The Use of Reinforcement Learning Algorithms in Traffic Control of High Speed Networks

Several approaches based on Computational Intelligence techniques that develop efficient solutions to some of the most significant traffic control problems of High Speed Networks have been proposed so far in the literature. In this chapter, using the experience obtained from already published works that regard the use of Reinforcement Learning Algorithms (RLA) in ATM networks, we try to form a general framework for encountering this kind of problems using RLA, making some general observations and remarks about the factors that affect considerably their performance, as well as classifying both the problems and the proposed solutions. Although this framework is developed using specific proposed mechanisms and Reinforcement Learning Algorithm, it can give some general but efficient guidelines that can be used irrespective of the RLA that is employed in each specific case, resulting in a better coping with this kind of problems.