Based on the paper by Shuchita Upadhyaya and Richa Setiya

1. OPTIMIZATION IN ANTNET ROUTING
ALGORITHM
Vinay
Kumar

2. BACKGROUND
• What is Routing?
 Routing is the process of selecting paths in a network to send data packets.
 It fully affects the throughput, reliability, and congestion of the network.
• What is an ideal routing algorithm?
 An algorithm which can deliver a packet from source to its destination with
min. amount of delay.
 It must be adaptive and intelligent to make decisions according to
circumstances.

3. BACKGROUND…
• Routing algorithms
 OSPF, RIP, BGP are some routing algorithms in use.
 Not sufficient to tackle the complexity of networks.
 Not adaptive, not intelligent and fault intolerant.

4. INTRODUCTION
• Ant is a common insect found in your home.
• Its foraging behavior
 Discovers shortest path to a food source.
 Shares the info to other ants by stigmergy.
 Stigmergy is a form of indirect communication used by ant for problem solving activity.
 It drops pheromone through the way it goes. It follows the backtrack path with maximum
pheromones. That path is actually the shortest path.

5. INTRODUCTION…
• Ant Colony Optimization [ACO] – A family of optimization algorithms based on real ant
behavior.
• AntNet Routing Algorithm
 A routing algorithm based on ACO
 A group of artificial ants (agents) create the shortest path between a pair of nodes.
 They update the routing tables.
 Data packets are directed towards destination using the updated routing table.

6. DATA STRUCTURE MAINTAINED AT EACH NODE
1. Routing table [Tk] –
 It contains the information specifying which is the next node should be taken in order to
reach a destination.
 Row represents all adjacent nodes to current node.
 Column represents all possible destinations in the network.
 Table entry is represented by Pnd ,
Where Pnd = goodness of choosing n as next node in order to reach destination d
OR
Pheromone concentration along the link from current node to next node
n for destination d

7. DATA STRUCTURE MAINTAINED AT EACH NODE
2. Local traffic statistics
• It follows the traffic fluctuations as seen by current node
• Stored as an array where each entry contains,
 µd=estimated mean
 σ2=estimated variance
 Wd=moving observation window
for a destination node d.

8. ANTNET ROUTING ALGORITHM
• The operation involves two types of agents-
 Forward agents [Fs->d] who gather information about the state of network.
 Backward agents [Bd->s] who use the collected information from Fs->d to update the routing
tables of routers on their path.

9. ANTNET ROUTING ALGORITHM
• Fs->d is launched towards destination d to discover a shortest and least-cost path at
regular time intervals.
• Fs->d maintain a stack which stores –
 Identifier of every visited node i
 Time elapsed in arriving at node i from originating node
At any node i, Fs->d selects the next adjacent node j by using the probability value,

10. ANTNET ROUTING ALGORITHM
• If Fs->d reaches an already visited node(cycle) then the information of this node is popped
from the stack.
• If cycle is greater than half the Fs->d lifetime then the Fs->d is destroyed.
• When Fs->d reaches the destination node d then a backward agent Bd->s is created.
• Fs->d transfers its stack information to Bd->s and dies.
• Bd->s starts from d towards s following the same path as Fs->d by popping the stack at each
node.
• As Bd->s reaches a node ,it updates both the data structure of the node.

11. UPDATING LOCAL TRAFFIC STATISTICS
• The estimated mean and variance are updated as follows-
 id   id   ( o i  d   id )
 id   id   (( o i  d   id )   id )
2 2 2 2
Where, Oi->d is the observed ant’s trip time from node i to destination d
η weighs the number of most recent samples that will really affect average

12. UPDATING ROUTING TABLE
• If Bd->s arrives at a node i from a node f ,then the entry in the routing table is updated by
the formula-
Pifd  Pifd  r (1  Pifd )
Where r is reinforcement value depending on tables of node i, Ti and Mi
• Simultaneously, the pheromone value for nodes other than f is decreased so that total
sum of probability still comes 1.
Pind  Pind  rPind , n  f
• As soon as Bd->s finishes, the data packet can be sent into the network according to
updates routing tables..

13. LIMITATIONS
• Stagnation is the major problem in antnet algorithm. If a link z keeps good condition for a
long time, then its probability value will be high thus leading to more and more packets to
be sent via link z. It reduces the probability of other links.
• Stagnation is very critical problem because-
 Link z may lose its optimality.
 If network gets failed, then link z may become unavailable.
 Other long paths may become optimal which otherwise should not.

14. OPTIMIZATION
Optimization operation is performed on Routing table -
• The column wise values in the table are picked up and a sorting algorithm is executed on
these values.
• The sorted values ranging from higher to lower are stored in a temporary array.
• The difference d, amongst the adjacent values is calculated and is compared to some
threshold value say pm.
• If the difference d is less than pm then those values are selected and comparison amongst
the adjacent values is continued until difference is greater than pm.
• Otherwise at the very first occurrence of difference greater than pm , the comparison is
stopped and the corresponding values in the array are selected.
• The interfaces corresponding to these values are stored in a new routing table which will
have the same structure as the original one, but obviously, the new table will have less
number of rows.

15. WORKING

16. WORKING…
Destination nodes
E G F A
A .25 .39 .8 .34
Next nodes
C .45 .32 0 .34
D .17 .11 .1 .20
F .13 .28 .1 .12
Ist column Assumption: pm =0.1 & Node B Routing table shown
.45 .25 .17 .13
{.45} is selected

17. WORKING
Destination nodes
E G F A
A .25 .39 .8 .34
Next nodes
C .45 .32 0 .34
D .17 .11 .1 .20
F .13 .28 .1 .12
IInd column
.39 .32 .28 .11
{.39,.32,.28} is selected

18. WORKING
Destination nodes
E G F A
A .25 .39 .8 .34
Next nodes
C .45 .32 0 .34
D .17 .11 .1 .20
F .13 .28 .1 .12
IIIrd column
.8 .1 .1 0
{.8} is selected

19. WORKING
Destination nodes
E G F A
A .25 .39 .8 .34
Next nodes
C .45 .32 0 .34
D .17 .11 .1 .20
F .13 .28 .1 .12
IVth column
.34 .34 .20 0.12
{.34,.34} is selected

20. WORKING
E G F A
Next nodes
A .39 .8 .34
C .45 .32 .34
F .28
Result- Thus, more than one optimal paths if exist are identified.

21. CONCLUSIONS
• We find more than one optimal outgoing path.
• It can explore new and better paths even if network topologies change frequently.
• The problem of stagnation is resolved.
• The throughput of network is improved.

22. REFERENCES
[1] Shuchita Upadhyaya and Richa Setiya , “Identifying multiple optimal paths in Antnet Routing
Algorithm with negligible overhead,” IJCSNS International Journal of Computer
Science and Network Security, VOL.9 No.2, February 2009
[2] F. Tekiner, F. Z. Ghassemlooy and S. Al-khayatt, “The Antnet Routing Algorithm – A
Modified Version”
[3] soe.northumbria.ac.uk
Thanks …. Any Questions?