RL Based Queue Selection Algorithm for Input Queued Switches: A Theoretical Approach

In modern computer networks, packet switching is the predominant method of transmitting data between devices. As a result, efficient packet queuing and routing strategies are essential to ensure high network performance and user satisfaction. Traditionally, queue selection algorithms have been designed using static rules and heuristics that do not adapt to changes in network traffic and topology. However, the increasing complexity and dynamism of modern networks require more sophisticated approaches that can learn and adapt to changing conditions. The main purpose of this article proposes a new queue selection algorithm that utilizes
$RL$
to learn optimal policies for selecting output ports. The algorithm we need is based on Q-learning, a popular
$RL$
algorithm that learns the action value by updating the best estimate of the
$Q$
value for each (state, action) pair. We present a high-level algorithmic description of our approach and discuss its key components and parameters. We also describe a simulation -based evaluation of our algorithm using a realistic network model and compare its performance with several baseline methods.