Network Intent Decomposition and Optimization for Energy-Aware Radio Access Network

TL;DR

This paper introduces a novel framework for decomposing and optimizing network intents in 6G energy-aware radio access networks, enhancing automation and reducing energy consumption through intent modeling and deep reinforcement learning.

Contribution

It presents a comprehensive intent representation framework, a novel intent decomposition algorithm, and a deep Q-network-based optimization scheme for energy-efficient 6G networks.

Findings

The intent decomposition algorithm reduces conflict analysis time.

Deep Q-network scheme improves energy efficiency.

Framework enhances automation in network management.

Abstract

With recent advancements in the sixth generation (6G) communication technologies, more vertical industries have encountered diverse network services. How to reduce energy consumption is critical to meet the expectation of the quality of diverse network services. In particular, the number of base stations in 6G is huge with coupled adjustable network parameters. However, the problem is complex with multiple network objectives and parameters. Network intents are difficult to map to individual network elements and require enhanced automation capabilities. In this paper, we present a network intent decomposition and optimization mechanism in an energy-aware radio access network scenario. By characterizing the intent ontology with a standard template, we present a generic network intent representation framework. Then we propose a novel intent modeling method using Knowledge Acquisition in automated Specification language, which can model the network ontology. To clarify the number and types of network objectives and energy-saving operations, we develop a Softgoal Interdependency Graph-based network intent decomposition model, and thus, a network intent decomposition algorithm is presented. Simulation results demonstrate that the proposed algorithm outperforms without conflict analysis in intent decomposition time. Moreover, we design a deep Q-network-assisted intent optimization scheme to validate the performance gain.