Artificial Intelligence/Machine Learning (AI/ML) methods for network automation have been widely adopted by the standardization community since 5G, due to its capability to learn patterns from continuously changing network data. Presently, AI in the current 3GPP releases is specified to be deployed as an overlay or add-on feature on top of conventional Network Functions (NFs), emitting a few limitations. First, NFs need to consult the AI-overlay to derive and execute intelligent actions, thereby affecting system and AI performance and lacking intelligent autonomy. Second, the capability of NFs and NF instances to added or removed based on changing network conditions is severely hindered, resulting in limited flexibility of NF deployment. To mitigate these limitations, this article explores standard enablers for evolution from AI-overlay in 5G to native AI in 6G, where AI will be an embedded functionality of 6G-NFs. Via envisioned standards enablers-AI/ML Data Plane and Control Plane-we conceive an Integrated Distributed AI Automation Layer (INTDAI), where each 6G-NF is equipped with the capabilities of an Intelligent Agent (IA)-independent data collection, intelligent decision-making and independent action execution. A special 6G-NF, the AI Controller, controls and manages a group of IAs to ensure overall system performance. The behavior and benefits of INTDAI are illustrated using a dedicated network automation use case, Dynamic User Plane Task Migration. By means of numerical analysis, we demonstrate the gain of INTDAI, motivating the need for future standardization work for AI-native 6G.

In-Network Computing (INC) is a currently emerging paradigm. Realizing INC in 6G networks could mean that user plane entities (UPEs) carry out computations on packets while transmitting them. These computations may have specific requirements in terms of their completion time. In case of high compute pressure at one UPE, migrating computations to another UPE may be beneficial, in order to avoid exceeding the completion time requirement. Centralized migration approaches suffer from increased signaling and are prone to react too slow. Therefore, this paper investigates the applicability of distributed intelligence to tackle the problem of compute task migration in the 6G user plane. Each UPE is equipped with an intelligent agent, enabling autonomous decisions on whether computations should be migrated to another UPE. To enable the intelligent agents to learn and apply an optimal task migration policy, we investigate and compare of two state-of-the-art Deep Reinforcement Learning (DRL) approaches: Advantage Actor-Critic (A2C) and Double Deep Q-Network (DDQN). We show, via simulations, that the performance of both solutions, in terms of the percentage of tasks exceeding their completion time requirement, is near-optimal and training A2C is at least 60% faster.Â

Our solution evaluation is based on synthetic data-sets generated by our internally-developed simulation platform. This means we do not use any stand-alone, external or internal data-sets. Our platform code is not public, but we describe in detail the design of the platform and the data-set generator module, for other researchers to reproduce our results. The code is written in 100% Python, leveraging object-oriented approaches for code reusability.  This article has been accepted for publication in IEEE Transactions on Network and Service Management. This is the author’s version which has not been fully edited and content may change prior to final publication. Citation information: DOI 10.1109/TNSM.2023.3333875 © 2023 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See https://www.ieee.org/publications/rights/index.html for more information.