Scribd
Upload
19 views

Antalgorithm

The document describes how an ant colony algorithm was used to solve the travelling salesman problem. The algorithm models artificial ants that probabilistically choose paths between cities based on pheromone trails. As ants traverse paths, they locally update pheromone levels to explore alternatives, and the shortest path receives global reinforcement. The algorithm performed well compared to other optimization methods on symmetric and asymmetric travelling salesman problems.

Uploaded by

Sediqa Fahimi
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
19 views

Antalgorithm

The document describes how an ant colony algorithm was used to solve the travelling salesman problem. The algorithm models artificial ants that probabilistically choose paths between cities based on pheromone trails. As ants traverse paths, they locally update pheromone levels to explore alternatives, and the shortest path receives global reinforcement. The algorithm performed well compared to other optimization methods on symmetric and asymmetric travelling salesman problems.

Uploaded by

Sediqa Fahimi
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

A travelling salesperson has the task of visiting clients living in a number of different cities.

The
travelling salesman problem is : In what order should the cities be visited so that the total
distance travelled is minimized? In order to understand and fully appreciate the potential
complexity of the problem you are advised to try to solve the problem yourself. Here is a
travelling salesman applet where you can try to beat a simulated annealing algorithm in a TSP
competition.

This section describes how an Ant Colony Algorithm was used to solve the travelling salesman
problem. The solution is based on the work of Marco Dorigo and Luca Maria Gambardella, Ant
colonies for travelling salesman problem, published in BioSystems 1997. <>

The algorithm

The artificial ant is in this case an agent which moves from city to city on a TSP graph. The
agents travelling strategy is based on a probabilistic function that takes two things into account.
Firstly it counts the edges it has travelled accumulating their length and secondly it senses the
trail(pheromone) left behind by other ant agents. Each agent modifies the environment in two
different ways :

 Local trail updating: As the ant moves between cities it updates the amount of pheromone
on the edge
 Global trail updating: When all ants have completed a tour the ant that found the shortest
route updates the edges in its path

The purpose of the local updating is mainly to avoid very strong pheromone edges to be chosen
by every ant and hence to increase exploration and hopefully avoid locally optimal solutions.
The global updating function gives the shortest path higher reinforcement, that is the amount of
pheromone on the edges of the path is increased. There are three main ideas that this ant colony
algorithm has adopted from real ant colonies:

1. The ants have a probabilistic preference for paths with high pheromone value
2. Shorter paths tend to have a higher rate of growth in pheromone value
3. It uses an indirect communication system through pheromone in edges

In addition the agents were provided with a few capabilities not present in real ants, but likely to
help solving the problem at hand. For example each ant is able to determine how far away cities
are, and they all have a memory of which cities they have already visited.

For further details and exact definitions of probability functions see the online report (in
postscript)*

<>Results

The algorithm was applied to both symmetric and asymmetric versions of the TSP problem and
its performance was tested against several other naturally inspired global optimization methods,
such as neural nets, genetic algorithms and simulated annealing. The performance was quite
impressive. The ant algorithm usually found equally good or better solutions than the other
methods, and in the remaining cases it was outperformed by a very narrow margin. For very
large city numbers the algorithm was modified slightly. For details and further result reports I
refer to the online report *.

This type of application does not make use of the adaptive properties of the Ant Colony
methology. More advanced applications, such as in telecommunication network routing (see next
section), depend on this behaviour.

*These direct postscript links tend to be a bit unstable. If they take too long you might want to
try downloading the online report from the provider site or make an enquiry to the the person
maintaining the page.

You might also like