE POTENTIAL (FLEET ONLY)
$50 per seat per month
x
9,000 buses in Dhaka
5% 10% 25% 50% 75% 100%
$270K $540K $1.4M $2.7M $4.1M $5.4M
penetration rate
annual revenue
Current Bus Riders in Dhaka 16% female (of those counted)
100% with a mobile phone (18% with smartphone, 50% with internet-enabled multimedia phone)
Young, Male, Captive, Mobile, Hates Crowding 85% surveyed btwn 24-34 years
57% ride at least 5 times a week
* Potential flock bias
Happiness
2.7
Most common complaint about buses (23%)
Long waits (21%) and Too few buses (20%) were also common
Determinants of Happiness crowding Rider Happiness
slowness
Crowding and Happiness
Happiness 5.0 4.5
y = 0.0493x + 3.1012 R² = 0.21825
Significant correlation between crowding and happiness
4.0 3.5 3.0
#27
2.5 2.0
#36 (20)
Crowded
(15)
(10)
(5)
y = 0.0514x + 2.0214 R² = 0.52836
1.5 1.0
-
Full
5
10
15
Empty Seats
20
Empty
one-way commute Average OVERALL
Sample Size
Std Dev
%Std Dev
Min
Max
1:01
24
0:18
30%
0:30
1:42
Inbound
1:02
15
0:16
26%
0:46
1:42
Outbound
0:59
9
0:21
36%
0:30
1:39
Weekday
1:03
20
0:18
29%
0:30
1:42
0:53
4
0:12
23%
0:39
1:10
12.4 km
1!
9!
Weekend BUS TRAVEL TIMES
variability
#36
wi-fi bus stops 12.4 km
9!
24
11.4 km
23
8!
8.0 km
Purobi Bus Stand, Section 11
7!
30
Shewrapara Bus Stand, Shewrapara
6!
6.5 km 5.1 km
2.5 km
0.6 km
ASAUB, Agargaon
4!
38
40
BUS CROWDING
Agargaon High School, Agargaon
5!
3341
3.2 km Avg Bus Size
#36
Pallabi Model School, Pallabi
32
Asad Gate, Jatiya Sangsad Bhaban
3!
36 27
New Model Degree College, Dhanmondi
2!
Dhaka College, New Market
1!
Home Economics College, Azimpur
BUS ROUTING
Dimensions of Data Itself
Data Value
High
Data Collection Predictable Dash
Dimensions of Data Collection
Qualitative + Quantitative
Real-Time
All the Data
Ubiquitous
Flocksou
Crowdsour Analog
Low
Unpredictable
Quantitative Only
Slow-Time
Sampled
Bus Survey
1
Marketing
Transport survey on the pedestrian bridge in Mirpur 1, Jan 2012
Bus Travel Times
Weekday 2:07
Bad day
Weekend 1:50
Average
1:04 0:43
Uttara 20 km
1:47
1:25
#27
8 am 10 am
Good day
6 pm
2 (Real-Time)
User Services *Data based on 42 Rides in March 2012
Bus Speed Map Live Bus Location Map
3
(Real-Time) Operator Services
4 Responsive City Planning
Dhaka Bus Dashboard
Updated March 2012
Bus health Indicators
Accessibility
2
Current Ridership
marketing slowness
1
Rider Happiness
Future Ridership
crowding Affordability of alternatives
operator profitability
Uttara #27
#36 Pallabi
Dhanmondi
New Market
Slowness
Gazipur
Accessibility
2.5 hours Most painful commute
Uttara
Most popular commutes
Banani
Dhanmondi
1.3 hours
Average one-way commute time
Azimpur
#27
Happiness by bus company #27
3.6
BRTC
2.8
Suchona
2.3
VIP
#36
2.5 2.3
Bikolpa Safety
crowding
#27
Bigger buses = happier passengers and more women!
BRTC
52 seats per bus
2.8
Suchona
48 seats per bus
2.3
VIP
39 seats per bus
3.6
Urban data collection techniques Qualitative + Quantitative (vs. Only Quantitative)
Flocksourcing
Analog Crowd Sourcing
Real-Time (vs. Slow)
Ubiquitous Sensing All the Data (vs. Sampled)