Sunteți pe pagina 1din 10

CVEN 2401/2402 Workshop: Week 6

Trip Assignment

This week we will review the final step of the four step transportation forecasting model, trip
assignment (or otherwise commonly known as traffic assignment).

Trip Assignment
The final step of the forecasting process is trip assignment which concerns the selection of
routes for the trips distributed (accounting for different modes) between origins and
destinations of the transport network. Trip assignment answers the question Which route
will people choose to get from their origin to their destination? Trip assignment outputs
measures of network performance, such as total system travel time.
Trip assignment can be vastly complex problem (consider the Sydney area where there are
an estimated 1.8M vehicle trips during the morning peak!). Therefore, planning models
usually make use of more simplified models based on Wardrops (descriptive) principle of
equilibrium. Recall:

The travel time on all used paths between an origin and a destination point are equal and
less than those which would be experienced by a single vehicle on any unused path, i.e., no
traveller can improve their travel time by unilaterally changing routes.

Question 1

Solving for an equilibrium solution


Consider the network shown in Figure 1.

CVEN2401/2402 Workshop Week 5: Trip Assignment 1


1

3 4

2
Figure 1. The network of interest

Assume that the demand between nodes 1 and 4 is 14 = 2000 / and between
nodes 2 and 4 is 24 = 1000 /. The link performance functions are given below,
where represents travel time on a link and represents link flow.

13 = 0.5 + 0.00113

23 = 0.5 + 0.00123

14 = 1.0 + 0.000514

24 = 1.0 + 0.00124

34 = 1.0 + 0.00234

First, write down the User Equilibrium equations. Then solve for equilibrium volumes and
travel times.

Solution

User equilibrium means that the travel times on all used routes are the same. Therefore:

14 = 13 + 34

24 = 23 + 34

But we need more equations. Consider the demand constraints:

14 + 13 = 14

24 + 23 = 24

Also consider the network. The flow on 1,3 and 2,3 must be equal to the flow on 1,4:

13 + 23 = 34

So with five unknowns and five equations it is time to solve.

Consider:

14 13 = 24 23

1 + 0.000514 0.5 0.00113 = 1 + 0.00124 0.5 0.00123

CVEN2401/2402 Workshop Week 5: Trip Assignment 2


The 1s and 0.5s cancel. Then:

0.514 13 = 24 23

Rearrange the equation like so:

0.5(14 + 13 ) 1.5 13 = (24 + 23 ) 223

And now we can substitute our demands

0.5(2000) 1.513 = 1000 223


4
13 = 3 23 [A]

Now return to another equation.

14 = 13 + 34

1.0 + 0.000514 = 0.5 + 0.00113 + 34 = 1.0 + 0.00234

Cancel and rearrange a bit:

0.514 = 500 + 13 + 234

Substitute again for 34 and 14

0.5(14 13 ) = 500 + 13 + 2(13 + 23 )

Substitute [A]

3
500 = 3.513 + 13
2

13 = 100

Then find that 13 = 0.6 also find that 23 = 75 and therefore 23 = 0.575

24 = 24 + 23 24 = 925 24 = 1.925

And

34 = 175 34 = 1.35

So total travel time

= 13 13 + 23 23 + 14 14 + 24 24 + 34 34

When you sub everything = 5824.375

Question 2

Determining the shortest path

Part of solving for the conditions of user equilibrium involves solving for the shortest path.

CVEN2401/2402 Workshop Week 5: Trip Assignment 3


a) Using the network in Figure 2, solve for the shortest path trees for nodes 1 and node
7. Assume links are bidirectional and have the same cost for each direction. The
number shown by the link is the cost.

Solution

Apply the Djikstra shortest path algorithm

1. Set labels for all nodes excluding the origin node to a very large number
(infinity) and set the origin label to be zero.

2. Select the minimum label from all the remaining nodes (initially among all of
the nodes) and set as a permanent label.

3. Check neighbouring node labels and if the cost to reach that


neighbouring node is less than the current label of that neighbouring
node, update the label to this new lower cost. Once an update is made keep
track of the predecessor node that resulted in the update.

4. Repeat step 2 until all nodes are permanently labelled.

Here is the shortest path tree solution. Let students have a try and then go over it by hand.
You just solve by adding up link costs. The numbers listed by the nodes are the costs.

CVEN2401/2402 Workshop Week 5: Trip Assignment 4


b) Now consider the following demand:

Origin node Destination node Demand

1 16 19

1 13 42

Assign this demand to the shortest path. What is the flow on each link? Which links are the
most congested? Discuss how the resulting cost in this problem is different from what
happens on actual traffic networks

Solution:

The most congested links are from 1 to 5, 5 to 6, and 6 to 7


It is important to note that with this network, the link costs are static not accounting
for any congestion effects which does not reflect realistic traffic conditions. Generally,
the more the demand is for a link, the greater the cost to travel on a link. Thus, after a
network is loaded, the shortest path may no longer be the shortest and an iterative
process is undertaken to achieve equilibrium conditions.

CVEN2401/2402 Workshop Week 5: Trip Assignment 5


Question 3

Method of Successive Averages

Consider the network below:

2 24
2 4

3 1

1 1 2 2
2 3 4

1 2 13

1 3 3
3

And assume that there are 10 units traveling from 1-4, and 10 units traveling from 1-5. Use
the MSA method to find the equilibrium flows.

Solution:

The general structure of the algorithm is as follow:

Iteration 0:
Set flows to 0 (free flow) and calculate link costs:
Iteration 1:
Find shortest path given fixed costs

CVEN2401/2402 Workshop Week 5: Trip Assignment 6


Update link costs given total link flows (over all paths)
Iteration 2:
Find shortest paths given fixed costs from previous iteration
Update link costs given total link flows (over all paths)
Repeat until converged.

So, we begin with Iteration 0. If we set all the flows to zero, we obtain the following link costs:

2
2 4

1 1 2

1
3

Now on to iteration 1:

- The shortest paths will be:


1 to 4 -> 1-3-5-4
1 to 5 -> 1-3-5

Path Cost Old X bar New


1-2-4 5 - 0 0
1-2-3-5-4 7 - 0 0
1-3-5-4 4 - 10 10
1-3-5 2 - 10 10
1-2-3-5 5 - 0 0

If we assign all 10 units of flow of each OD to the shortest paths, the new costs will be:

CVEN2401/2402 Workshop Week 5: Trip Assignment 7


2
2 4

1 1 32

41

1
3

Now on to iteration 2:
- The shortest paths will be:
1 to 4 -> 1-2-4
1 to 5 -> 1-2-3-5

Path Cost Old X bar New


1-2-4 5 0 10 5
1-2-3-5-4 97 0 0 0
1-3-5-4 134 10 0 5
1-3-5 102 10 0 5
1-2-3-5 65 0 10 5

We average the solution from the previous iteration by 1/2 and the new shortest path flows
by 1/2.
Using these path flows we update the link costs and get the following costs:

2 4

13

1 11 1

21

4
3

CVEN2401/2402 Workshop Week 5: Trip Assignment 8


Now on to iteration 3,

- The shortest paths will be:


1 to 4 -> 1-2-4
1 to 5 -> 1-3-5

IMPORTANT NOTE: Just because the shortest paths were the same in two consecutive
iterations does not mean the problem has converged.

We will now AVERAGE our previous two solutions (2/3 weight to solution 2, 1/3 weight to the
new paths).

Path Cost Old X bar New


1-2-4 20 5 10 6.66666667
1-2-3-5-4 87 0 0 0
1-3-5-4 84 5 0 3.33333333
1-3-5 67 5 10 6.66666667
1-2-3-5 70 5 0 3.33333333

Using these path flows we update the link costs and get the following costs:
.

8.66667
2 4

13

1 7.666667 12

21

41
3 5

Now on to iteration 4,

- The shortest paths will be:


1 to 4 -> 1-2-4
1 to 5 -> 1-2-3-5

We will now AVERAGE our previous two solutions (3/4 weight to solution 3, 1/4 weight to the
new paths).

CVEN2401/2402 Workshop Week 5: Trip Assignment 9


Path Cost Old SP New
1-2-4 21.66667 6.666667 10 7.5
1-2-3-
5-4 73.66667 0 0 0
1-3-5-4 74 3.333333 0 2.5
1-3-5 62 6.666667 0 5
1-2-3-5 61.66667 3.333333 10 5

And we continue to do this procedure until convergence. The convergence of the algorithm
will be seen when the change between solution "n" and solution "n+1" is small enough.
Again, this is not to be confused with the change between the shortest path at iteration n and
the shortest path at iteration n+1, as this will not indicate convergence.

CVEN2401/2402 Workshop Week 5: Trip Assignment 10

S-ar putea să vă placă și