ytong general brochure1 · H=J>=;LAGF J=KMDLK >JGE 9LL=FLAGF LG
C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective...
Transcript of C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective...
![Page 1: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/1.jpg)
C. G. Cassandras and Y. Geng
Division of Systems Engineering
Dept. of Electrical and Computer Engineering
Center for Information and Systems Engineering
Boston University
Christos G. Cassandras CODES Lab. - Boston University
BUILDING A CYBER-PHYSICAL
INFRASTRUCTURE
FOR THE SMART CITY:
THE CASE OF “SMART PARKING”
![Page 2: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/2.jpg)
CYBER-PHYSICAL SYSTEMS
Christos G. Cassandras CISE - CODES Lab. - Boston University
INTERNET
CYBER
PHYSICAL
Data collection:
relatively easy…
Control:
a challenge…
![Page 3: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/3.jpg)
“SMART CITY” INFRASTRUCTURE
Christos G. Cassandras CISE - CODES Lab. - Boston University
Decision Making
Data collection
Energy
Management
Safety
Security
Control and
Optimization
Actions
Information
Processing
Privacy
THE CASE OF
“SMART PARKING”
TRAFFIC LIGHT
CONTROL
Geng, Y., and Cassandras, C.G.,
“Traffic Light Control Using
Infinitesimal Perturbation
Analysis”, subm. to 51st IEEE
Conf. Decision and Control, 2012
![Page 4: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/4.jpg)
30% of vehicles on the road in the downtowns of major cities are
cruising for a parking spot. It takes the average driver 7.8 minutes to
find a parking spot in the downtown core of a major city.
R. Arnott, T.Rave, R.Schob, Alleviating Urban Traffic Congestion. 2005
Over one year in a small Los Angeles business
district, cars cruising for parking created the
equivalent of 38 trips around the world, burning
47,000 gallons of gasoline and producing 730
tons of carbon dioxide.
Donald Shoup, The High Cost of Free Parking. 2005
“SMART PARKING” - MOTIVATION
Christos G. Cassandras CISE - CODES Lab. - Boston University
![Page 5: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/5.jpg)
1)Autonomous Reconfigurability via Robust Optimization (Paschalidis, Gong)
2)Autonomous Reconfigurability via Perturbation Analysis and Concurrent Estimation (Cassandras, Gong)
3)A Game-theoretic Approach to Cross-level Reconfigurability (Bestavros, Gong)
4)Sensor Network Adaptivity/Reconfigurability (All)
Node Localization (Paschalidis, Gao)
1)Event-Driven Reconfigurable Information Acquisition System (Gao, Gong)
2)Anomaly Detection (Paschalidis, Gong, Cassandras)
3)Test Beds (All)
4)Education component (All)
5)Final Report (All)
1)Autonomous Reconfigurability via Robust Optimization (Paschalidis, Gong)
2)Autonomous Reconfigurability via Perturbation Analysis and Concurrent Estimation (Cassandras, Gong)
3)A Game-theoretic Approach to Cross-level Reconfigurability (Bestavros, Gong)
4)Sensor Network Adaptivity/Reconfigurability (All)
Node Localization (Paschalidis, Gao)
1)Event-Driven Reconfigurable Information Acquisition System (Gao, Gong)
2)Anomaly Detection (Paschalidis, Gong, Cassandras)
3)Test Beds (All)
4)Education component (All)
5)Final Report (All)
OPTIMAL
PARKING SPOT
Minimize a
function of
COST and
DISTANCE from
A
“SMART PARKING” - CONCEPT
![Page 6: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/6.jpg)
1)Autonomous Reconfigurability via Robust Optimization (Paschalidis, Gong)
2)Autonomous Reconfigurability via Perturbation Analysis and Concurrent Estimation (Cassandras, Gong)
3)A Game-theoretic Approach to Cross-level Reconfigurability (Bestavros, Gong)
4)Sensor Network Adaptivity/Reconfigurability (All)
Node Localization (Paschalidis, Gao)
1)Event-Driven Reconfigurable Information Acquisition System (Gao, Gong)
2)Anomaly Detection (Paschalidis, Gong, Cassandras)
3)Test Beds (All)
4)Education component (All)
5)Final Report (All)
1)Autonomous Reconfigurability via Robust Optimization (Paschalidis, Gong)
2)Autonomous Reconfigurability via Perturbation Analysis and Concurrent Estimation (Cassandras, Gong)
3)A Game-theoretic Approach to Cross-level Reconfigurability (Bestavros, Gong)
4)Sensor Network Adaptivity/Reconfigurability (All)
Node Localization (Paschalidis, Gao)
1)Event-Driven Reconfigurable Information Acquisition System (Gao, Gong)
2)Anomaly Detection (Paschalidis, Gong, Cassandras)
3)Test Beds (All)
4)Education component (All)
5)Final Report (All)
DESTINATION
OPTIMAL
PARKING SPOT
“SMART PARKING” - CONCEPT
![Page 7: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/7.jpg)
GUIDANCE-BASED PARKING – DRAWBACKS…
Drivers:
• May not find a vacant space
• May miss better space
• Processing info while driving
City:
• Imbalanced parking utilization
• May create ADDED CONGESTION
(as multiple drivers converge
to where a space exists)
Searching for parking Competing for parking
![Page 8: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/8.jpg)
SMART PARKING – NEW FEATURES
Driver makes decisions System makes optimal
decisions for driver
• System finds BEST parking space for driver
(based on PROXIMITY to destination + parking COST)
• Space RESERVED guaranteed parking space
• System continuously IMPROVES assigned parking space
• System ensures FAIRNESS in parking space allocation
• Parking space UTILIZATION INCREASES
![Page 9: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/9.jpg)
GUIDANCE-BASED PARKING v “SMART PARKING”
COLLECTING DATA IS NOT “SMART”,
JUST A NECESSARY STEP TO BEING
“SMART”
PROCESSING DATA TO MAKE
GOOD DECISIONS IS “SMART”
INFO
INFO ACTION
![Page 10: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/10.jpg)
• Parking space availability
detection
• Vehicle localization
• System-Driver
communication
• Parking reservation
SMART PARKING – IMPLEMENTATION
Standard sensors
(e.g., magnetic, cameras)
Wireless sensor networking
GPS
Smartphone
Vehicle navigation system
Folding/Retreating barrier
Red/Green/Yellow light system
![Page 11: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/11.jpg)
Allocation
Fail
Allocation
Succeed
Allocation
WAIT
RESERVE
Request
Departure
. . .
1
2
N
PROBLEM FORMULATION
![Page 12: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/12.jpg)
DYNAMIC RESOURCE ALLOCATION
FREE RESERVED BUSY
)](),(),([)1( twttft uxx
WAIT
RESERVE u(t)
WAIT
RESERVE
New
request
TIME
RESOURCE
0))(),(( s.t.
)(min)(
tt
tJi
it
uxg
u
)( STATE tx
• Assign resource
• De-assign resource
• Change resource price
• …
CONTROL
• New request
• Cancel request
• Resource freed
• …
EVENTS
![Page 13: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/13.jpg)
)(min)()()( )(
kJxkJ ij
kRkWi kj
ijX
i
Objective function
at kth decision point:
jresourcetoassignedisiuserif
jresourcetoassignedNOTisiuserifxij
1
0
Decision variables:
User cost function:
i
ij
i
i
ij
iijD
D
M
kMJ )1(
)(
weight cost upper bound
max proximity to dest.
OBJECTIVE FUNCTION
![Page 14: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/14.jpg)
)(),()(}10{
)(,),(|
,)(|),(0)(
)()()(
)(1
)(1
)(1
s.t.
)1()(min
)(
)1(
)(
)(
)(
)()(
)()()()()(
kjkRkWix
kWmttkjmm
kjiikjxx
kRikJkJx
kRix
kWix
kjx
xkJx
iij
ijmjm
imj
kn
in
kiqij
kj
ij
kj
ij
kj
ij
kRkWi
ij
kj
ij
kWikj
ijij
kRkWi
i
i
i
i
i
ii
,
Reservation Guarantee
MIXED INTEGER LINEAR PROBLEM (MILP)
Reservation Upgrade
Fairness
Satisfied User Cost Unsatisfied User Cost
![Page 15: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/15.jpg)
Time-driven Event-driven
Immediate response
Computational cost
Abandonment
Optimal assignment
X
X
X
X
DECISION INTERVAL
![Page 16: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/16.jpg)
“SMART PARKING” TEST BED
Christos G. Cassandras CODES Lab. - Boston University
![Page 17: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/17.jpg)
![Page 18: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/18.jpg)
On-street parking spaces Off-street parking spaces
Points of interest
SIMULATION CASE STUDY
![Page 19: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/19.jpg)
CASE STUDY RESULTS
Traffic
Traffic
Traffic
Traffic
SP: BU Smart Parking system G: Parking using guidance-based systems
NG: No guidance (status quo)
![Page 20: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/20.jpg)
1. 10-20% higher parking utilization
HIGHER REVENUE,
LOWER CONGESTION
2. % drivers searching for parking (wandering) < 2%
HIGHER REVENUE,
LOWER CONGESTION
3. 50% reduction in parking time under heavy traffic
LOWER CONGESTION,
LESS FUEL,
DRIVER COMFORT
KEY CONCLUSIONS
![Page 21: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/21.jpg)
“Smart Parking” proof-of-concept study implemented in a small (27
space) garage at Boston University during summer 2011:
- Smart Parking Allocation Center (SPAC): Server located in CODES Lab
SPAC determines optimal allocation for request (if one exists) and notifies
driver through iPhone app showing the identity of reserved spot
- Garage gateway: Laptop computer located in garage
-Sensor and light system device: Custom-built
device affixed on ceiling over each parking spot.
- Parking request through iPhone app.
IMPLEMENTATION
![Page 22: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/22.jpg)
http://www.bu.edu/buniverse/view/?v=1zqb6NnD
http://www.necn.com/09/23/11/JoeBattParkingapp/landing_scitech.html?blockID=566574&feedID=4213
![Page 23: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/23.jpg)
TEAM: Yanfeng Geng (PhD student), Ted Grunberg (Undergrad. Student),
Andy Ochs, Mikhail Gurevich, Greg Berman (BU SOM students)
• 2011 IBM/IEEE Smarter Planet Challenge competition, team won 2nd
place prize
• Best Student Paper Award, Finalist, 2011 IEEE Multi-Conference
on Systems and Control
• Third prize poster on “Smart Parking”, INFORMS 2011 Northeastern
Conference
• Ongoing implementation under BU OTD “Ignition Award” • Working with City of Boston under IBM Award for “Combating Climate
Change Through Smarter Urban Transportation Policies”
PROJECT TEAM, RECOGNITION
• Geng, Y., and Cassandras, C.G., “Dynamic Resource Allocation in Urban Settings: A “Smart Parking”
Approach”, Proc. of 2011 IEEE Multi-Conference on Systems and Control, Oct. 2011.
• Geng, Y., and Cassandras, C.G., “A New “Smart Parking” System Based on Optimal Resource Allocation
and Reservations”, Proc. of 14th IEEE Intelligent Transportation Systems Conf., pp. 979-984, Nov. 2011.
![Page 24: C. G. Cassandras and Y. Geng - bu.edu · J k x J ij k i W k R k j k ij X i ¦ ¦ : Objective function at kth decision point: ¯ ® if user i is assigned to resource j if user i is](https://reader035.fdocuments.in/reader035/viewer/2022071019/5fd33c1a11e0241b732ac1cc/html5/thumbnails/24.jpg)
http://www.bu.edu/buniverse/view/?v=1zqb6NnD