You are on page 1of 23

5.

Network Analysis
 Network Configuration
 Traffic Flow
 Supply Function
 Shortest Path
 Assignment Principle
 Static Assignment

1
Network configuration
 Network
 consists of nodes and links

j
◎ o o destination
o ← link(directional) node
o ◎
o
o o ← node link
o ◎
(directional)
◎ o ◎ ← origin/
o destination i
o ◎
origin

2
Node and Link
 Node
 boundary of uniform highway sections
 junction of links
 intersection (surface streets)
 interchange (motorways)
 point of geometric change
 lane reduction, vertical grade
 origins and destinations
 called as “centroid”
 Link
 section connecting two adjacent nodes
 consider “direction” for traffic flow

3
Traffic flow
 loaded demand onto a network
 flow representing variables
 Q jd : traffic demand (rate) from origin node j to
destination node d
 xij : traffic flow on link (i,j)
 xijd : traffic flow on link (i,j) toward destination
node d

4
Flow conservation principle
 Flow variables must satisfy the flow
conservation principle
 The amount of flow cannot disappear but must be
conserved.
 Flow conservation at node j with respect to
flow toward destination d:

 k
x  i x  Q
d
jk
d
ij
jd

 left side: total flow exiting from node j


 right side: entering flow into node j
+ demand generated at j toward destination d
5
 summed up for all the destinations d,

 k
x jk   i xij  R j  S j

 Rj(t) : generated demand at node j


 Sj(t) : absorbed demand at node j

6
Supply function
 describe the service level (= performance) of
a link when flow is loaded
 light traffic – smooth travel with desired speed
 heavy traffic – slow speed and more travel time
 also called as “cost function”

7
Flow-dependency
 If link cost depends upon its flow
 “flow-dependent” cost function
 Cij(xij): cost of link (i, j) when link flow is xij
 If link cost stays constant regardless of its flow
 “flow-dependent” cost function
 Cij : constant cost of link (i, j)
cost Cij(xij) flow-dependent

flow-independent

flow xij
8
Monotonic function
 Normally, flow-dependent cost function is
assumed monotonic increasing in link flow xij
 travel cost get larger as more congested
 travel time, fuel, discomfort etc.

flow-dependent
cost Cij(xij)

More cost due to


longer travel time,
more fuel etc.

flow xij

9
Non-monotonic function
 some cases of non-monotonic cost function
 public transportation
 high demand  delay, discomfort
 low demand  fewer service  long waiting time
 goods delivery service
passenger cost flow-dependent
Cij(xij)

More cost due to


longer dwell time,
Longer waiting time discomfort in a crowded bus
due to fewer services
flow xij

10
Shortest path
 Assuming flow-independent cost function,
how to obtain shortest path from an origin to
a destination?
 path: route (a series of links) from origin to
destination
 Several algorithm have been developed
mostly based on Bellman’s dynamic
programming principle (動的計画法).
 A complicated problem is solved by breaking it
down into simpler sub-problems and solving them
recursively.

11
 Suppose the shortest path P from an origin to
a destination.
according to Bellman’s principle
 the shortest path from any node on path P to
the same destination must belong to path P.

path P

origin
destination

12
Settings
 node j is connected by link (i,j)s
 the shortest path from origin o to node is are
already determined
 d(o,i): the shortest costs from o to i
 the shortest cost to node j must be given as
d(o,i) = Mini {d(o,i) + Cij}
node i
Cij
d(o,i)
node j
origin

13
Dijkstra method
 calculates the shortest paths from an origin to
every nodes simultaneously
 nodes are classified into two groups:
 Np: nodes to which the shortest paths have been
already determined
 Nt: nodes to which the shortest paths have not
been confirmed

14
Dijkstra’s algorithm (1)
 Step 1: Initialization
 d(o,j) := ∞, and p(j) = ∞, for all node j  origin
node o
 Np :=  , Nt := {1, 2, 3, .........., N}
 d(o,o) := 0, and p(o) = 0
 i=o
 Step 2:
 Add node i into Np and remove from Nt
 If Nt is empty, end.

15
Dijkstra’s algorithm (2)
 Step 3:
 For every node j connected by link (i,j) from
node i, calculate Mini {d(o,i) + Cij}. And update
d(o,j) as follows:
 if d(o,j) > Mini {d(o,i) + Cij},
 d(o,j) = Mini {d(o,i) + Cij} and p(j) = i.
 Step 4:
 Among node j in Nt, find node k which gives
the minimum cost: d(o,k) = Minj d(o,j), j  Nt.

16
Dijkstra’s algorithm (3)
 Step 5:
 Set i = k, return to Step 2.

 Tracing preceded nodes p(i)’s, the shortest


path to any node from the origin is confirmed.
 Note: If p(i) is infinity, it means node i cannot be
accessed from the origin.

17
Label modifying method
 repeatedly revises paths from an origin node
based on the Bellman’s principle
 the shortest path to one node;
 Dijkstra algorithm – determined in each iteration
 Label modifying algorithm – not determined
 However, Label modifying algorithm does not
require to find a node of the minimum cost
d(o,j) in Step 4 in the Dijkstra algorithm.

18
Label modifying algorithm (1)
 Step 1: Initialization
 d(o,j) := ∞, and p(j) = ∞, for all node j  origin
node o,
 NL := {o}, where NL is a list of nodes,
 d(o,o) := 0, and p(o) = 0,
 Step 2:
 Find node i at the top of NL and remove node i
from NL.

19
Label modifying algorithm (2)
 Step 3:
 For every node j connected by link (i,j), calculate Mini
{d(o,i) + Cij}. And update d(o,j) as follows:
 if d(o,j) > Mini {d(o,i) + Cij},
 d(o,j) = Mini{d(o,i) + Cij}, p(j) = i, and
add node j at the bottom of NL if j is not in NL.
 Step 4:
 If NL is empty, end. Otherwise, return Step 2.

 A node could be added to and removed from NL


several times.
 In the Dijkstra method, once a node is moved from Nt
to Np, the node is never move back again.
20
Example
 Obtain minimum cost from origin 1 to all other
nodes.

21
Dijkstra method
ref. d(o,i) p(i)
node
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
0 ∞ ∞ ∞ ∞ ∞ ∞ ∞ ∞ ∞ 0 ∞ ∞ ∞ ∞ ∞ ∞ ∞ ∞ ∞

22
Label modifying method
ref. d(o,i) p(i) NL
node 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

0 ∞ ∞ ∞ ∞ ∞ ∞ ∞ ∞ ∞ 0 ∞ ∞ ∞ ∞ ∞ ∞ ∞ ∞ ∞ 1

23

You might also like