AntNet is a novel distributed, mobile agents-based Monte Carlo system for adaptive learning of routing tables in communications networks. Inspired by ant colony algorithms and stigmergy (indirect communication through environmental modifications), AntNet uses agents to explore and exchange information in a network. Agents communicate indirectly via the network, and their behavior is guided by probabilistic routing tables and local models of network status. AntNet outperforms six state-of-the-art routing algorithms in various experimental conditions, showing superior performance and stability. The algorithm adapts routing tables based on collected information, using a combination of probabilistic entries and heuristic corrections to improve routing decisions. AntNet is tested on realistic and artificial IP networks with varying traffic patterns and node numbers. The algorithm's performance is evaluated using a simulation environment that mimics real-world network characteristics. AntNet's success is attributed to its ability to adapt to dynamic network conditions and its use of indirect communication and probabilistic models. The paper discusses the algorithm's characteristics, compares it with other routing methods, and presents experimental results demonstrating its effectiveness in various scenarios. AntNet is a distributed, adaptive routing algorithm that uses a mobile agents-based approach to solve routing problems in dynamic networks.AntNet is a novel distributed, mobile agents-based Monte Carlo system for adaptive learning of routing tables in communications networks. Inspired by ant colony algorithms and stigmergy (indirect communication through environmental modifications), AntNet uses agents to explore and exchange information in a network. Agents communicate indirectly via the network, and their behavior is guided by probabilistic routing tables and local models of network status. AntNet outperforms six state-of-the-art routing algorithms in various experimental conditions, showing superior performance and stability. The algorithm adapts routing tables based on collected information, using a combination of probabilistic entries and heuristic corrections to improve routing decisions. AntNet is tested on realistic and artificial IP networks with varying traffic patterns and node numbers. The algorithm's performance is evaluated using a simulation environment that mimics real-world network characteristics. AntNet's success is attributed to its ability to adapt to dynamic network conditions and its use of indirect communication and probabilistic models. The paper discusses the algorithm's characteristics, compares it with other routing methods, and presents experimental results demonstrating its effectiveness in various scenarios. AntNet is a distributed, adaptive routing algorithm that uses a mobile agents-based approach to solve routing problems in dynamic networks.