Optimal radio resource management in 5G NR featuring network slicing

3GPP 5G New Radio (NR) has introduced several new features that the network slicing concept can leverage to guarantee the heterogeneous requirements in terms of throughput and delays of expected 5G network services. Mainly, these features are (i) Mixed-numerology to control the time slot duration, hence guaranteeing low latency requirements; (ii) Bandwidth Parts (BWP) to control the number of radio resources allocated to users, hence satisfying different throughput requirements. However, efficient radio resource management is already complex, and adding these new dimensions will further increase this complexity. In this paper, we first propose modeling radio resource management in 5G NR featuring network slicing through a Mixed Integer Linear Program (MILP). For our best knowledge, this is the first MILP modeling of the radio resource management featuring network slicing taking into account (i) Mixed-numerology, (ii) both latency and throughput requirements (iii) multiple slice attach per UE (iv) Inter-Numerology Interference (INI). After showing that solving the problem takes an exponential time, we considered a new approach to solve it in a polynomial time, which is highly required when scheduling radio resources. The new approach consists of formalizing this problem using a Deep Reinforcement Learning (DRL)-based solver. We evaluate the use of RL to solve the problem for different network configurations and compared its performance with the optimal solution obtained by solving the MILP problem.