Chapter 7: Conclusions and Prospects for the Future

7.1 What Is Known About ACO?

ACO is a well-recognized member of the family of metaheuristic methods for discrete optimization problems.
The theory developed has little practical use. For example, we know that both MMAS and ACS converge in value, both with and without local search.
ACO algorithms are state-of-the-art for many problems.
Companies such as EuroBios and AntOptima use ACO for tasks such as scheduling and vehicle routing.

Dynamic problems such as network routing or the dynamic TSP.
Stochastic problems such as network routing.
Multi objective optimization problems (MOOP) such as tardiness scheduling with changeover cost. Multiple ant colonies might be used for the different objectives.
Understanding effective parallelization and the resulting performance improvement.
Understanding and characterizing behavior better. For example, ACO can be studied under controlled and simplified experimental conditions. Often, it is insightful to find circumstances under which ACO performs poorly.

An ant algorithm is a multiagent system inspired by the observation of real ant colony exploiting stigmergy.

Foraging and Path Marking. Pheromone trails could be used to cover a graph of unknown topology. This might be useful in the context of Internet search.
Brood Sorting. This might be useful for classifying web documents based on their similarity to one another.
Division of Labor. One applications is choosing a paint booth for trucks coming out of an assembly line in a truck factory.
Cooperative Transport. One application is cooperative box pushing or object pulling.