C. G. Cassandras and Y. Geng

Transcription

1 BUILDING A CYBER-PHYSICAL INFRASTRUCTURE FOR THE SMART CITY: THE CASE OF SMART PARKING C. G. Cassandras and Y. Geng Dvson of Systems Engneerng Dept. of Electrcal and Computer Engneerng Center for Informaton and Systems Engneerng Boston Unversty Chrstos G. Cassandras CODES Lab. - Boston Unversty

2 CYBER-PHYSICAL SYSTEMS INTERNET CYBER Data collecton: relatvely easy PHYSICAL Control: a challenge Chrstos G. Cassandras CISE - CODES Lab. - Boston Unversty

3 SMART CITY INFRASTRUCTURE Prvacy Data collecton Securty TRAFFIC LIGHT CONTROL Geng, Y., and Cassandras, C.G., Traffc Lght Control Usng Infntesmal Perturbaton Analyss, subm. to 51st IEEE Conf. Decson and Control, 2012 Control and Optmzaton Actons Informaton Processng Safety Decson Mang Energy Management THE CASE OF SMART PARKING Chrstos G. Cassandras CISE - CODES Lab. - Boston Unversty

4 SMART PARKING - MOTIVATION 30% of vehcles on the road n the downtowns of major ctes are crusng for a parng spot. It taes the average drver 7.8 mnutes to fnd a parng spot n the downtown core of a major cty. R. Arnott, T.Rave, R.Schob, Allevatng Urban Traffc Congeston Over one year n a small Los Angeles busness dstrct, cars crusng for parng created the equvalent of 38 trps around the world, burnng 47,000 gallons of gasolne and producng 730 tons of carbon doxde. Donald Shoup, The Hgh Cost of Free Parng Chrstos G. Cassandras CISE - CODES Lab. - Boston Unversty

5 SMART PARKING - CONCEPT OPTIMAL PARKING SPOT Mnmze a functon of COST and DISTANCE from A

6 SMART PARKING - CONCEPT DESTINATION OPTIMAL PARKING SPOT

7 GUIDANCE-BASED PARKING DRAWBACKS Drvers: May not fnd a vacant space May mss better space Processng nfo whle drvng Cty: Imbalanced parng utlzaton May create ADDED CONGESTION as multple drvers converge to where a space exsts Searchng for parng Competng for parng

8 SMART PARKING NEW FEATURES System fnds BEST parng space for drver based on PROXIMITY to destnaton + parng COST Space RESERVED guaranteed parng space System contnuously IMPROVES assgned parng space System ensures FAIRNESS n parng space allocaton Parng space UTILIZATION INCREASES Drver maes decsons System maes optmal decsons for drver

9 GUIDANCE-BASED PARKING v SMART PARKING COLLECTING DATA IS NOT SMART, JUST A NECESSARY STEP TO BEING SMART INFO PROCESSING DATA TO MAKE GOOD DECISIONS IS SMART ACTION INFO

10 SMART PARKING IMPLEMENTATION Parng space avalablty detecton Standard sensors e.g., magnetc, cameras Wreless sensor networng Vehcle localzaton GPS System-Drver communcaton Smartphone Vehcle navgaton system Parng reservaton Foldng/Retreatng barrer Red/Green/Yellow lght system

11 PROBLEM FORMULATION Request WAIT Allocaton Fal Allocaton RESERVE 1 2. Departure Allocaton Succeed N

12 DYNAMIC RESOURCE ALLOCATION RESOURCE New request FREE RESERVED BUSY WAIT WAIT RESERVE ut RESERVE TIME STATE x t x t 1 f [ x t, u t, w t] mn J t u t s.t. g x t, u t 0 CONTROL Assgn resource De-assgn resource Change resource prce EVENTS New request Cancel request Resource freed

13 OBJECTIVE FUNCTION Objectve functon at th decson pont: J mn x X j W R j J j Decson varables: 0 x j 1 f f user user s s NOT assgnedto resource assgnedto resource j j User cost functon: J j M j 1 M D D j cost upper bound max proxmty to dest. weght

14 , {0 1},,,, s.t. 1 mn 1 j R W x W m t t j m m j j x x R J J x R x W x j x x J x j j mj m mj n n q j j j j j j j R W j j j W j j j R W, Reservaton Guarantee MIXED INTEGER LINEAR PROBLEM MILP Reservaton Upgrade Farness Satsfed User Cost Unsatsfed User Cost

15 DECISION INTERVAL Tme-drven Event-drven Immedate response X Computatonal cost X Abandonment X Optmal assgnment X

16 SMART PARKING TEST BED Chrstos G. Cassandras CODES Lab. - Boston Unversty

17

18 SIMULATION CASE STUDY On-street parng spaces Off-street parng spaces Ponts of nterest

19 CASE STUDY RESULTS Traffc Traffc Traffc Traffc SP: BU Smart Parng system NG: No gudance status quo G: Parng usng gudance-based systems

20 KEY CONCLUSIONS % hgher parng utlzaton HIGHER REVENUE, LOWER CONGESTION 2. % drvers searchng for parng wanderng < 2% HIGHER REVENUE, LOWER CONGESTION 3. 50% reducton n parng tme under heavy traffc LOWER CONGESTION, LESS FUEL, DRIVER COMFORT

21 IMPLEMENTATION Smart Parng proof-of-concept study mplemented n a small 27 space garage at Boston Unversty durng summer 2011: - Parng request through Phone app. - Smart Parng Allocaton Center SPAC: Server located n CODES Lab SPAC determnes optmal allocaton for request f one exsts and notfes drver through Phone app showng the dentty of reserved spot - Garage gateway: Laptop computer located n garage -Sensor and lght system devce: Custom-bult devce affxed on celng over each parng spot.

22

23 PROJECT TEAM, RECOGNITION TEAM: Yanfeng Geng PhD student, Ted Grunberg Undergrad. Student, Andy Ochs, Mhal Gurevch, Greg Berman BU SOM students 2011 IBM/IEEE Smarter Planet Challenge competton, team won 2nd place prze Best Student Paper Award, Fnalst, 2011 IEEE Mult-Conference on Systems and Control Thrd prze poster on Smart Parng, INFORMS 2011 Northeastern Conference Ongong mplementaton under BU OTD Ignton Award Worng wth Cty of Boston under IBM Award for Combatng Clmate Change Through Smarter Urban Transportaton Polces Geng, Y., and Cassandras, C.G., Dynamc Resource Allocaton n Urban Settngs: A Smart Parng Approach, Proc. of 2011 IEEE Mult-Conference on Systems and Control, Oct Geng, Y., and Cassandras, C.G., A New Smart Parng System Based on Optmal Resource Allocaton and Reservatons, Proc. of 14th IEEE Intellgent Transportaton Systems Conf., pp , Nov

24