This paper introduces AntNet, a novel approach to adaptive learning of routing tables in communications networks. Inspired by the ant colony metaphor, AntNet is a distributed, mobile agents-based Monte Carlo system that explores the network concurrently and exchanges information indirectly and asynchronously through stigmergy. The algorithm is compared with six state-of-the-art routing algorithms from telecommunications and machine learning fields, and its performance is evaluated over realistic testbeds. Results show that AntNet outperforms its competitors under all experimental conditions, demonstrating superior performance and stability. The paper discusses the main characteristics of AntNet and explains its superior performance.This paper introduces AntNet, a novel approach to adaptive learning of routing tables in communications networks. Inspired by the ant colony metaphor, AntNet is a distributed, mobile agents-based Monte Carlo system that explores the network concurrently and exchanges information indirectly and asynchronously through stigmergy. The algorithm is compared with six state-of-the-art routing algorithms from telecommunications and machine learning fields, and its performance is evaluated over realistic testbeds. Results show that AntNet outperforms its competitors under all experimental conditions, demonstrating superior performance and stability. The paper discusses the main characteristics of AntNet and explains its superior performance.