Modeling Cruising for Parking Itzhak Benenson bennya@post.tau.ac.il - - PowerPoint PPT Presentation

▶
modeling cruising for parking
SMART_READER_LITE
LIVE PREVIEW

Modeling Cruising for Parking Itzhak Benenson bennya@post.tau.ac.il - - PowerPoint PPT Presentation

Jun 30, 2023 166 likes •746 views

Modeling Cruising for Parking Itzhak Benenson bennya@post.tau.ac.il http://www.tau.ac.il/~bennya/ http://geosimlab.tau.ac.il/ Geosimulation and Spatial Analysis Lab, Department of Geography and Human Environment, Tel Aviv University, Israel

slide-1
SLIDE 1

Modeling Cruising for Parking

Itzhak Benenson

bennya@post.tau.ac.il http://www.tau.ac.il/~bennya/ http://geosimlab.tau.ac.il/

Geosimulation and Spatial Analysis Lab, Department of Geography and Human Environment, Tel Aviv University, Israel

MTS Summer School 2015

slide-2
SLIDE 2

Parking is a long lasting urban blight

 Cruising for parking is the last leg of the car-based trip  Cruising is linked to many urban externalities:

  • Congestion, air pollution and noise, loss of space, social inequity

 Parking has mostly been left to engineers to solve supply

and operational issues by building lots

2 MTS Summer School 2015

slide-3
SLIDE 3

BASIC MODEL OF CRUISING FOR PARKING

MTS Summer School 2015 3

slide-4
SLIDE 4

MTS Summer School 2015

Basic model of cruising for parking: DEFINITIONS

Parameters Arrivals a cars/min Departure rate d/min Maximal search time t min Total parking places R

t  t + 1

  • Fraction d of the parking cars depart
  • cars that cruise longer than t min

depart

  • a cars arrive
  • vacant parking places are occupied

by the cruising cars Initial conditions: All parking places are occupied

4

slide-5
SLIDE 5

MTS Summer School 2015

  • State vector of the system, M(t) = <M0(t), M1(t), … Mt-1(t)>,

Mm(t) is the numbers of cars cruising for m minutes

  • Total number of cruising cars N(t) = M0(t) + M1(t) + … + Mt-1(t)
  • O(t) – Number of occupied parking places
  • F(t) - Cars that failed to find a parking place
  • p(t) - probability to park

p(t) = min{1, [number of free places]/[number of cruising cars]} p(t) = min{1, (R – (1 - d)*O(t))/N(t)}

Basic model of cruising for parking: DEFINITIONS

5

slide-6
SLIDE 6

MTS Summer School 2015

  • 1. Two auxiliary equation to estimate parameters at the start of a next minute

N(t) = M0(t) + M1(t) + … + Mt-2(t) + Mt-1(t) p(t) = min{1, (R – (1 – d)*O(t))/N(t)}

  • 2. The dynamics of M(t), O(t) and F(t) are described by the recurrence equations:

M0(t + 1) = a M1(t + 1) = M0(t)*[1 – p(t)] … Mt-1(t + 1) = Mt-2(t)*[1 – p(t)] F(t + 1) = F(t) + Mt-1(t)*[1 – p(t)] O(t + 1) = min{(1 – d)*O(t) + N(t), R}

*N. Levy, K. Martens, I. Benenson, 2013, Transportmetrica A, 9 (9), 773–797

Basic model of cruising for parking: EQUATIONS OF SYSTEM DYNAMICS*

6

slide-7
SLIDE 7

MTS Summer School 2015

Basic model of cruising for parking: SOLUTION

Arrivals: 4/min

  • Dep. rate: 0.05

Capacity: 100 Arrivals: 6/min

  • Dep. rate: 0.05

Capacity: 100 Arrivals: 10/min

  • Dep. rate: 0.05

Capacity: 100

7

slide-8
SLIDE 8

MTS Summer School 2015

Basic model of cruising for parking: EXTENSIONS

Close to destination: More attractive, High parking prices Far from destination: Less attractive, Low parking prices

Drivers’ reaction to the lack parking places or to the high price of parking

  • Avoid areas with no parking:

Number of arriving cars depends on the density of cruising cars or average cruising time

  • Avoid areas of expensive parking:

Number of arriving cars or maximal cruising time depends on prices

  • Park further from the destination:

Tradeoff between price of the parking place and distance to destination

8

slide-9
SLIDE 9

MTS Summer School 2015

Basic model of cruising for parking: EXTENSION

Max Arrivals = 10

k = 0.01

9

slide-10
SLIDE 10

MTS Summer School 2015

Basic model of cruising for parking: EXTENSION

Max Arrivals = 10

k = 0.2

10

slide-11
SLIDE 11

MTS Summer School 2015

Basic model of cruising for parking: EXTENSION

Max Arrivals = 10

k = 0.9

Three types of parking dynamics

  • Monotonous convergence
  • Non-monotonous convergence
  • Steady cycles

11

slide-12
SLIDE 12

MTS Summer School 2015

  • Parking search is either

easy or takes relatively long time

Basic model of cruising for parking: CONCLUSIONS

  • Drivers’ reaction to cruising or regulator’s measures can decrease the

number of cruising cars and parking failures. But, cruising will always be either almost zero or essential

  • If drivers’ reaction to cruising is weak or intermediate then the system

stabilizes, maybe non-monotonously. Strong feedback cause fluctuations.

12

slide-13
SLIDE 13

MTS Summer School 2015

Local aspects (driver vs destination)

  • Parking supply is defined by many factors: type of

parking (on-street/lot), price, distance to destination, parking control

  • Parking demand, parking supply, and drivers’ behavior

are all essentially heterogeneous, in space and in time

Global aspects (urban parking policy)

  • In a long term, drivers consider parking as only one

component of a trip

  • Parking supply is defined by the urban land-use policy

Parking reality: SPATIO-TEMPORAL HETEROGENEITY

Parking planning (future transportation)

  • What will be parking demand in the future?
  • What is parking demand of the autonomous cars?

13

slide-14
SLIDE 14

Parking spatial pattern Drivers’ parking behavior Parking demand and supply Parking dynamics in space and time Parking management and policy assessment

14 MTS Summer School 2015

WE STUDY THE CURRENT STATE WE AIM AT FORECASTING

COMPONENTS OF PARKING REALITY

slide-15
SLIDE 15

15 MTS Summer School 2015

Parking demand and supply Parking spatial pattern

slide-16
SLIDE 16

URBAN GIS, AERIAL PHOTOS

PARKING DEMAND AND SUPPLY*

GIS + Aerial photos + Population Census

DEMAND Night: Number of households multiplied car ownership rate Day: Office area/20 or proportional to Shops’ turnover SUPPLY Curb: Length of streets /5 minus prohibited places Lots: Lot floor area /12 (5 m – length of a car, 12m2 = 8m2 car + 4m2 pass

16 MTS Summer School 2015

Parking demand and supply

Parking spatial pattern

*Levy, N., Benenson, I, 2015, Journal of Transport Geography, 46, 220–231

slide-17
SLIDE 17

MTS Summer School 2015

PARKING DEMAND AND SUPPLY*

Parking turnover: Field surveys

Residents Visitors Average occupancy (weekdays) STD Average occupancy (weekdays) STD 61.8% 0.94% 17.4% 1.77%

Parking demand and supply

Parking spatial pattern

For a certain day of the week and hour of a day, parameters of the parking system are stable

*N. Levy, K. Martens, I. Benenson, 2013, Transportmetrica A, 9 (9), 773–797

17

slide-18
SLIDE 18

Parking demand and supply

18 MTS Summer School 2015

PARKING PATTERNS*

Destination-parking place distance: Field surveys

Parking spatial pattern

The distance between the parking place and the destination Estimated based on the data

  • f owners’ addresses

Municipality GIS, Remote Sensing data, population census and field surveys provide reliable estimates of parking spatial patterns, demand, supply, and turnover at high spatio-temporal resolution

*Levy, N., Benenson, I, 2015, Journal of Transport Geography, 46, 220–231

slide-19
SLIDE 19

19 MTS Summer School 2015

Drivers’ parking behavior

slide-20
SLIDE 20

20 MTS Summer School 2015

Drivers’ parking behavior

DRIVERS’ PARKING SEARCH BEHAVIOR

GPS data logging, GIS analysis, interviews with drivers

5000 10000 15000 20000 25000 30000 20 40 60 80 100 120

Disstance from home Speed (km/h)

Car speed versus distance to parking

20 40 60 80 100 120 15:40:35 15:42:09 15:42:57 15:43:45 15:44:33 15:45:21 15:46:09 15:46:57 15:47:45 15:48:33 15:49:21 15:50:09 15:50:57 15:51:45 15:52:33 15:53:21 15:54:09 15:54:57 15:55:45 15:56:33 15:57:21 15:58:09 16:00:07 16:01:42 16:02:36 16:04:23 16:05:49 16:07:34

Car speed during the trip

Driver’s speed during parking search is 12-16 km/h (3 - 4 m/s).

slide-21
SLIDE 21

Revealed/Stated preferences: Parking Lot Survey

  • Survey of 100 random drivers

immediately after they parked in private and municipal parking lots in the CBD area of Tel Aviv

  • Aim: understand the drivers’ parking

behaviour after long term adaptation to conditions and prices.

Private Muni Private Muni

21 MTS Summer School 2015

Drivers’ parking behavior

slide-22
SLIDE 22

Factor Category N Visitors Residents Sig. Parking type choice Private

43 60.0% 20.5%

2=15.6 p<0.001 Municipal

58 40.0% 79.5%

Closeness to destination At destination

49 47.4% 50.0%

2=0.2 p>0.1 Up to 5 min walk

34 33.3% 34.1%

More than 5 min walk

18 15.9% 15.9%

Cruising Yes

35 37.2% 35.2%

2=0.04 p>0.1 No

62 62.8% 64.8%

Parking duration (min) mean

95 193 121

t=-2.4; p=0.02 Price (ILS/hr)

95 21.0 9.4

t=3.2; p=0.002 Willingness to pay (ILS/hr)

95 13.9 5.8

t=2.7; p=0.008

Two distinct groups – residents and visitors…but with similar behaviors

~2/3 of drivers do not consider cruising a viable parking choice

General conclusion: Majority of drivers exhibit post-adaptive

  • behaviours. How do they adapt to parking conditions?

Parking lot survey: RESULTS

22 MTS Summer School 2015

Drivers’ parking behavior

slide-23
SLIDE 23

ParkGame: THE IDEA and IMPLEMENTATION

Aim: Understand cruising behaviour and the choice between curb and lot parking based on simulated cruising experience

23 MTS Summer School 2015

Drivers’ parking behavior

slide-24
SLIDE 24
  • User experience with

real occupancy and turnover rates

  • Adjustable game

duration, speeds (car/walk), prices, penalties.

  • Given GIS road

network layers, adjustable to any city

  • Imitates driver’s

limitations (possible view ahead = 5 cars)

24 MTS Summer School 2015

ParkGame: DESIGN

Drivers’ parking behavior

slide-25
SLIDE 25

Driver’s position and speed are recorded every second of the game

Time SegmentID DistanceFro Free Occupied PType Cruise Con

  • Dist. to Des

Speed 240 1065280-1065283 3 0 OFF 428 14.4 239 1065280-1065283 3 0 OFF 428 14.4 238 1065280-1065283 3 0 OFF 428 14.4 237 1065280-1065283 3 0 OFF 428 14.4 236 1065280-1065283 3 0 OFF 428 14.4 235 1065280-1065283 1.28 3 0 OFF 429 14.4 234 1065280-1065283 2.56 3 0 OFF 430 14.4 233 1065280-1065283 3.52 3 0 OFF 431 14.4 232 1065280-10652834.8 3 0 OFF 433 14.6 231 1065280-1065283 6.13 3 0 OFF 434 15.8

ParkGame: OUTPUT

25 MTS Summer School 2015

Drivers’ parking behavior

slide-26
SLIDE 26

ParkGame is freely downloadable

You can start playing and collecting data yourselves

http://www.inplanning.eu/en/news/new-inplanning-simulation-parkgame/

26 MTS Summer School 2015

Drivers’ parking behavior

slide-27
SLIDE 27

Area 3 N=28 Cruising Time (sec) Walk time (sec) % Lot Parked 93% 247.5 145.4 14% 95% 210.9 84.0 39% 98% 247.2 122.5 57% p 0.003 >0.1 0.002

Area 3

44 m 82 m

27 MTS Summer School 2015

Origin and destinations are established in a way that typical time of driving between them is 3.5 minutes Goal of a player: To be in time at the meeting that starts in 4 minutes (+ up to 2 min delay).

  • Initial budget: 20 ILS,
  • Curb parking: 5 ILS,
  • Lot parking: 15 ILS

93% = 1/15 free 95% = 1/20 free 98% = 1/50 free

ParkGame: reaction to occupation rate

 Rise in occupancy pushes drivers to park in lots and increases

cruising time… but the relative location of lot to destination influences behaviour.

 If a lot is close by - drivers will cruise more time as they

perceive the lot as “fail safe” option (myopic behaviour)

 If a lot is further away – drivers cruise less time because the

probability of being late increases…but lot is less attractive because of distance (more rational).

Drivers’ parking behavior

slide-28
SLIDE 28

Occupancy, %

  • High price ratio and medium-high
  • ccupancy rates require drivers

to think. Otherwise the parking problem is simple to solve

Perception

100 98 95 93

  • Drivers’ myopic perception

underweights the real occupancy rate related to the number of

  • bserved free spaces en-route.

Lot/Curb parking price ratio

1 2 3 4 5 6 7 8

1.00 0.95 0.90 0.85

Average occupation

rate 0.93

Search for curb parking

Look for curb parking

  • n the way to the lot

0.98

28 MTS Summer School 2015

More experiments required with different pricing polices

Towards conceptual model of cruising for parking

Drivers’ parking behavior

slide-29
SLIDE 29

Parking dynamics in space-time

29 MTS Summer School 2015

Parking management and policy assessment

slide-30
SLIDE 30

PARKFIT: Approximating parking dynamics in space

30 MTS Summer School 2015

Parking dynamics in space and time

3 7

*Levy, N., Benenson, I, 2015, Journal of Transport Geography, 46, 220–231

slide-31
SLIDE 31

MTS Summer School 2015

Parking dynamics in space and time

PARKFIT results fits very well to the Tel Aviv field data

31

PARKFIT – free download from

https://www.researchgate.net/publication/274835167_ArcGIS-based_Python_Tool_for_Assessing_City_Parking_Patterns

slide-32
SLIDE 32

MTS Summer School 2015

Fast and frugal estimates of parking patterns in heterogeneous city

Average distance to destination and the number of failures

Parking management and policy assessment

*Levy, N., Benenson, I, 2015, Journal of Transport Geography, 46, 220–231

32

slide-33
SLIDE 33

MTS Summer School 2015

Parking management and policy assessment

Fast and frugal estimates of parking patterns in heterogeneous city

Average distance to destination and the number of failures

*Levy, N., Benenson, I, 2015, Journal of Transport Geography, 46, 220–231

33

slide-34
SLIDE 34

MTS Summer School 2015

Parking management and policy assessment

Fast and frugal estimates of parking patterns in heterogeneous city

Average distance to destination and the number of failures

*Levy, N., Benenson, I, 2015, Journal of Transport Geography, 46, 220–231

34

slide-35
SLIDE 35

MTS Summer School 2015

Parking management and policy assessment

Fast and frugal estimates of parking patterns in heterogeneous city

Average distance to destination and the number of failures

*Levy, N., Benenson, I, 2015, Journal of Transport Geography, 46, 220–231

35

slide-36
SLIDE 36

MTS Summer School 2015

Parking management and policy assessment

Fast and frugal estimates of parking patterns in heterogeneous city

Average distance to destination and the number of failures

*Levy, N., Benenson, I, 2015, Journal of Transport Geography, 46, 220–231

36

slide-37
SLIDE 37

MTS Summer School 2015

Parking management and policy assessment

Fast and frugal estimates of parking patterns in heterogeneous city

Average distance to destination and the number of failures

*Levy, N., Benenson, I, 2015, Journal of Transport Geography, 46, 220–231

37

slide-38
SLIDE 38

D/S = 0.75

38 MTS Summer School 2015

PARKFIT: Bat Yam parking pattern today

Bat Yam, average distance to parking place

Parking management and policy assessment

Distance to destination, 2012

D/S  0.75

Survey data PARKFIT

PARKFIT is based on

  • veroptimistic estimates of the parking search

process and ignores its uncertainty Nonetheless, PARKFIT’s outputs fit well to the tested sets of the field data...

*Levy, N., Benenson, I, 2015, Journal of Transport Geography, 46, 220–231

slide-39
SLIDE 39

39 MTS Summer School 2015

At a level of a city, total demand is the same as a supply: PARKFIT: Planning parking in Bat-Yam city (200,000) in 2030

67,000 Total parking places in 2030 65,000 Total cars in 2030 There are no parking places within a 400 m distance for 5000 cars!

TAZ Growth of car

  • wnership

Growth of parking supply Parking deficit

3601

560 ~ 100 ~ 460 ~

3602

400 ~ 400 ~

3603

650 ~ 50 600 ~

3605

950 ~ 950 ~

3609

330 ~ 330 ~

3701

380 ~ 380 ~

3703

720 ~ 360 360 ~

3704

360 ~ 360 ~

3705

325 ~ 325 ~

Parking management and plicy assessment

*Levy, N., Benenson, I, 2015, Journal of Transport Geography, 46, 220–231

slide-40
SLIDE 40

Every parking inspector is an agent

40 MTS Summer School 2015

Parking dynamics in space and time

PARKAGENT Modelling parking dynamics in space-time

Every car that searches for parking or parks is an agent

(1)I. Benenson, K. Martens and Birfir, S. 2008, CEUS, 32, 431–439

(2)N. Levy, K. Martens, I. Benenson, 2013, Transportmetrica A, 9 (9), 773–797

slide-41
SLIDE 41

41 MTS Summer School 2015

Parking dynamics in space and time

PARKAGENT is a spatially explicit model

slide-42
SLIDE 42

Cruising time No of issued tickets per route/area Occupancy rate per street segment/any area Distance to destination

42 MTS Summer School 2015

Parking dynamics in space and time

DRIVERS of several types, PARKING INSPECTORS

Residents Commuters Guests Customers

PARKAGENT generates great variety

  • f parking

statistics

slide-43
SLIDE 43
  • Estimate the fraction pfreeof free parking places on the way as:

pfree = Nfree/(Nfree + Noccupied)

  • Estimate expected number of free parking places on the way to destination:

Fexpected = pfree*Distance to destination/length of the parking place

  • Decide whether to park or to continue driving towards destination
  • Continuously update pfree and Fexpected while driving to destination

1

Probability to continue driving towards destination

F1 ~ 1 F2 ~ 3

43 MTS Summer School 2015

Parking dynamics in space and time

Drivers’ decision to park on the way to destination

Linear dependency seems sufficient…

slide-44
SLIDE 44

44 MTS Summer School 2015

Parking dynamics in space and time

PARKAGENT: Driver’s search area grows in time

slide-45
SLIDE 45

PARKAGENT: User interface

45 MTS Summer School 2015

Parking dynamics in space and time

slide-46
SLIDE 46

Tel Aviv city street network Simplified grid network Antwerp city street network Ramat Gan, diamond stock exchange

46 MTS Summer School 2015

Parking dynamics in space and time

PARKAGENT: Easily adjustable to a new city

slide-47
SLIDE 47

MTS Summer School 2015

Parking dynamics in space and time

PARKAGENT: Fits very well to the field data

47

slide-48
SLIDE 48

MTS Summer School 2015

Parking dynamics in space and time

48

PARKAGENT: Reveals universal parking laws

slide-49
SLIDE 49

MTS Summer School 2015 49

Occupancy = 85% Occupancy = 95% Occupancy = 99%

PARKAGENT reveals spatial details of parking patterns

Few residual parking places are always somewhere else…

PARKAGENT adequately describes parking dynamics in space and in time and reveals universal parking laws PARKFIT provides initial approximation of the parking spatial pattern PARKAGENT and PARKFIT are easily adjusted to a new city and fit well to the field data

Parking dynamics in space and time

slide-50
SLIDE 50

Different development plans were interpreted as the models scenarios

1 – 10: Planned office buildings

Search time Distance to destination Parked in the Bialik Garage Average parking search Average distance to the office Scenario, certainty

348 8.8 min 150 A, low 514 6.6 min 243 B, low 600 5.5 min 214 C, high Example of the model outputs

For each scenario, estimates of occupancy rate, distance to destination, search time

50 MTS Summer School 2015

PARKAGENT: Cost-benefit analysis of new parking facility

Multi-level garage under the main road

Parking management and policy assessment

*Levy, N., Render, M., I. Benenson, 2015, Transport Policy, 39 p 9-20

slide-51
SLIDE 51

The signpost system directs drivers to the lots that have vacant places and decreases search time by 30%

51 MTS Summer School 2015

PARKAGENT: New garage would not justify itself unless signpost system is introduced

Parking management and policy assessment

*Levy, N., Render, M., I. Benenson, 2015, Transport Policy, 39 p 9-20

slide-52
SLIDE 52

52 MTS Summer School 2015

Should we adjust parking facilities in the city center to demand?

(1)Greg Marsden, The evidence base for parking policies—a review:

The research base, in many instances, does not support, or provides evidence counter to, the assumption that parking restraint makes centers less attractive. From that on, this claim is reconfirmed many times

How to restrain cruising, then? By means of adapting parking prices!

(2)Donald Shoup strongly advocates adaptive parking prices, with

an objective to preserve 1 per 7 places free. Implemented in a famous San Francisco SF-PARK experiment in 2011-2013, (3)Millard-Ball et al, 2014.

What can be positive incentive to change the transportation mode?

Give residents the right to sell their parking permits, (4)van Ommeren et al, 2014

1 G Marsden, 2006, Transport Policy 13, 447–457 2 D. Shoup, 2004, Reg. Sci. Urban Econ. 34 (6), 753–784 3 Millard-Ball et al, 2014 Transportation Research Part A 63, 76–92 4 J. van Ommeren et al, 2014, Reg. Sci. Urban Econ. 45 (1) 33–44)

Parking management and policy assessment

Parking policy: Recent ideas and implementations

slide-53
SLIDE 53

SF-Park really worked!

It reduced 43% search time, 30% mileage, 30% emissions! But…cost $18M!

  • Millard-Ball, Adam, Rachel R. Weinberger, and Robert C. Hampshire, Is the curb 80% full or 20% empty? Assessing the impacts of San

Francisco’s parking pricing experiment. Transportation Research Part A: Policy and Practice 63 (2014): 76-92.

  • SFMTA’s evaluation of the SFpark pilot project: http://sfpark.org/resources/docs_pilotevaluation/

53 MTS Summer School 2015

Parking management and policy assessment

slide-54
SLIDE 54

Parking is a heterogeneous phenomenon

Occupancy before and after Readiness to pay is heterogeneous too

2011 2013

54 MTS Summer School 2015

Parking management and policy assessment

slide-55
SLIDE 55

To spread the demand uniformly we must react locally

55 MTS Summer School 2015

Parking management and policy assessment

slide-56
SLIDE 56

MTS Summer School 2015

Collaborators

ParkAgent -

  • Dr. Nadav Levy (TAU, Tel Aviv),
  • Dr. Evgeny Medvedev (TAU),
  • Dr. Karel Martens (Radboud U, Nijmegen)

ParkFit -

  • Dr. Nadav Levy (TAU)

ParkGame -

  • Dr. Eran Ben Elia (BGU, Beer Sheba),
  • Dr. Evgeny Medvedev (TAU),
  • Mr. Shay Ashkenazi (TAU)

Parking management and policy assessment Parking dynamics in space and time

Drivers’ parking behavior Parking spatial pattern

Parking demand and supply Basic parking model

56

slide-57
SLIDE 57

MTS Summer School 2015

Thank you!

Anticipating the future…

57