A Slice Admission Policy Based on Reinforcement Learning for a 5G Flexible RAN

We present a slice admission policy based on reinforcement learning able to maximize the profit of an infrastructure provider. Results show that when tenants request slices with different latency requirements, the proposed policy outperforms benchmark heuristics by up to 54.5%.

[1]  L. Wosinska,et al.  Priority-Aware Service Orchestration Using Big Data Analytics for Dynamic Slicing in 5G Transport Networks , 2017, 2017 European Conference on Optical Communication (ECOC).

[2]  Lena Wosinska,et al.  Dynamic slicing approach for multi-tenant 5G transport networks [invited] , 2018, IEEE/OSA Journal of Optical Communications and Networking.

[3]  Lena Wosinska,et al.  Machine Learning Aided Orchestration in Multi-tenant Networks , 2018, 2018 IEEE Photonics Society Summer Topical Meeting Series (SUM).

[4]  Srikanth Kandula,et al.  Resource Management with Deep Reinforcement Learning , 2016, HotNets.