The Cloud Continuum Framework (CCF) extends computing capabilities across near-edge, far-edge, and extremeedge nodes beyond the traditional edge to meet the diverse performance demands of emerging 6G applications. While Deep Reinforcement Learning (DRL) has demonstrated potential in automating Virtual Network Function (VNF) migration by learning optimal policies, centralized DRL-based orchestration faces challenges related to scalability and limited visibility in distributed, heterogeneous network environments. To address these limitations, we introduce MARC-6G (Multi-Agent Reinforcement Learning for Distributed Context-Aware Service Function Chain (SFC) Deployment and Migration in 6G Networks), a novel framework that leverages decentralized agents for distributed, dynamic, and service-aware SFC placement and migration. MARC-6G allows agents to monitor different portions of the network, collaboratively optimize network control policies via experience sharing, and make local decisions that collectively enhance global orchestration under time-varying traffic conditions. We show through simulations that MARC-6G improves SFC deployment efficiency, reduces migration costs by $\mathbf{3 4 \%}$, and lowers energy consumption by $\mathbf{1 2. 5 \%}$ compared to the state-of-the-art centralized DRL baseline.

In this work, we aim to address the challenge of slice provisioning in edge-based mobile networks. We propose a solution that learns a service function chain placement policy for Network Slice Requests, to maximize the request acceptance rate, while minimizing the average node resource utilization. To do this, we consider a Hierarchical Multi-Armed Bandit problem and propose a two-level hierarchical bandit solution which aims to learn a scalable placement policy that optimizes the stated objectives in an online manner. Simulations on two real network topologies show that our proposed approach achieves 5% average node resource utilization while admitting over 25% more slice requests in certain scenarios, compared to baseline methods.

The orchestrators’ deployment problem presents numerous challenges in 6G Network Radio Access Networks due to their large-scale, dynamic conditions, and variable user demands. Most works propose single- or hierarchical-orchestrator solutions, which offer poor resiliency, high signaling overhead, and slow adaptation to variable network dynamics. To tackle these challenges, we propose an online, data-driven, fully decentralized, Multi-Agent Reinforcement Learning (MARL)-based, self-organization orchestrator deployment system for 6G networks, which jointly optimizes the tradeoff between user throughput and fairness, based on time-varying system conditions. In the proposed approach, a flexible variable number of decentralized, cooperative, peer self-organization agents autonomously adapt their associated orchestrator’s deployment location and activity to optimize network operation, without requiring centralized coordination. Simulations show improvements of up to 77% in user throughput compared to Hierarchical and Single Orchestrator baselines in a broad range of realistic scenarios.

Open-Radio Access Network (O-RAN) facilitates the scalability of cellular networks by introducing a RAN Intelligent Controller (RIC) component whose functions can be flexibly distributed over large-scale 6G networks. Artificial Intelligence (AI) is effective in optimizing RIC placement in 6G O-RAN, mitigating the limited adaptability of non-data-driven methods in complex time-varying network conditions. However, the centralized orchestration of current approaches for RIC placement hinders scalability. This work introduces a data-driven DEcentralized Reinforced RAN Intelligent Controller orchestration (DERRIC) method for 6G networks, leveraging the online learning capabilities of decentralized multi-agent Reinforcement Learning (RL) orchestration to solve the RAN Intelligent Controller Placement Problem (CPP). DERRIC is a two-layer network management scheme with decentralized orchestrators that adapt to network conditions, deploy controllers, and allocate resources. These orchestrators manage distributed controllers to optimize RAN parameters, such as user transmission power. DERRIC's main goal is to increase the system's overall user Packet Delivery Ratio (PDR) by optimal controller deployment and operation. Optimal controller deployment reduces controller-user latency and accelerates user-transmission-power control decisions, leading to further enhancement to user PDR. We show that DERRIC reduces the controller-user latency and power consumption by up to 66% and 29% and increases user PDR by up to 14% compared to state-of-the-art baselines in a broad range of simulated scenarios.

Federated Learning (FL) is widely applied in privacy-sensitive domains, such as healthcare, finance, and education, due to its privacy-preserving properties. However, implementing FL in dynamic wireless networks poses substantial communication challenges. Central to these challenges is the need for efficient communication strategies that can adapt to fluctuating network conditions and the growing number of participating devices, which can lead to unacceptable communication delays. In this article, we propose Stochastic Client Selection for Tree All-Reduce Federated Learning (CSTAR-FL), a novel approach that combines a probabilistic User Device (UD) selection strategy with a tree-based communication architecture to enhance communication efficiency in FL within densely populated wireless networks. By optimizing UD selection for effective model aggregation and employing an efficient data transmission structure,CSTAR-FLsignificantly reduces communication time and improves FL efficiency. Additionally, our approach ensures high global model accuracy under scenarios where data distribution is heterogeneous from User Device (UD)s. Extensive simulations in dynamic wireless network scenarios demonstrate thatCSTAR-FLoutperforms existing state-of-the-art methods, reducing model convergence time by up to 40% without losing the global model accuracy. This makesCSTAR-FLa robust solution for efficient and scalable FL deployments in high-density environments.

The next generation of Virtual Reality (VR) applications is expected to provide advanced experiences through Six Degree-of-Freedom (6DoF) technology. However, 6DoF VR applications require latency and throughput guarantees. This article presents a novel intra-domain routing algorithm with throughput guarantees for minimizing the overall end-to-end (E2E) latency for all flows deployed in the network.We investigate the Joint Flow Allocation (JFA) problem to find paths for all flows in a network such that it determines the optimal path for each flow in terms of throughput and latency. The JFA problem is NP-hard. We use a mixed integer linear programming to model the system, along with a heuristic, Flow Latency and Throughput Aware Sensitive Routing (FLATWISE), which is one order of magnitude faster than optimally solving the JFA problem. FLATWISE introduces an adaptive routing approach that dynamically adjusts the path calculation based on E2E latency. The novelty of FLATWISE lies in its unique ability to precisely tune the routing path by either constraining or relaxing the path criteria to align the E2E latency of the selected path with the latency demands of each VR flow. This approach ensures that the latency of the calculated path approximates the latency of each VR flow, enabling more flexible and efficient network routing to meet diverse latency requirements. Our evaluation considers 6DoF VR application flows, which demand high throughput and ultra-low E2E latency. Extensive simulations demonstrate that FLATWISE significantly reduces flow latency, over-provisioned latency, E2E latency, and algorithm execution time when network flows are processed randomly. FLATWISE improves flow throughput and frame rate compared to related work approaches.

Fifth-generation (5G) mobile networks are expected to support the dynamic provisioning of services with heterogeneous Quality of Service requirements through Network Slicing. However, the uncertainty in the resource requirements of the tenant’s future Network Slice Requests raises the problem of how Network Slices can be admitted onto the mobile network infrastructure. To address this, we investigate the Slice Admission Control problem in virtualization-enabled mobile networks. Specifically, we focus on the scenario in which a controller needs to select an Admission Control policy (or algorithm) based on the patterns of previous Network Slice Requests and formulate such a problem as a Multi-Armed Bandit problem. By leveraging Online Learning (OL), we propose a framework, Drift-AwaRe upper confIdence bOund (DARIO), that adaptively selects and learns the performance of online Slice Admission Control (SAC) policies by monitoring for changes in the underlying patterns of Network Slice Request (NSR) features. We evaluate the performance of our framework in terms of the relative gains in average revenue, acceptance ratio, and average resource utilization when compared to both static and adaptive baselines and show that we outperform the considered baselines for the considered metrics.

The key impediments to deploying deep neural networks (DNN) in IoT edge environments lie in the gap between the expensive DNN computation and the limited computing capability of IoT devices. Current state-of-the-art machine learning models have significant demands on memory, computation, and energy and raise challenges for integrating them with the decentralized operation of heterogeneous and resource-constrained IoT devices. Recent studies have proposed the cooperative execution of DNN models in IoT devices to enhance the reliability, privacy, and efficiency of intelligent IoT systems but disregarded flexible finegrained model partitioning schemes for optimal distribution of DNN execution tasks in dynamic IoT networks. In this paper, we propose DISNET, a distributed micro-split deep learning scheme for heterogeneous dynamic IoT. DISNET accelerates inference time and minimizes energy consumption by combining vertical (layer-based) and horizontal DNN partitioning to enable flexible, distributed, and parallel execution of neural network models on heterogeneous IoT devices. DISNET considers the IoT devices’ computing and communication resources and the network conditions for resource-aware cooperative DNN Inference. Experimental evaluation in dynamic IoT networks shows that DISNET reduces the DNN inference latency and energy consumption by up to 5.2× and 6×, respectively, compared to two state-of-the-art schemes without loss of accuracy.

The next generation of Virtual Reality (VR) applications is expected to provide advanced experiences through Six Degrees of Freedom (6DoF) content, which requires higher data rates and ultra-low latency. In this article, we refactor 6DoF VR applications into atomic services to increase the computing capacity of VR systems aiming to reduce the end-to-end (E2E) of 6DoF VR applications. Those services are chained and deployed across Head-Mounted Displays (HMDs) and Multi-access Edge Computing (MEC) servers in high mobility scenarios over realedge network topologies. We investigate the Distributed Service Chain Problem (DSCP) to find the optimal service placement of services from a service chain such that its E2E latency does not exceed 5 ms. The DSCP problem is NP-hard. We provide an integer linear program to model the system, along with a heuristic, namely disTributed sErvice chaiN orchEstraTor (TENET), which is one order of magnitude faster than optimally solving the DSCP problem. We compare TENET to DSCP implementation and well-known service migration algorithms in terms of E2E latency, power consumption, video resolution selection based on E2E latency, context migrations, and execution time. We observe a significant reduction of E2E latency and gains in more advanced video resolution selection and accepted context service migrations when using TENET’s deployment strategy on VR services.

Network Slicing has emerged as a powerful technique to enable cost-effective, multi-tenant communications and services over a shared physical mobile network infrastructure. One major challenge of service provisioning in slice-enabled networks is the uncertainty in the demand for the limited network resources that must be shared among existing slices and potentially new Network Slice Requests. In this paper, we consider admission control of Network Slice Requests in an online setting, with the goal of maximizing the long-term revenue received from admitted requests. We model the Slice Admission Control problem as an Online Multidimensional Knapsack Problem and present two reservation-based policies and their algorithms, which have a competitive performance for Online Multidimensional Knapsack Problems. Through Monte Carlo simulations, we evaluate the performance of our online admission control method in terms of average revenue gained by the Infrastructure Provider, system resource utilization, and the ratio of accepted slice requests. We compare our approach with those of the online First Come First Serve greedy policy. The simulation's results prove that our proposed online policies increase revenues for Infrastructure Providers by up to 12.9 % while reducing the average resource consumption by up to 1.7% In particular, when the tenants' economic inequality increases, an Infrastructure Provider who adopts our proposed online admission policies gains higher revenues compared to an Infrastructure Provider who adopts First Come First Serve.

Federated Learning (FL) is gaining popularity in trajectory prediction field due to its privacy-preserving and scalability capabilities. However, deploying FL on resource-constrained devices and varying wireless network conditions can be challenging. Moreover, the design of FL’s distributed neural architectures is complex requiring expert knowledge. To address these limitations, we propose the network-adaptive FEDerated learning for reinFORCEd mobility prediction (FedForce) system. FedForce uses reinforcement learning to design a transformer neural network that jointly optimizes prediction accuracy, training time, and transmission time based on the unique features of the mobility dataset, the client’s computing capacity, and the available network throughput. FedForce achieves an average displacement error of 0.20m on the ETH+UCY dataset and 76% accuracy on the Orange dataset (−0.02m and 10% better than the best-performing existing baselines, respectively), while reducing FL training and transmission time by a factor of 3. FedForce can save up to 80% of computational resources and 96% of communication overheads with negligible accuracy decrease.

In dynamic settings, fully distributed gossip-based learning schemes have recently gained interest due to their better scalability, robustness, and enhanced privacy protection compared to server-based architectures. However, existing approaches to their performance characterization either assume stable connectivity among nodes or are ad-hoc for specific trace-based mobility patterns. Thus, in dynamic settings, there is currently a poor understanding of the conditions under which gossip-based learning schemes are feasible, and of their main performance tradeoffs. In this work, we start addressing this issue by performing a first baselining of Gossip Learning (GL) on random Time-Varying Graphs (TVG), to get a first-order characterization of their main performance patterns in dynamic settings. The use of random TVG enables a fine-grained and accurate characterization of GL effectiveness as a function of the main system parameters while abstracting from scenario-specific features of patterns of communication and mobility (e.g., induced by road grids or measured mobility traces). Our results suggest that GL schemes are robust to node mobility and comparable in accuracy and convergence speed to Federated Learning architectures, over a wide range of operational conditions. We show that the final model accuracy is robust against data dispersion across nodes as well as against very low rates of exchanges across nodes.

The deployment of Internet of Things (IoT) temperature sensors in urban areas is essential for the monitoring and understanding of the thermal environment. However, accurate temperature measurements can be compromised by factors such as direct sunlight, leading to overheating and inaccurate readings. We propose a Machine Learning-based approach that addresses this challenge by dynamically ventilating the sensor environment using small fans, enabling accurate and energy-efficient temperature measurements. This paper focuses on two interconnected problems: predicting steady-state temperature using a limited window of initial temperature measurements and investigating the impact of ventilation time. We employ various DNNs suitable for low-power IoT sensor devices to predict temperature using multivariate time series from different sensors and compare their accuracy. Furthermore, we highlight the tradeoff between prediction accuracy, which is correlated to the length of the observed input sequence, and energy consumption dependent on ventilation time. By adopting advanced prediction techniques, we can develop efficient IoT systems for accurate and energy-efficient environment monitoring in smart cities.

Positioning estimations of wireless sensors can be enhanced via sensor collaboration. To enable this, various methods have been proposed; yet, most do not leverage the entire collective knowledge, which also involves the estimation’s uncertainty. In this article, we introduce Anchor-free Ranging-Likelihood-based Cooperative Localization (ARLCL); a novel anchor-free and technology-agnostic localization algorithm that utilizes inter-exchanged ranging signals from sensors to enable their simultaneous positioning. Ranging technologies with easy-to-model propagation properties, such as UWB or LiDAR are among the first beneficiaries that ARLCL is targeting. To examine its applicability, however, even to signals that are noisier and often unsuitable for ranging, we assess ARLCL with real-world BLE RSS measurements. At the same time, we consider deployments that typically induce flip-ambiguity, being a major problem in cooperative localization. We provide an extensive comparison against the most widely-adopted optimization method (Mass-Spring) but also against the recent likelihood-based approach (Maximum Likelihood - Particle Swarm Optimization). The results showed that ARLCL outperformed the baselines in almost all scenarios. Our gain in positioning accuracy is also found to be positively correlated to both the swarm’s size and the signal’s quality, reaching an improvement of 40%.

Large Machine Learning (ML) models require considerable computing resources and raise challenges for integrating them with the decentralized operation of heterogeneous and resource-constrained Internet of Things (IoT) devices. Running ML tasks on the cloud can introduce network delay, throughput, and privacy concerns, whereas running ML tasks on IoT devices is penalized by their constrained resources. For this reason, recent research proposed cooperative execution of ML tasks over IoT networks but disregarded resource variability and the IoT devices’ energy constraints simultaneously. In this paper, we propose Early Exit of Computation (EEoC), an adaptive, energy-efficient, low-latency inference scheme over IoT networks. EEoC adaptively distributes the inference computation load between the IoT device and the edge server, based on estimated communication and computation resources, to jointly minimize prediction latency and energy consumption. We evaluate our solution’s latency and energy profile on a real testbed running two widely used neural networks. Results show that EEoC can reduce latency and energy consumption up to 24.6% and 46.5%, respectively, compared to other state-of-the-art solutions without sacrificing accuracy.

Perceptual studies show that the Quality of Service (QoS) of large-scale Mobile Virtual Reality (MVR) applications is positively correlated to video frame rate and the duration of the immersive experience. These metrics depend on the sum of computation and network latency needed to generate and deliver a video frame to the Head-Mounted Display (HMD) and the power consumption on the HMD. Recent research shows that Multi-access Edge Computing (MEC) can support mobile HMDs to reduce their computing latency, but its potential to maintain the acceptable end-to-end (E2E) latency and reduce power consumption under high mobility conditions remains unexplored. This paper proposes Service Chaining Graph (SCG), an orchestrator to split VR applications into atomic services and deploy them across HMDs and MEC servers according to an optimization problem that aims to jointly minimize latency and energy consumption. Through simulations, we show that SCG reduces E2E latency by up to 74% in three high-mobility user-dense scenarios compared to four widely used service-migration strategies against a moderate increase in power consumption. Unlike other approaches, SCG can find a balance between average latency and energy consumption by migrating services between MEC servers and HMDs according to a policy depending on application requirements.

Anticipating future locations of mobile users plays a pivotal role in intelligent services supporting mobile networks. Predicting user trajectories is a crucial task not only from the perspective of facilitating smart cities but also of significant importance in network management, such as handover optimization, service migration, and the caching of services in a mobile and edge-computing network. Convolutional Neural Networks (CNNs) have proven to be successful to tackle the forecasting of mobile users’ future locations. However, designing effective CNN architectures is challenging due to their large hyper-parameter space. Reinforcement Learning (RL)-based Neural Architecture Search (NAS) mechanisms have been proposed to optimize the neural network design process, but they are computationally expensive and they have not been used to predict user mobility. In large urban scenarios, the rate at which mobility information is generated makes it a challenge to optimize, train, and maintain prediction models for individual users. However, considering that user trajectories are not independent, a common trajectory-prediction model can be built and shared among a set of users characterized by similar mobility features. In the present work, we introduce Reinforcement Convolutional Transfer Learning (RC-TL), a CNN-based trajectory-prediction system that clusters users with similar trajectories, dedicates a single RL agent per cluster to optimize a CNN neural architecture, trains one model per cluster using the data of a small user subset, and transfers it to the other users in the cluster. Experimental results on a large-scale dataset show that our proposed RL-based CNN achieves up to 12% higher trajectory-prediction accuracy, with no training speed reduction, over other state-of-the-art approaches on a large-scale, real-world mobility dataset. Moreover, RC-TL’s clustering strategy saves up to 90% of the computational resources needed for training compared to single-user models, in exchange for a 3% accuracy reduction.

Predicting mobile users' trajectories accurately is essential for improving the performance of wireless networks and autonomous systems. In this paper, we tackle the problem of tra-jectory prediction in a multi-agent scenario where the social inter-action among users is taken into consideration. We propose Intra-Cluster Reinforced Social Transformer (INTRAFORCE), a novel system to design and train Social-Transformer neural networks that learn the spatio-temporal interactions among neighboring mobile users and predict their joint future trajectories. Unlike state-of-the-art social-aware trajectory predictors that either miss the large-distance interactions or are computationally expensive due to the pooling of all users' interactions, INTRAFORCE clusters users with similar trajectories and learns their interactions. INTRAFORCE performs Neural Architecture Search to optimize each transformer's architecture to fit each cluster's user mobility features using Reinforcement Learning. Through experimental validation, we show that INTRAFORCE outperforms several state-of-the-art trajectory predictors on five widely used small-scale pedestrian mobility datasets and one large-scale privacy-oriented cellular mobility dataset by achieving lower prediction error. training time, and computational complexity.

Distributed training of Machine Learning models in edge Internet of Things (IoT) environments is challenging because of three main points. First, resource-constrained devices have large training times and limited energy budget. Second, resource heterogeneity of IoT devices slows down the training of the global model due to the presence of slower devices (stragglers). Finally, varying operational conditions, such as network bandwidth, and computing resources, significantly affect training time and energy consumption. Recent studies have proposed Split Learning (SL) for distributed model training with limited resources but its efficient implementation on the resource-constrained and decentralized heterogeneous IoT devices remains minimally explored. We propose Adaptive REsource-aware Split-learning (ARES), a scheme for efficient model training in IoT systems. ARES accelerates training in resource-constrained devices and minimizes the effect of stragglers on the training through device-targeted split points while accounting for time-varying network throughput and computing resources. ARES takes into account application constraints to mitigate training optimization tradeoffs in terms of energy consumption and training time. We evaluate ARES prototype on a real testbed comprising heterogeneous IoT devices running a widely-adopted deep neural network and dataset. Results show that ARES accelerates model training on IoT devices by up to 48% and minimizes the energy consumption by up to 61.4% compared to Federated Learning (FL) and classic SL, without sacrificing model convergence and accuracy.

In modern networks, edge computing will be responsible for processing and learning from the critical network- and user-generated data, such as wireless link usage, mobility information, application requests, and many others. The presence of Artificial Intelligence-based (AI) applications at the edge of the network will enable the network to predict necessary user behavior and its impact on network infrastructure, such as base station overloading. One of the main strategies for offloading users and base stations is to deploy UAV base stations, or flying base stations, which can dynamically provide service and connectivity. In this article, we introduce a framework for distributed learning over Multi-access Edge Computing (MEC), which manages data applications in a fully distributed setting across edge servers, thus reducing the cost of collecting user information in a centralized server. We couple the proposed distributed learning with a novel similarity metric for user trajectories, which can aggregate neural network models with similar costs as other model aggregation techniques. However, the aggregation technique can achieve much higher accuracy. Furthermore, we apply the proposed distributed learning scheme to manage and deploy flying base stations to areas that experience high demand or poor user connectivity, thus optimizing connectivity in terms of user satisfaction, delay, and network throughput.

Although many target applications in VANETs are information-centric, the performance of Named Data Networking (NDN) in vehicular ad-hoc networks is severely hampered by persistent network partitioning, typical of many vehicular scenarios. Existing approaches try to address this issue by relying on opportunistic communications. However, they leave open the crucial issue of how to guarantee content persistence and tight QoS levels while optimizing the resource utilization in the vehicular environment. In this work we propose DeepNDN, a communication scheme based on the joint application of NDN and of probabilistic spatial content caching, which enables content retrieval in fragmented and dynamic network topologies with tight delay constraints. We present a data-based approach to DeepNDN management, based on locally modulating content replication and delivery in order to achieve a target hit ratio in a resource-efficient manner. Our management algorithm employs a Convolutional Neural Network (CNN) architecture for effectively capturing the complex relations between spatio-temporal patterns of mobility and content requests and DeepNDN performance. Its numerical assessment in realistic, measurement-based scenarios suggest that our management approach achieves its target set goals while outperforming a set of reference schemes.

5G-enabled vehicular networks will soon allow their users to exchange safety and non-safety related information over heterogeneous communication interfaces. Routing vehicular data flows over multi-hop Vehicle-to-Vehicle (V2V) communications is one of the hardest challenges in vehicular networking, and it has been tackled in literature by using distributed algorithms. The distributed approach has shown significant inefficiencies in such dynamic vehicular scenarios, mainly due to poor network congestion control. To overcome the complexity of the envisioned architecture, and the inefficiency of distributed routing algorithms, we hereby propose to leverage the coordination capabilities of Software-Defined Networking (SDN) to determine optimal V2V multi-hop paths and to offload traffic from the Vehicle-to-Infrastructure-to-Vehicle (V2I2V) to the V2V communications, using both cellular and Wi-Fi technologies. In order to achieve this goal, we propose Multi-Flow Congestion-Aware Routing (MFCAR), a centralized routing algorithm that relies on graph theory to choose short and uncongested V2V paths. Realistic simulations prove that MFCAR outperforms well- established centralized routing algorithms (e.g. Dijkstra's) in terms of Packet Delivery Ratio (PDR), goodput and average packet delay, up to a five-fold performance gain.

Floating Content (FC) has recently been proposed as an attractive application for mobile networks, such as VANETs, to operate opportunistic and distributed content sharing over a given geographic area, namely Anchor Zone (AZ). FC performances are tightly dependent on the AZ size, which in literature is classically chosen by the node that generates the floating message. In the present work, we propose a method to improve FC performances by optimizing the AZ size with the support of a Software Defined Network (SDN) controller, which collects mobility information, such as speed and position, of the vehicles in its coverage range.