Artificial ants (AntNet)

AntNet is an algorithm to adapt the best effort routing in IP networks.
AntNet's design is based on ant colony optimization (ACO), which explores the mechanisms behind the behavior of ants, using the shortest way to define a meta-heuristic inspired by nature for combinatorial optimization.
The ACO is characterized by a multi-agent using estigmergia communication between agents (distributed transactions), using a stochastic policy decision to build solutions, estigmergia learning the parameters of the policy decision. It has been successfully applied to a variety of combinatorial problems. AntNet was the first algorithm
ACO for routing in packet-switched networks. This work is based on the work of Dr. Gianni Di Caro in AntNet, under the supervision of Prof. Marco Dorigo.
AntNet, as well as most other algorithms based on ACO, displays a series of interesting properties: it works in a fully distributed, it is highly adaptable to changing network and traffic, uses lightweight mobile agents (called ants) for sampling path assets, is robust to failures of the agent, has several routing paths, and automatically takes care of loading data from spreading.

The performance of AntNet has been extensively tested in simulations, considering the different networks and traffic patterns, and compared to various routing algorithms (state-of-the-art). In the vast majority of situations, AntNet far surpassed all its competitors, showing excellent adaptability and robustness. AntNet has also been tested in small physical networks, confirming the good results and performance in real world tests.