Date post: | 07-Mar-2016 |
Category: |
Documents |
Upload: | albert-ching |
View: | 220 times |
Download: | 2 times |
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