Proactive Mobility Load Balancing through Interior-point Policy Optimization for Open Radio Access Networks

The future B5G/6G wireless networks are required to intelligently operate and optimize in all ranges of scenarios, however, most of the functions in self-organizing network (SON) architectures are rule-based and reactive, these rigid functions can not keep up with the dynamic nature of the future wireless networks. In this paper, we propose an active load balancing framework considering mobility load balancing (MLB) and mobility robust optimization (MRO), which are two essential functions to ensure a seamless user experience for the future B5G/6G. The proposed method is based on random subspace Bayesian additive regression tree (RS-BART) and interior point policy optimization (IPO). Specifically, user trajectory is firstly forecasted by utilizing RS-BART, and then the joint problem of MLB and MRO is further solved by IPO. Simulations based on open radio access network (O-RAN) reveal that the proposed method reduces load deviation and radio link failure ratio by 40% and 50% compared to the previous ones.