The Network Digital Twins (NDTs) have become a frontier thanks to their real-time monitoring, analysis, prediction, and optimisation capabilities. However, their architecture has not been designed for the scale of next-generation networks that will ensure ubiquitous connectivity. At this, different NDT applications may require distinct levels of quality of service requirements to be met. With increasing size and heterogeneity in the networks, the processing load at the NDTs escalates, and these requirements may not be met. Therefore, we propose a microservice-based architecture with application-oriented slicing. Here, we present the scaling methodology, which enables scaling at both microservice and slice levels. Then, we evaluate the throughput, delay, and quality of service requirement violation rate metrics under two scenarios. Thanks to the slicing approach with microservice-based implementation, the end-to-end delay and the QoS requirement violation rate are reduced while having higher throughput performance.

Although the emergence of 6G IoT networks has accelerated the deployment of enhanced smart city services, the resource limitations of IoT devices remain as a significant problem. Given this limitation, meeting the low-latency service requirement of 6G networks becomes even more challenging. However, existing 6G IoT management strategies lack real-time operation and mostly rely on discrete actions, which are insufficient to optimise energy consumption. To address these, in this study, we propose a Digital Twin (DT)-guided energy management framework to jointly handle the low latency and energy efficiency challenges in 6G IoT networks. In this framework, we provide the twin models through a distributed overlay network and handle the dynamic updates between the data layer and the upper layers of the DT over the Real-Time Publish Subscribe (RTPS) protocol. We also design a Reinforcement Learning (RL) engine with a novel formulated reward function to provide optimal data update times for each of the IoT devices. The RL engine receives a diverse set of environment states from the What-if engine and runs Deep Deterministic Policy Gradient (DDPG) to output continuous actions to the IoT devices. Based on our simulation results, we observe that the proposed framework achieves a 37% improvement in 95th percentile latency and a 30% reduction in energy consumption compared to the existing literature.

Even though the emergence of 6G IoT systems has accelerated the deployment of hyper-connected networks, the inherent resource limitations of IoT sensors remain a significant problem. In addition, maintaining energy efficiency and low response times in such environments has become more challenging. However, the existing management methods often lack the real-time adaptability and intelligence to optimize energy consumption in 6G IoT networks. To tackle this, we propose a DT-based collaborative management consisting of a multi-agent twin layer, a collaboration protocol and a Reinforcement Learning (RL)-based learner model. In the multi-agent twin layer, each physical network sensor is modelled as an individual agent for enhanced granularity in the management. The collaboration protocol ensures information sharing among the sensors and, thus, lowers response times. Furthermore, in the learner model, we utilize a multi-agent Deep Deterministic Policy Gradient (MADDPG) algorithm to optimise actions according to the novel energy-aware reward function. According to our simulation results, the proposed DT-based collaborative management surpasses the traditional method by 27 % for longer battery levels and 65 % more rapid responses.

Digital Twins (DT) have become crucial to achieve sustainable and effective smart urban solutions. However, current DT modelling techniques cannot support the dynamicity of these smart city environments. This is caused by the lack of right-time data capturing in traditional approaches, resulting in inaccurate modelling and high resource and energy consumption challenges. To fill this gap, we explore spatiotemporal graphs and propose the Reinforcement Learning-based Adaptive Twining (RL-AT) mechanism with Deep Q Networks (DQN). By doing so, our study contributes to advancing Green Cities and showcases tangible benefits in accuracy, synchronisation, resource optimization, and energy efficiency. As a result, we note the spatiotemporal graphs are able to offer a consistent accuracy and 55% higher querying performance when implemented using graph databases. In addition, our model demonstrates right-time data capturing with 20% lower overhead and 25% lower energy consumption.

Complete topology discovery is the most important MAC service in broadband networks but it holds spatial and temporal complexities for the network-wide, i.e in the data link layer it requires considerable time amount to update all link status management information. Moreover, many service providers are complaining about high operational time and resource usage in the complete topology discovery process. Additionaly, consuming high resources leads to a huge amount of management traffic on the links. At this point, partial topology discovery arises as an alternative solution as a more efficient MAC service for the next generation broadband networks to reduce complexities and maintain smooth functioning. However, manual execution of partial topology discovery leads to risky situations for the network arising from human intervention. Therefore, in this paper, we propose an AI-driven partial topology discovery approach to maintain a global MAC service which serves both physical and virtual connections in a broadband network. Besides, with this approach, we not only preserve the network resources but also have the ability of forecasting the device-based network statistics. For this aim, we use Hidden Markov Model in order to estimate the path to be discovered regarding the arrived log patterns of the devices. Thanks to the partial path estimation, we eliminate the usage of every node in the discovery and achieve up-to-date topology information more rapidly. Consequently, according to our simulations, we succeed in a significant reduction in the number of nodes used by 60%, required time to have up-to-date topology by 35%. And finally, as a consequence of using less amount of nodes, we reduce the management traffic on the links on average 50%.