DataPDF Available
Maritime & Transport Technology
Dual-Mode Vehicle Routing
in Mixed Autonomous &
Non-Autonomous Zone Networks
Breno Beirigo
Asst. Prof. Frederik Schulte
Prof. Rudy Negenborn
Prins Hendrik Laan, Utrecht - NL.1
Is infrastructure prepared
to accommodate AVs?
Arc de Triomphe, Paris - FR.
Is infrastructure prepared
to accommodate AVs?
2
Maritime & Transport Technology
Content
Problem definition
Dual-mode routing, real-world input, zone networks
MILP formulation
Preprocessing and site-dependent DARP
Scenarios for AV deployment
Zones, cost scenarios, and transportation demand
Numerical study
Test cases and results
Maritime & Transport Technology
Content
Problem definition
Dual-mode routing, real-world input, zone networks
Considering heterogeneous networks
impacts vehicle routing.
1 2 3 4 5 6
7 8 9 10 11 12
13 14 15 16 17 18
19 20 21 22 23 24
25 26 27 28 29 30
31 32 33 34 35 36
jConventional
Driving (CD)
Infrastructure
4
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
Considering heterogeneous networks
impacts vehicle routing.
1 2 3 4 5 6
7 8 9 10 11 12
13 14 15 16 17 18
19 20 21 22 23 24
25 26 27 28 29 30
31 32 33 34 35 36
jConventional
Driving (CD)
Infrastructure
Shortest path
CD
j
4
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
(6)
Considering heterogeneous networks
impacts vehicle routing.
1 2 3 4 5 6
7 8 9 10 11 12
13 14 15 16 17 18
19 20 21 22 23 24
25 26 27 28 29 30
31 32 33 34 35 36
jConventional
Driving (CD)
Infrastructure
Shortest path
CD
j
4
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
Considering heterogeneous networks
impacts vehicle routing.
1 2 3 4 5 6
7 8 9 10 11 12
13 14 15 16 17 18
19 20 21 22 23 24
25 26 27 28 29 30
31 32 33 34 35 36
jConventional
Driving (CD)
jAutonomous
Driving (AD)
Infrastructure
Shortest path
CD
j
4
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
Considering heterogeneous networks
impacts vehicle routing.
1 2 3 4 5 6
7 8 9 10 11 12
13 14 15 16 17 18
19 20 21 22 23 24
25 26 27 28 29 30
31 32 33 34 35 36
jConventional
Driving (CD)
jAutonomous
Driving (AD)
Infrastructure
Shortest path
CD
j
4
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
(10)
Considering heterogeneous networks
impacts vehicle routing.
1 2 3 4 5 6
7 8 9 10 11 12
13 14 15 16 17 18
19 20 21 22 23 24
25 26 27 28 29 30
31 32 33 34 35 36
jConventional
Driving (CD)
jAutonomous
Driving (AD)
Infrastructure
Shortest path
CD
j
4
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
(10)
Considering heterogeneous networks
impacts vehicle routing.
1 2 3 4 5 6
7 8 9 10 11 12
13 14 15 16 17 18
19 20 21 22 23 24
25 26 27 28 29 30
31 32 33 34 35 36
jConventional
Driving (CD)
jAutonomous
Driving (AD)
Infrastructure
Shortest path
CD
j
jAD
4
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
(8)
(10)
Considering heterogeneous networks
impacts vehicle routing.
1 2 3 4 5 6
7 8 9 10 11 12
13 14 15 16 17 18
19 20 21 22 23 24
25 26 27 28 29 30
31 32 33 34 35 36
jConventional
Driving (CD)
jAutonomous
Driving (AD)
Infrastructure
Shortest path
CD
j
jAD
4
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
(8)
(10)
Considering heterogeneous networks
impacts vehicle routing.
1 2 3 4 5 6
7 8 9 10 11 12
13 14 15 16 17 18
19 20 21 22 23 24
25 26 27 28 29 30
31 32 33 34 35 36
jConventional
Driving (CD)
jAutonomous
Driving (AD)
Infrastructure
Shortest path
CD
j
jAD
jDual-mode
(CD+AD)
4
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
(8)
(6)
(10)
5
3
2’
2
A
3’ 1’
D
C
1
AVZ
Pick-up DeliveryVehicle
(capacity 2)
City map, vehicles,
and requests
Real-world input is transformed into
a viable transportation network.
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
1 passenger
5
3
2’
2
A
3’ 1’
D
C
1
AVZ
Pick-up DeliveryVehicle
(capacity 2)
City map, vehicles,
and requests
Real-world input is transformed into
a viable transportation network.
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
1 passenger
Transportation
network
3
2’
2
3’ 1’
D
C
1
A
Dual-mode
Automated
Conventional
Driving
mode
Maritime & Transport Technology
Content
MILP formulation
Preprocessing and site-dependent DARP
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018 7
3
2’
2
3’ 1’
D
C
1
A
A
Vehicles
(capacity 2) D C
Request
origins 1 2 3
Request
destinations 1’ 2’ 3’
Site-dependent dial-a-ride problem: Requests can
be serviced according to vehicle’s driving capabilities.
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018 8
(
D
D
1
)
(
D
2
1
(
A
3
2
)
(
C
1
2
)
(
C
2
2’
)
(
D
D
2
)
(
D
2
3
(
A
3
1
)
(
C
1
3
)
(
C
2
3’
)
(
D
D
3
)
(
D
2
1’
(
A
3
1’
)
(
C
1
1’
)
(
C
2’
3
)
(
D
1
2
)
(
D
2
2’
(
A
3
2’
)
(
C
1
2’
)
(
C
2’
1
)
(
D
1
3
)
(
D
2
3’
(
A
3
3’
)
(
C
1
3’
)
(
C
2’
3’
)
(
D
1
1’
)
(
D
2’
3
(
A
3’
2
)
(
C
1’
2
)
(
C
2’
1’
)
(
D
1
2’
)
(
D
2’
3
(
A
3’
1
)
(
C
1’
3
)
(
D
1
3’
)
(
D
2’
1’
(
A
3’
2’
)
(
C
1’
2’
)
(
D
1’
2
)
(
D
2’
3’
(
A
3’
1’
)
(
C
1’
3’
)
(
D
1’
3
)
(
A
A
1
(
A
2
3
)
(
C
3
2
)
(
D
1’
2’
)
(
A
A
2
(
A
2
1
)
(
C
3
1
)
(
D
1’
3’
)
(
A
A
3
(
A
2
1’
)
(
C
3
1’
)
(
D
3
2
)
(
A
1
2
(
A
2
2’
)
(
C
3
2’
)
(
D
3
1
)
(
A
1
3
(
A
2
3’
)
(
C
3
3’
)
(
D
3
1’
)
(
A
1
1’
(
A
2’
3
)
(
C
3’
2
)
(
D
3
2’
)
(
A
1
2’
(
A
2’
1
)
(
C
3’
1
)
(
D
3
3’
)
(
A
1
3’
(
A
2’
3’
)
(
C
3’
2’
)
(
D
3’
2
)
(
A
1’
2
(
A
2’
1’
)
(
C
3’
1’
)
(
D
3’
1
)
(
A
1’
3
(
C
C
1
)
(
C
2
3
)
(
D
3’
2’
)
(
A
1’
2’
(
C
C
2
)
(
C
2
1
)
(
D
3’
1’
)
(
A
1’
3’
(
C
C
3
)
(
C
2
1’
)
3
2’
2
3’ 1’
D
C
1
A
Preprocessing input data in SDDARP reduces model complexity.
Available (k,i,j) trips (infeasible vehicle load):
A
Vehicles
(capacity 2) D C
Dual-mode
Automated
Conventional
Driving
mode
Request
origins 1 2 3
Request
destinations 1’ 2’ 3’
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018 9
Preprocessing input data in SDDARP reduces model complexity.
A
Vehicles
(capacity 2) D C
(
D
D
1
)
(
D
2
1
(
A
3
2
)
(
C
1
2
)
(
C
2
2’
)
(
D
D
2
)
(
D
2
3
(
A
3
1
)
(
C
1
3
)
(
C
2
3’
)
(
D
D
3
)
(
D
2
1’
(
A
3
1’
)
(
C
1
1’
)
(
C
2’
3
)
(
D
1
2
)
(
D
2
2’
(
A
3
2’
)
(
C
1
2’
)
(
C
2’
1
)
(
D
1
3
)
(
D
2
3’
(
A
3
3’
)
(
C
1
3’
)
(
C
2’
3’
)
(
D
1
1’
)
(
D
2’
3
(
A
3’
2
)
(
C
1’
2
)
(
C
2’
1’
)
(
D
1
2’
)
(
D
2’
3
(
A
3’
1
)
(
C
1’
3
)
(
D
1
3’
)
(
D
2’
1’
(
A
3’
2’
)
(
C
1’
2’
)
(
D
1’
2
)
(
D
2’
3’
(
A
3’
1’
)
(
C
1’
3’
)
(
D
1’
3
)
(
A
A
1
(
A
2
3
)
(
C
3
2
)
(
D
1’
2’
)
(
A
A
2
(
A
2
1
)
(
C
3
1
)
(
D
1’
3’
)
(
A
A
3
(
A
2
1’
)
(
C
3
1’
)
(
D
3
2
)
(
A
1
2
(
A
2
2’
)
(
C
3
2’
)
(
D
3
1
)
(
A
1
3
(
A
2
3’
)
(
C
3
3’
)
(
D
3
1’
)
(
A
1
1’
(
A
2’
3
)
(
C
3’
2
)
(
D
3
2’
)
(
A
1
2’
(
A
2’
1
)
(
C
3’
1
)
(
D
3
3’
)
(
A
1
3’
(
A
2’
3’
)
(
C
3’
2’
)
(
D
3’
2
)
(
A
1’
2
(
A
2’
1’
)
(
C
3’
1’
)
(
D
3’
1
)
(
A
1’
3
(
C
C
1
)
(
C
2
3
)
(
D
3’
2’
)
(
A
1’
2’
(
C
C
2
)
(
C
2
1
)
(
D
3’
1’
)
(
A
1’
3’
(
C
C
3
)
(
C
2
1’
)
Available (k,i,j) trips (infeasible vehicle load & vehicle/zone compatibility) :
3
2’
2
3’ 1’
D
C
1
ADual-mode
Automated
Conventional
Driving
mode
Request
origins 1 2 3
Request
destinations 1’ 2’ 3’
10
𝒎(𝒌) =
Type of vehicle
𝑘.
𝒕𝒊,𝒋
𝒎(𝒌) =
Travel time from node
𝑖to node 𝑗in mode 𝑚 𝑘 for vehicle 𝑘.
Parameters (abridged):
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
𝒙𝒊,𝒋
𝒌=
1, if vehicle
𝑘travels from point 𝑖to point 𝑗. The trip (𝑘, 𝑖, 𝑗)
is
feasible
(after preprocessing).
𝒓𝒊
𝒌=
Time spent by request
𝑖in vehicle 𝑘. The visit (𝑘, 𝑖) is valid
(after preprocessing).
Variables (abridged):
MILP formulation: Vehicle types have different
travel times between pickup and delivery points.
11
Operational cost
Distance rate (0,001 €/s)
Base fare (3,00)
Maximize:
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
(𝑘, 𝑖, 𝑗)
𝑘, 𝑖, 𝑗
& request i
MILP objective: Maximize overall profit.
11
Subject to (abridged):
Min. travel time
Trip duration
Ride delay (10 min)
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
request 𝑖, visit (𝑘, 𝑖)
MILP objective: Maximize overall profit.
Maritime & Transport Technology
Content
Scenarios for AV deployment
Zones, cost scenarios, and transportation demand
13
Autonomous Vehicle Zones (AVZs) configurations
are generated for Delft, The Netherlands.
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
45
Automated driving (AD) Conventional driving (CD)
AD coverage:
{10%, 25%, 50%}
#AVZ centers:
{1, 2, 4}
14
Large price
premium
Scenario 1
DV AV CV Scenario 2
Moderate price
premium
DV AV CV Scenario 3
Minimal price
premium
DV AV CV
Cost scenarios: AV technologies become increasingly
affordable and labor/operational costs remain constant.
= 0,001 €/s = Operational costs = Labor costs= Automation costs
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
15
CVZ AVZ CVZ CVZAVZ AVZ
High Moderate Low
The demand distributions are based on transportation
patterns reflecting three zone-crossing frequencies:
= 10% demand / AVZ = autonomous vehicle zone / CVZ = conventional vehicle zone
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
Maritime & Transport Technology
Content
Numerical study
Test cases and results
17
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
INPUT
Mixed-zone network
Cost scenario
Demand configuration
Fleet
Requests
Optimal results were obtained in
91% of 14.580 test cases.
17
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
INPUT
Mixed-zone network
Cost scenario
Demand configuration
Fleet
Requests
Optimal results were obtained in
91% of 14.580 test cases.
Preprocessing
MILP
17
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
INPUT
Mixed-zone network
Cost scenario
Demand configuration
Fleet
Requests
Fleet mix
Routing plan
Managerial insights
Infrastructural insights
OUTPUT
Optimal results were obtained in
91% of 14.580 test cases.
Preprocessing
MILP
18
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
Fleet composition is sensitive to automated driving
coverage and passenger movements between zones.
High
zone-crossing
Moderate
zone-crossing
Low
zone-crossing
AD coverage
AV CV DV
19
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
Fleet composition is sensitive to automated driving
coverage and passenger movements between zones.
higher zone-crossing
frequency
more Dual-mode
vehicles
High
zone-crossing
Moderate
zone-crossing
Low
zone-crossing
AD coverage
AV CV DV
High
zone-crossing
Moderate
zone-crossing
Low
zone-crossing
AD coverage
20
wider automated
driving coverage
more AVs
fewer CVs
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
Fleet composition is sensitive to automated driving
coverage and passenger movements between zones.
AV CV DV
High
zone-crossing
Moderate
zone-crossing
Low
zone-crossing
AD coverage
21
low zone-crossing
frequency
fewer dual-mode
vehicles
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
Investigating user movement patterns can help
reducing the number of dual-mode vehicles.
AV CV DV
22
Conclusions
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
22
Conclusions
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
The functions of traffic planning and fleet management may
converge in the future.
22
Conclusions
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
The functions of traffic planning and fleet management may
converge in the future.
Data analysis will be essential for the quality of future
(partially) autonomous mobility services.
Breno Beirigo
B.AlvesBeirigo@tudelft.nl Asst. Prof. Frederik Schulte
F.Schulte@tudelft.nl Prof. Rudy Negenborn
R.R.Negenborn@tudelft.nl
THANK YOU
FOR YOUR ATTENTION
Maritime & Transport Technology
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018 26
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018 27
28
Summary of test case settings
totaling 14.580 instances:
Breno Beirigo, Frederik Schulte, Rudy Negenborn ITSC2018
Parameter
Values
Operational cost scenario
{S01, S02, S03}
Number of vehicles
{15, 30, 60}
AD coverage percentage
{10%, 25%, 50%}
Number of AD zones
{1, 2, 4}
Number of zone deployments
5
Number of requests
{10, 20, 40}
Zone
-crossing frequency
{high, moderate, low}
Time interval
(min)
{1, 5, 10, 20}
36 demand
configurations
45 mixed-zone
network configurations
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.