Abstract
Network management tools are usually inherited from one generation toanother. This was successful since these tools have been kept in check andupdated regularly to fit new networking goals and service requirements.Unfortunately, new networking services will render this approach obsolete andhandcrafting new tools or upgrading the current ones may lead to complicatedsystems that will be extremely difficult to maintain and improve. Fortunately,recent advances in AI have provided new promising tools that can help solvingmany network management problems. Following this interesting trend, the currentarticle presents LEASCH, a deep reinforcement learning model able to solve theradio resource scheduling problem in the MAC layer of 5G networks. LEASCH isdeveloped and trained in a sand-box and then deployed in a 5G network. Theexperimental results validate the effectiveness of LEASCH compared toconventional baseline methods in many key performance indicators.