Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Impacts of Autonomous Vehicles on Parking and Congestion
Sina Bahrami, Ph.D. CandidateMatthew J. Roorda, Professor
2/35
AVs legislation and policy
3/35
Speed up the integration of AVs
4/35
Street Access
STR
EET
DR
IVEW
AY
Building
Conventional Car-parks
5/35
AV Car-parks
Street Access
STR
EET
DR
IVEW
AY
Building
6/35
1- Design Demand 2- Plot Dimensions
Optimal Parking Facility Geometry
7/35
Any vehicle can be discharged at any given point in time
Relocation Policy
8/35
Vehicle Relocation in Larger Islands
9/35
Expected Relocations Per Vehicle Retrieval
10/35
Solution Methodology
A mixed integer program with a non-linear objective function.
The purpose of the [MP] is to iteratively generate different layouts until the best layout is found.
The [SP] finds the optimal allocation of the demand between the islands.
11/35
Impact of Demand on Optimal Layout
12/35
Plot Shape Analysis
Capacity560
Capacity540
Capacity500
13/35
Parking capacity increase
14/35
Where to park?
15/35
Full information scenario
All arrival and departure times are known in
advance.
The problem is modelled as an integer
program.
16/35
Full information scenario[𝐴𝐴1, 𝐴𝐴2,𝐴𝐴3,𝐴𝐴4,𝐴𝐴5,𝐷𝐷4,𝐴𝐴6,𝐴𝐴7, 𝐴𝐴8,𝐴𝐴9,𝐴𝐴10,𝐴𝐴11,𝐷𝐷9,𝐴𝐴12,𝐷𝐷7, 𝐷𝐷2,𝐷𝐷3,𝐷𝐷5,𝐷𝐷8,𝐷𝐷1,𝐷𝐷12,𝐷𝐷6,𝐷𝐷10,𝐷𝐷11]
1
2
7
1
2
33
4
5
44
5
66
7
88
99
1010
11
91212
7
2
3
5
8 1
12
6
10
1111
17/35
Partial information scenario
Sequential stochastic optimization model
Infinite state space
Test and compare different policies using a
simulation model
18/35
Allocation policies
Arrival time
oOnly considers the arrival time
Clustering based on dwell time
oCluster AVs as short term vs long term
Blockage probability
o Blockage probability based on average dwell
times19/35
Key operational findings
Blocking probability is the best scenario when all the islands are sizeable or arrival rate is high.
Arrival policies compete with blockage probability because they consider future arrivals.
Considering Retrieving vehicles from the rear side does not reduce the number of relocations.
20/35
Regular Vehicle Parking Autonomous Vehicle Parking
21/35
Parking options
Home
o 𝐶𝐶ℎ = 2𝑥𝑥ℎ𝑐𝑐𝑡𝑡
Car-park
o 𝐶𝐶𝑝𝑝 = 2𝑥𝑥𝑝𝑝𝑐𝑐𝑡𝑡 + 𝑟𝑟𝑝𝑝(𝑡𝑡𝑝𝑝 − 2𝑥𝑥𝑝𝑝)
Cruise
o 𝐶𝐶𝑐𝑐 = 𝑡𝑡𝑝𝑝𝑐𝑐𝑡𝑡
𝑥𝑥 Travel time𝑐𝑐 Travel cost𝑟𝑟 Parking rate𝑡𝑡 Activity time
22/35
Hypothetical city
23/35
Base case scenario with 𝑟𝑟𝑝𝑝 = 3[ $ℎ𝑟𝑟
] and 𝑡𝑡𝑝𝑝 = 12[ $ℎ𝑟𝑟
]
24/35
Parking cost sensitivity analysis
25/35
Travel cost sensitivity analysis
26/35
Parking location analysis
Daily spatial distribution of cruising
Daily spatial distribution of Parking
27/35
Key findingsNo
policy
Same parking
price
Zero-occupant
toll
5 pm traffic flow snapshot
18 min
12 min
47 min
-
30 min
10 min
50 min
+1 %
15 min
11.5 min
43 min
- 3.5 %
Maximum cruising
timeAverage
travel time to
car-parks
Maximum travel
time to car-parks
Change in VKT
28/35
Vehicle to Vehicle
Vehicle to Infrastructure
Capacity enhancement
29/35
Relation between link capacity and AV proportion
30/35
The Equilibrium Condition
The equilibrium condition can be formulated
as NCP.
The UE does not have a unique solution
because the travel time function changes
regarding HV and AV flows is not symmetric.
31/35
A simple example
32/35
Best User Equilibrium flow
33/35
Traffic management policies
HV exclusive, AV exclusive, or shared links. There are 3 𝐴𝐴 different scenarios for a
network 𝐺𝐺(𝑉𝑉,𝐴𝐴). System optimal traffic assignment is used as
the lower bound. For a real size network, policies can decrease
the gap between user equilibrium and system optimal to less than 1%.
34/35
Review
Car-park dimensionsDesign demand
Optimal car-park layout Design
Arrival and departure information of each individual
Optimal operation of the car-park
Location and price of each car-park
Optimal parking policiesOptimal traffic management policies
Parking Design
Parking Operation
Parking and Network PolicyParking Choice
35/35
Thank You!
Sina BahramiEmail: [email protected]: www.linkedin.com/in/bahrami-sina