Introduction to multi-agent modified Q-learning routing for computer networks

Routing is an important issue in designing modern communication environments as well as in working of distributed systems. This paper concerns the concept of using reinforcement learning in the case of routing in computer networks. For this problem we propose a multi-agent approach called modified Q-learning routing. The proposed idea is compared with actually working solutions: distance-vector and deterministic routing. The properties and some experiments are presented, and it is shown that the proposed algorithm achieves better results than its opponents.