Machine Learning Based Network Slicing and Resource Allocation for Electric Vehicles (EVs)

Abstract

The introduction of electric vehicles (EVs) brings out various challenges like deploying more charging stations, building its supportive infrastructure, managing the EVs and their various resource requests. In order to maintain all the resource requests of the EVs, network slicing is used which provides an efficient way to satisfy various use case demands of the EVs. In this work, we perform network slicing that partitions the physical network into three slices, infotainment and safety message slices that belong to downlink communication and charge state information slice that belongs to uplink communication. If there are large number of resource requests made by the EVs to the slices, it might lead to collision in the channel. Unsupervised machine learning is performed on the EVs using local scaling as the scaling parameter which handles multi scale data and also performs better clustering. For an efficient communication between the clustered EVs and charging station, slice leaders of every cluster is determined. We have also proposed an algorithm that efficiently performs resource allocation to the EVs to increase the throughput with low latency. Slice leaders forward the resource requests made by the EVs of the respective clusters to the charging station through RSUs and a slice block allocation is performed by giving higher preference to the critical message requests.