Network Assignment

The “All-or-Nothing” Assignment

Assumptions:
• People choose the shortest path to travel from an origin to destination
(the “shortest” path can be the path with least travel time or least
generalized cost)
• Un-congested conditions exist – the travel times/costs for each link of a
roadway are a constant over the time-horizon of interest

Procedure:
• Determine the shortest path between each origin-destination (OD) pair
• Assign all trips between the OD pair to this path
 Network Assignment
The “All-or-Nothing” Assignment: Example
• A,B,C, and D are zone centroids (origins and destinations of trips)
• 1,2,3,..,13 are nodes of a highway network
• The stretch of roadway between any pair of adjacent nodes is a roadway link
(e.g., 1-2, 5-6, 3-7, 8-11, etc.)
• The numbers in italics on the links are the travel times on the links (e.g.: the
travel time on link1-2 is 5 minutes)

5 6 2
A 1 2 3 4

10 4 8 4

B 5 3 6 6 7 4 8 C

3 5 8

9 10 10 5 11

2 3

12 13 D
2
 Network Assignment
The “All-or-Nothing” Assignment: Example

The following is the demand (OD matrix):

Destination
A B C D
A 0 0 400 200
Origin

B 0 0 300 100
C 0 0 0 0
D 0 0 0 0

Determine the link flows on the highway network

 Network Assignment
The “All-or-Nothing” Assignment: Example

STEP (1): find the shortest paths from A-C, A-D, B-C, and B-D

5 6 2
A 1 2 3 4

10 4 8 4

B 5 3 6 6 7 4 8 C

3 5 8

9 10 10 5 11

2 3

12 13 D
2
 Network Assignment
The “All-or-Nothing” Assignment: Example
5 6 2
A 1 2 3 4

10 4 8 4

B 5 3 6 6 7 4 8 C

3 5 8

9 10 10 5 11

2 3

12 13 D
2

A to C: 1-2-3-4-8 with a travel time of 17 minutes.

A to D: 1-2-6-7-10-12-13 with a travel time of 24 minutes
B to C: 5-6-7-8 with a travel time of 13 minutes
B to D: 5-9-10-12-13 with a travel time of 17 minutes
 Network Assignment
The “All-or-Nothing” Assignment: Example

STEP (2): Assign all demand between a zonal pair to the shortest path between the
zones
Destination Link Flow Link Flow
A B C D
A 0 0 400 200 1 to 2 7 to 8
Origin

B 0 0 300 100 2 to 3 5 to 9
C 0 0 0 0
D 0 0 0 0 3 to 4 7 to 10

1 to 5 8 to 11

2 to 6 9 to 10
A to C: 1-2-3-4-8
3 to 7 10 to 11
A to D: 1-2-6-7-10-12-13
4 to 8 10 to 12
B to C: 5-6-7-8
5 to 6 11 to 13
B to D: 5-9-10-12-13
6 to 7 12 to 13
 Network Assignment
The “All-or-Nothing” Assignment: Example

STEP (2): Assign all demand between a zonal pair to the shortest path between the
zones
Link Flow Link Flow
Destination
A B C D 1 to 2 600 7 to 8 300
A 0 0 400 200
2 to 3 400 5 to 9 100
Origin

B 0 0 300 100
C 0 0 0 0
3 to 4 400 7 to 10 200
D 0 0 0 0
1 to 5 0 8 to 11 0

2 to 6 200 9 to 10 100
A to C: 1-2-3-4-8 3 to 7 0 10 to 11 0
A to D: 1-2-6-7-10-12-13
4 to 8 400 10 to 12 300
B to C: 5-6-7-8
5 to 6 300 11 to 13 0
B to D: 5-9-10-12-13
6 to 7 500 12 to 13 300
 Network Assignment

The “All-or-Nothing” Assignment: Summary

• Relatively simple approach –

• However determination of the shortest paths require computational
time
• There are systematic procedures for determination of the shortest
paths for large-scale networks: for eg. The Dijkstra’s algorithm
• Can lead to unreasonable assignment of flows when there are multiple
paths between an OD pair with small differences in their travel times
• Assumes un-congested conditions – the travel times/costs for each link of
a roadway are a constant over the time-horizon of interest irrespective of
the number of vehicles on that link
 Network Assignment

The User-Equilibrium Assignment

Problem Setting:
• People choose the shortest path (least travel time) to travel from an origin
to destination
• The travel time for each link of the network, in turn, depends upon how
many vehicles are on that link (i.e., the flow)
• The relationship between the travel time on a link and the flow on the
link is given by a “link performance function”
 Network Assignment

The User-Equilibrium: Illustrative Example

t1  4  (2q1 )
Link 1 Demand from A to B
A B = 8 vehicle trips
Link 2

t 2  2  ( 4q 2 )

# people # people Travel time for Travel time for

choosing link 1 choosing link 2 link 1 link 2
q1 q2 t1 t2

People traveling along link 2

4 4 = 4 + (2*4) = 12 = 2 + (4*4) = 18 have an incentive to switch to
link 1 and save travel time

People traveling along link 1

6 2 = 4 + (2*6) = 16 = 2 + (4*2) = 10 have an incentive to switch to
link 2 and save travel time
State of User Equilibrium
No incentive for anyone to
5 3 = 4 + (2*5) = 14 = 2 + (4*3) = 14 switch their paths
 Network Assignment

The User Equilibrium Principle (Wardrop, 1952)

• A situation in which no traveler can improve their travel time by

unilaterally changing paths
• Occurs when every traveler follows the minimum travel time path
• At user-equilibrium, the travel times on all used paths between an origin
and destination are equal and less than those which would be
experienced by a single vehicle on any unused path
 Network Assignment

Solving The User Equilibrium Problem: Example 1

N1
Link 1 Link 4
Demand from O to D =
100 vehicle trips
O Link 3 D

Link 5
Link 2
N2

t1  2  (8q1 ) N1 The link performance functions

t 4  7  q4
are given for each link.
O t3  2q3 D t = travel time on link

t5  5  (4q5 ) q = flow on link

t 2  4q2 N2
What are the equilibrium flows on
each link of the network?
 Network Assignment

Solving The User Equilibrium Problem: Example 1

t1  2  (8q1 ) N1 Assume that the flows on the links

t 4  7  q4
are respectively
O t3  2q3 D q1 , q2 , q3 , q4 , and q5
t5  5  (4q5 )
t 2  4q2 N2

(1) Flow Conservation Equations:

At node O: q1  q2  100 At node N2: q2  q3  q5

At node N1: q1  q3  q4 At node D: q4  q5  100

 Network Assignment

Solving The User Equilibrium Problem: Example 1

t1  2  (8q1 ) N1 Assume that the flows on the links

t 4  7  q4
are respectively
O t3  2q3 D q1 , q2 , q3 , q4 , and q5
t5  5  (4q5 )
t 2  4q2 N2

(2) Apply User Equilibrium Principle: All used paths have the same travel time

Path Travel Time

(1) O-N1-D (links 1 and 4) T1  t1  t 4  (2  8q1 )  (7  q4 )  9  8q1  q4
(2) O-N2-D (links 2 and 5) T2  t 2  t5  5  4q2  4q5
(3) O-N2-N1-D (links 2, 3, and 4) T3  t 2  t3  t 4  7  4q2  2q3  q4

Assuming all three paths are used at T1  T2  T3

equilibrium:
 Network Assignment

Solving The User Equilibrium Problem: Example 1

q1  q2  100
Flow conservation equations
q1  q3  q4
q2  q3  q5

T1  T2   9  8q1  q4    5  4q2  4q5 

User equilibrium
T1  T3   9  8q1  q4    7  4q2  2q3  q4 
principle
 Network Assignment

Solving The User Equilibrium Problem: Example 1

q1  q2  100 q1  q2  100
q1  q3  q4 q1  q3  q4  0
q2  q3  q5 q2  q3  q5  0
T1  T2   9  8q1  q4    5  4q2  4q5  8q1  4q2  q4  4q5  4

T1  T3   9  8q1  q4    7  4q2  2q3  q4  8q1  4q2  2q3  2

1q1  1q2  0q3  0q4  0q5  100

1q1  0q2  1q3  1q4  0q5 0
0q1  1q2  1q3  0q4  1q5  0
8q1  4q2  0q3  1q4  4q5  4
8q1  4q2  2q3  0q4  0q5  2
 Network Assignment

Solving The User Equilibrium Problem: Example 1

1 1 0 0 0   q1  100 
     
 1 0 1  1 0 q
  2  0 
 0 1 1 0 1  * q    0 
   3  
8  4 0 1  4   q4    4 
8  4  2 0 0  q   2
   5  

1
 q1   1 1 0 0 0  100 
     
q
 2  1 0 1  1 0   0 
 q    0 1 1 0 1   0 
 3    
 q4   8  4 0 1  4   4 
q    2
 5  8  4  2 0 0   
 Network Assignment

Solving The User Equilibrium Problem: Example 1

t1  2  (8q1 ) N1
t 4  7  q4
O t3  2q3 D

t5  5  (4q5 )
t 2  4q2 N2

Link flows at user equilibrium

Travel times for each path
q1  38.11
q2  61.89 T1  9  8q1  q4  381.6
q3  29.64 T2  5  4q2  4q5  381.6
q4  67.64 T3  7  4q2  2q3  q4  381.6
q5  32.26
 Network Assignment

System Performance at User Equilibrium: Example

N1 Assume all paths are used at UE

t1  50  q1 t 4  10q4
Path 1: O-N1-D (links 1 and 4)
O t3  10  q3 D Path 2: O-N2-D (links 2 and 5)

t5  50  q5 Path 3: O-N2-N1-D (links 2, 3, and 4)

t 2  10q2 N2

q1  q2  6
q1  q3  q4 Flow conservation equations
q2  q3  q5

T1  T2   50  q1  10q4    50  10q2  q5  User equilibrium

T1  T3   50  q1  10q4   10  10q2  q3  10q4  principle
 Network Assignment

System Performance at User Equilibrium: Example

t1  50  q1 N1
t 4  10q4
Path 1: O-N1-D (links 1 and 4)
O t3  10  q3 D Path 2: O-N2-D (links 2 and 5)

t5  50  q5 Path 3: O-N2-N1-D (links 2, 3, and 4)

t 2  10q2 N2

UE solution : Travel times for each path / each

traveler
q1  2
q2  4 T1  50  q1  10q4  92 min.

q3  2 T2  50  10q2  q5  92 min.
q4  4 T3  10  10q2  q3  10q4  92 min.
q5  2 Total system travel time = 92 * 6 = 552 minutes
 Network Assignment

System Performance at User Equilibrium: Example

3 trips along
t1  50  q1 N1 O-N1-D
t 4  10q4
Path 1: O-N1-D (links 1 and 4)
O t3  10  q3 D Path 2: O-N2-D (links 2 and 5)

t5  50  q5
t 2  10q2 N2 3 trips along
O-N2-D

Alternate Solution : Travel times for each path / each

traveler
q1  3
q2  3 T1  50  q1  10q4  83 min.

q3  0 T2  50  10q2  q5  83 min.
q4  3
Total system travel time = 83 * 6 = 498 minutes
q5  3 This is less than the system travel time
corresponding to the UE case (552 minutes) !
 Network Assignment

Why is the System Optimum not naturally realized?

3 trips along
t1  50  q1 N1 O-N1-D
t 4  10q4
Path 1: O-N1-D (links 1 and 4)
O t3  10  q3 D Path 2: O-N2-D (links 2 and 5)

t5  50  q5 Path 3: O-N2-N1-D (links 2, 3, and 4)

t 2  10q2 N2 3 trips along
O-N2-D

Alternate Solution : Travel times for each path / each

traveler
q1  3
q2  3 T1  50  q1  10q4  83 min.

q3  0 T2  50  10q2  q5  83 min.
q4  3 T3  10  10q2  q3  10q4  70 min.
q5  3 People traveling on path O-N2-D
have an incentive to switch for
personal travel time savings
 Network Assignment

How to get to the system optimum conditions?

Ideas from an economist, A.C.Pigou (1920):

“Suppose there are two roads BDA and BCA both leading from B to A. If left
to itself, the traffic will be distributed so that the trouble involved in driving a
“representative” cart along each of the two roads would be equal.”

“But, in some circumstances, it would be possible, by shifting a few carts

from route D to C, to greatly lessen the trouble of driving those still left in D
while only slightly increasing the trouble of driving along C.”

“In these circumstances, a rightly chosen measure of differential taxation

against route D would create an “artificial” situation superior to the “natural”
one.”
From: M.Patriksson, Algorithms for Urban Traffic Network Equilibria, Linkoping, Sweden, 1991
 Network Assignment

How to get to the system optimum conditions?

“Suppose there are two roads BDA and BCA The “natural” situation is
both leading from B to A. If left to itself, the User Equilibrium
traffic will be distributed so that the trouble
involved in driving a “representative” cart along
each of the two roads would be equal.” This is not a typo,
remember, this result is
“But, in some circumstances, it would be possible, from 1920!
by shifting a few carts from route D to C, to greatly
lessen the trouble of driving those still left in D
while only slightly increasing the trouble of driving
The “artificial” situation is
along C.”
System Optimum

“In these circumstances, a rightly chosen

measure of differential taxation against route D
The way to achieve this
would create an “artificial” situation superior to
optimum is by differential
the “natural” one.”
taxation = congestion
pricing
From: M.Patriksson, Algorithms for Urban Traffic Network Equilibria, Linkoping, Sweden, 1991
 Four-Step Process

Trip Generation

Trip Distribution

Mode Choice

Number of vehicle Transit ridership between Number of walk/bike trips

(automobile) trips zones i & j between zones i & j
between zones i & j

Conversion of the
PA matrices to OD Transit Assignment
matrices

Transit ridership by route,

Roadway Network boarding, alighting, transit
Assignment revenue etc.

Link volumes, speeds, travel

times, toll revenue etc.