Reward-based routing and Braess´ paradox an efficient resource usage strategy

There exists evidence showing that Internet end-to-end performance is directly affected by well-known routing inefficiencies. Braess´ paradox is an example of the inefficient use of resources as a result of inadequate policies, thus, tackling it might be a necessary condition for new routing strategies. In this paper, we propose a routing strategy based on Q-routing that provides an improved exploration/exploitation tradeoff, which resulted in a dynamic load balancing scheme. We tested it under the conditions of the Braess´ paradox, using the NS-2 network simulator. It is shown that the new strategy avoids the Braess´ paradox, increasing network resource availability.