As augmented and virtual reality evolve, achieving seamless synchronization between physical and digital realms remains a critical challenge, especially for real-time applications where delays affect the user experience. This paper presents MetaLore, a Deep Reinforcement Learning (DRL) based framework for joint communication and computational resource allocation in Metaverse or digital twin environments. MetaLore dynamically shares the communication bandwidth and computational resources among sensors and mobile devices to optimize synchronization, while offering high throughput performance. Special treatment is given in satisfying end-to-end delay guarantees. A key contribution is the introduction of two novel Age of Information (AoI) metrics: Age of Request Information (AoRI) and Age of Sensor Information (AoSI) — integrated into the reward function to enhance synchronization quality. An open source simulator has been extended to incorporate and evaluate the approach. The DRL solution is shown to achieve the performance of full-enumeration brute-force solutions by making use of a small, task-oriented observation space of two queue lengths at the network side. This allows the DRL approach the flexibility to effectively and autonomously adapt to dynamic traffic conditions.

Recently, deep autoencoders have gained traction as a powerful method for implementing goal-oriented semantic communications systems. The idea is to train a mapping from the source domain directly to channel symbols, and vice versa. However, prior studies often focused on rate-distortion tradeoff and transmission delay, at the cost of increasing end-to-end complexity and thus latency. Moreover, the datasets used are often not reflective of real-world environments, and the results were not validated against real-world baseline systems, leading to an unfair comparison. In this paper, we study the problem of remote camera pose estimation and propose AdaSem, an adaptive semantic communications approach that optimizes the tradeoff between inference accuracy and end-to-end latency. We develop an adaptive semantic codec model, which encodes the source data into a dynamic number of symbols, based on the latent space distribution and the channel state feedback. We utilize a lightweight model for both transmitter and receiver to ensure comparable complexity to the baseline implemented in a real- world system. Extensive experiments on real-environment data show the effectiveness of our approach. When compared to a real implementation of a client-server camera relocalization service, AdaSem outperforms the baseline by reducing the end-to-end delay and estimation error by over 75% and 63%, respectively.

Network slicing is a key technique in 5G and beyond for efficiently supporting diverse services. Many network slicing solutions rely on deep learning to manage complex and high-dimensional resource allocation problems. However, deep learning models suffer limited generalization and adaptability to dynamic slicing configurations. In this paper, we propose a novel frame-work that integrates constrained optimization methods and deep learning models, resulting in strong generalization and superior approximation capability. Based on the proposed framework, we design a new neural-assisted algorithm to allocate radio resources to slices to maximize the network utility under inter-slice resource constraints. The algorithm exhibits high scalability, accommodating varying numbers of slices and slice configurations with ease. We implement the proposed solution in a system-level network simulator and evaluate its performance extensively by comparing it to state-of-the-art solutions including deep reinforcement learning approaches. The numerical results show that our solution obtains near-optimal quality-of-service satisfaction and promising generalization performance under different network slicing scenarios.

Transfer learning has emerged to address various challenges in machine learning by improving the efficiency and effectiveness of the learning process. We explore transfer learning in the context of reinforcement learning and apply it to optimize the resource allocation for network slicing for different industrial environments. We propose two transfer learning techniques and investigate whether the knowledge gained by a pre-trained deep reinforcement learning agent for an industrial environment can be leveraged when the environment changes. Experiments demonstrate that our proposed approach yields higher performance when compared with the training of a new agent for the changed environment by reducing training convergence time, improving performance stability and achieving higher reward.

Deep reinforcement learning (DRL) has been in-creasingly employed to handle the dynamic and complex re-source management in network slicing. The deployment of DRL policies in real networks, however, is complicated by heterogeneous cell conditions. In this paper, we propose a novel transfer learning (TL) aided multi-agent deep reinforcement learning (MADRL) approach with inter-agent similarity analysis for inter-cell inter-slice resource partitioning. First, we design a coordinated MADRL method with information sharing to intelligently partition resource to slices and manage inter-cell interference. Second, we propose an integrated TL method to transfer the learned DRL policies among different local agents for accelerating the policy deployment. The method is composed of a new domain and task similarity measurement approach and a new knowledge transfer approach, which resolves the problem of from whom to transfer and how to transfer. We evaluated the proposed solution with extensive simulations in a system-level simulator and show that our approach outperforms the state-of-the-art solutions in terms of performance, convergence speed and sample efficiency. Moreover, by applying TL, we achieve an additional gain over 27% higher than the coordinated MADRL approach without TL.

The legacy mobility robustness optimization (MRO) in self-organizing networks aims at improving handover performance by optimizing cell-specific handover parameters. However, such solutions cannot satisfy the needs of next-generation network with network slicing, because it only guarantees the received signal strength but not the per- slice service quality. To provide the truly seamless mobility service, we propose a deep reinforcement learning-based slice- aware mobility robustness optimization (SAMRO) approach, which improves handover performance with per-slice service assurance by optimizing slice-specific handover parameters. Moreover, to allow safe and sample efficient online training, we develop a two-step transfer learning scheme: 1) regularized offline reinforcement learning, and 2) effective online fine-tuning with mixed experience replay. System-level simulations show that compared against the legacy MRO algorithms, SAMRO significantly improves slice-aware service continuation while optimizing the handover performance.

Network slicing enables the operator to configure virtual network instances for diverse services with specific requirements. To achieve the slice-aware radio resource scheduling, dynamic slicing resource partitioning is needed to orchestrate multi-cell slice resources and mitigate inter-cell interference. It is, however, challenging to derive the analytical solutions due to the complex inter-cell interdependencies, inter-slice resource constraints, and service-specific requirements. In this paper, we propose a multi-agent deep reinforcement learning (DRL) approach that improves the max-min slice performance while maintaining the constraints of resource capacity. We design two coordination schemes to allow distributed agents to coordinate and mitigate inter-cell interference. The proposed approach is extensively evaluated in a system-level simulator. The numerical results show that the proposed approach with inter-agent coordination outperforms the centralized approach in terms of delay and convergence. The proposed approach improves more than two-fold increase in resource efficiency as compared to the baseline approach.

Real-time camera localization is a key enabler for interactive network service, e.g. visualizing network performance with augmented reality (AR) in user devices. We propose a deep learning and sensor fusion approach for real-time camera localization. A multi-input deep neural network is designed to regress the camera pose from a single image and motion sensor measurements. We perform a comprehensive analysis to find the best choices of input features, loss function, convolutional neural network model, and hyperparameters. We show that by adding features extracted from motion sensor data, our approach significantly outperforms the state-of-the-art visual-based camera localization approaches. In an indoor environment where we conduct a proof-of-concept of the proposed end-to-end AR-supported radio map visualization solution, our camera localization approach achieves an orientation error of 2.5179° and a position error of 0.0222 meters with an inference time lower than 4 ms per frame.

Digital twin technology is a new tool for automatic network planning and optimization. However, for indoor private networks, such as Industry 4.0 supported manufacturing plants, an exact “twin” of the real environment further leads to overly complex radio propagation modeling. We propose a feature and object-based monocular simultaneous localization and mapping (SLAM) algorithm called Simultaneous Localization and Mapping with Object REcognition (SLAMORE), that can reconstruct an “extracted” version of the radio propagation environment by detecting, tracking, localizing, and reconstructing the major obstacles for electromagnetic waves. The embedded convolutional object detector helps recognize and reconstruct major obstacles, and provides reference to estimate the room scale and real world coordinate. We conduct a proof of concept in an office environment and show that the mean absolute percentage errors for room size estimation and object position estimation achieve approximately 2.6-5.7% and 1.8-2.7%, respectively.

Accurate and efficient resource utilization predictions are of vital importance for the future generation of mobile wireless networks. By anticipating network resource demand, the operator can perform proactive resource allocation and predictive network control to improve network resource efficiency. In this paper, we exploit deep and transfer learning algorithms for multi-step resource utilization prediction in radio networks. In particular, we propose long short-term memory network-based architectures with transfer learning for the multi-step prediction task, in order to address scalability, computation time and data storage limitations of current implementations for large-scale networks. We carry out extensive experiments on a dataset collected from an LTE field network. When predicting physical resource block percentage utilization, our approach achieves state of the art results with root mean square error below 12 for a four-hour-ahead prediction, in half of the computation time required by deep learning methods without transfer learning.

Wireless devices need to adapt their transmission power according to the fluctuating wireless channel in order to meet constraints of delay sensitive applications. In this paper, we consider delay sensitivity in the form of strict packet deadlines arriving in a transmission queue. Packets missing the deadline while in the queue are dropped from the system. We aim at minimizing the packet drop rate under average power constraints. We utilize tools from Lyapunov optimization to find an approximate solution by selecting power allocation. We evaluate the performance of the proposed algorithm and show that it achieves the same performance in terms of packet drop rate with that of the Earliest Deadline First (EDF) when the available power is sufficient. However, our algorithm outperforms EDF regarding the trade-off between packet drop rate and average power consumption.

Fifth generation wireless networks (5G) will face key challenges caused by diverse patterns of traffic demands and massive deployment of heterogeneous access points. In order to handle this complexity, machine learning techniques are expected to play a major role. However, due to the large space of parameters related to network optimization, collecting data to train models for all possible network configurations can be prohibitive. In this paper, we analyze the possibility of performing a knowledge transfer, in which a machine learning model trained on a particular network configuration is used to predict a quantity of interest in a new, unknown setting. We focus on the tilt-dependent received signal strength maps as quantities of interest and we analyze two cases where the knowledge acquired for a particular antenna tilt setting is transferred to (i) a different tilt configuration of the same antenna or (ii) a different antenna with the same tilt configuration. Promising results supporting knowledge transfer are obtained through extensive experiments conducted using different machine learning models on a real dataset.

We propose novel resource allocation algorithms that have the objective of finding a good tradeoff between resource reuse and interference avoidance in wireless networks. To this end, we first study properties of functions that relate the resource budget available to network elements to the optimal utility and to the optimal resource efficiency obtained by solving max-min utility optimization problems. From the asymptotic behavior of these functions, we obtain a transition point that indicates whether a network is operating in an efficient noise-limited regime or in an inefficient interference-limited regime for a given resource budget. For networks operating in the inefficient regime, we propose a novel partial resource muting scheme to improve the efficiency of the resource utilization. The framework is very general. It can be applied not only to the downlink of 4G networks, but also to 5G networks equipped with flexible duplex mechanisms. Numerical results show significant performance gains of the proposed scheme compared to the solution to the max-min utility optimization problem with full frequency reuse.

The rapid growth in the number and variety of connected devices requires 5G wireless systems to cope with a very heterogeneous traffic mix. As a consequence, the use of a fixed transmission time interval (TTI) during transmission is not necessarily the most efficacious method when heterogeneous traffic types need to be simultaneously serviced. This work analyzes the benefits of scheduling based on exploiting scalable TTI, where the channel assignment and the TTI duration are adapted to the deadlines and requirements of different services. We formulate an optimization problem by taking individual service requirements into consideration. We then prove that the optimization problem is NP-hard and provide a heuristic algorithm, which provides an effective solution to the problem. Numerical results show that our proposed algorithm is capable of finding near-optimal solutions to meet the latency requirements of mission critical communication services, while providing a good throughput performance for mobile broadband services.

Mobile devices are becoming more chatty in emerging wireless networks due to the machine-type-communication and always-on applications. On the other hand, the densely deployed heterogeneous networks cause more frequent handovers. From the network's perspective, the increase in background traffic and handovers increases the signaling load due to frequent state transitions between idle and connected states. From the user's perspective, staying in connected state over a long period of time results in higher energy consumption. To analyze the tradeoff between the connected and idle states, we develop a closed-form mathematical model of state transition process, based on the framework of alternating renewal process. The mobility-aware user-specific model enables various types of optimization with respective to reduction of signaling overhead and energy consumption. In this paper we apply the developed model to achieve the minimum amount of signaling load by optimizing the inactivity timer, subjected to the energy consumption constraint.

Current cellular networks are mainly designed for the transmission of broadband packet payloads. Due to this, signalling mechanisms are inefficient for small packet payloads. Future trends predict an explosion of sporadic small packet transmissions due to machine type and always-ON type applications. Therefore, we target signalling minimization in 5G. In particular, we define user and service centric connection and mobility management by 5G radio access network (5G-RAN). The key idea is to anchor the core network connection for a user in 5G-RAN with the help of a user centric mobility area. This area is dynamically managed by 5G-RAN for each user. For a mobile user within this area, we minimize the core network signalling related to connection transitions, paging and handover. We present the performance assessment of our proposal with the help of simulations. The results show significant gains in terms of signalling reduction with respect to 4G-LTE approach.

Modern vehicles will have strong requirements with regard to seamless mobility support in future cellular systems, in order to enable advanced cooperative driver assistance and infotainment systems that guarantee traffic safety and efficiency. In this work, we introduce street-specific handover parameters for vehicular terminals. In particular, we propose an adaptive optimization algorithm that exploits vehicle context information in order to tune the handover parameters. Simulation results confirm that the proposed concept has the potential to improve handover performance significantly.

The mobility robustness optimization (MRO) problem in LTE self-organizing networks (SON) is a multi-objective optimization problem; it involves a set of non-convex contradicting objective functions that depend on multiple variables such as handover (HO) parameters and user mobility classes. This paper exploits the framework of stochastic processes to develop a novel method of successively choosing a sequence of multi-variate training points for multi-objective optimization. Combined with the collected statistics and a priori knowledge, the proposed method is used in the design of an efficient MRO algorithm. The performance of the algorithm is evaluated by simulations to illustrate significant improvements with respect to both HO-related ratio link failures (RLFs) and unnecessary HOs.

This paper presents a novel cell outage detection algorithm based on statistics and performance metrics, which enables a base station (BS) to detect a failure/outage of a neighbor cell. The algorithm is a weighted combination of three hypothesis tests based on: 1) the distribution of the channel quality indicator (CQI), 2) the time correlation of the CQI differential, and 3) the registration request (RRQ) frequency. The weights of the combined test are functions of the predicted traffic load in neighboring cells, which is motivated by the fact that the reliability of a individual test depends on the load state. To detect the change-point in the CQI distribution, we use an efficient discriminant function related to the “universal code” proposed by [1], which can be shown to be asymptotically optimal in the sense of the modified Neyman-Pearson criterion. The simulation results indicate that the proposed algorithm can detect the outage problem in a real-time and reliable manner.