Autonomous Resource Allocation for QoS Enhancement in Next-Generation Heterogeneous Wireless Systems

Abstract

The network heterogeneity, high dynamism of traffic demands, channel conditions and diverse Quality-of-Service (QoS) requirements in Beyond Fifth-Generation (B5G) and Sixth-Generation (6G) heterogeneous wireless networks create new difficulties for autonomous resource allocation. Static and reactive methods are not adaptable enough, and current learning based methods typically do not have a built-in mechanism for predicting, multi-objective, and temporally stable resource decisions. To fill this, this paper introduces a hybrid Decision Support System (DSS) for autonomous multi-objective resource allocation with Multi-Objective Optimization (MOOP), Deep Reinforcement Learning (DRL), Predictive Environment Model (PEM) and Stability-Aware Optimization (SAO) under a single decision-making framework. The proposed DSS integrates available network observation data with prediction of environmental conditions and is able to automatically infer resource-allocation decisions based on multiple and possibly competing criteria. DRL enables adaptive policy learning, PEM enables proactively making decisions for the future network state, MOOP enables balancing throughput, latency, fairness and energy efficiency, and SAO enables temporal consistency between successive allocation decisions. The framework also considers Signal-to-Interference-plus-Noise Ratio (SINR) and data rate to determine the quality of communication-links. A large number of simulation experiments are performed to compare the proposed DSS with static, reactive and conventional DRL-based resource-allocation methods. It shows that the results are better in the following aspects: Communication efficiency, Resource utilization, Latency, Fairness, and Energy efficiency. Convergence and temporal analyses also show quicker and more regular decision behavior and the SINR and data-rate analyses confirm better communication-link quality under the assumed operational conditions. As a whole, the results indicated that adding AL, PEWA, multi-objective optimization, and stability-aware decision-making to a DSS is a good way to autonomously manage the use of resources in a dynamic B5G and 6G wireless heterogeneous network while maintaining a balance between the conflicting goals.