Two-/multi-action discretised learning routing algorithms in interaction with threshold-based flow control for computer networks

The interactions between adaptive routing algorithms incorporating learning automata and a variable window flow control algorithm are examined. The routing algorithms examined use two new discretised learning automata to choose the minimum delay routes in the network. The first routing algorithm can choose between two possible candidate paths. The second algorithm, using a new fast and accurate multi-action discretised automaton can choose between as many candidate paths as desired. The flow control algorithm is a threshold-based variable window algorithm that decreases the window size when a predefined delay threshold is exceeded and increases it below this threshold. The interactions between the two algorithms are studied. The resulting scheme is compared with similar schemes reported in the literature via simulations. Simulation results are presented which show that the new scheme performs quite well in both normal and abnormal network conditions.