Vehicular Multi-slice Optimization in 5G: Dynamic Preference Policy using Reinforcement Learning

Network slicing, as an effective way of using heterogeneous network resources, is widely used in today's radio access network (RAN). However, because of the greater randomness of equipment capacity and mobility, the existing allocation schemes of network slicing do not make use of existing resources effectively. In this regard, this paper studies how to improve the efficiency of network slicing utilization in one base station (BS) area through deep Q-learning's allocation strategy. First, we propose an allocation strategy that uses the preference matrix to prioritize all network slices. Then, with low coupling, a real-time updating Q-learning model is developed to calculate the preference matrix. Finally, we demonstrate through simulation that our proposal can improve the efficiency of service delivery in a heterogeneous wireless network region.