State-Of-the-art deep transfer learning methods depend on exhaustive, trial-and-error fine-tuning of pre-trained models—a process that is both computationally expensive and unreliable when data in target domain are scarce. To overcome these limitations, we propose a domain-informed fine-tuning strategy built upon a novel Innately-Intelligent Neural Network (IINN) architecture. Unlike how state-of-the-art deep learning models are heuristically constructed, IINN constructs each layer in a domain informed manner by directly mapping the mathematical operations of analytical equations (e.g., 3GPP propagation models) into it’s network architecture prior to any training. This "innate" design strategy inherently aligns each layer with specific physical parameters, making the model fully interpretable. As a result, we can pre-identify the exact layers associated with parameters that change between source and target domains and fine-tune only those—eliminating the need for iterative layer-by-layer retraining. This targeted fine tuning approach reduces computational overhead and data requirements. We validated IINN on radio-propagation modelling for cellular networks, achieving faster adaptation and higher accuracy than the conventional fine-tuning approach. Experimental evaluations demonstrate that our proposed domain-aware transfer learning framework achieves up to 16.4% improvement in sector-based performance and approximately 10.3% gain in adapting to varying base station heights, with overall average gains in the 10–15% range over state-of-the-art DNN transfer learning approaches. The proposed framework offers a promising direction for data-efficient learning in next-generation wireless systems.

The ubiquitous deployment of 4G/5G technology has made it a critical infrastructure for society that will facilitate the delivery and adoption of emerging applications and use cases (extended reality, automation, robotics, to name but a few). These new applications require high throughput and low latency in both uplink and downlink for optimal performance, while coexisting with traditional downlink-heavy consumer applications. Successfully supporting these new use cases hinges on the network being able to allocate resources as efficiently as possible. In this paper, we utilize a 3GPP-compliant 5G testbed to analyze the limitations of legacy network resource allocation methods, which are based on instantaneous channel measurements, and examine the effect on throughput – a key performance indicator. We then propose a framework that allows resource allocation decisions to leverage predictions of network quality (computed at the connected devices), and study two different prediction methods that provide different degrees of reliability. We further validate our framework with real-world cellular data and demonstrate that with accurate channel metric forecast knowledge, the mean network throughput can improve by a factor of $\sim 1.8$ over the baseline reactive approach based on best CQI policy, for the considered scenario.

The trend towards denser base station deployment and multi-band operations in emerging cellular networks has made mobility management and handover (HO) optimization a formidable challenge. The challenge is further aggravated by the scarcity of practical multi-objective mobility management solutions optimizing both intra and inter frequency HO. This paper presents a holistic multi-objective mobility management solution for both intra and inter frequency HO employing multiple parameters of standardized HO events A2, A3, and A5. We formulate a multi-objective optimization problem to determine the optimal parameter settings that jointly optimize four key performance indicators: number of HO failures, HO latency, signaling overhead and number of radio link failures. We leverage soft actor-critic reinforcement learning (RL) to solve the multi-objective problem. To mitigate the risk of performance deterioration resulting from direct interactions between live network and RL-agent during training, this paper proposes a mobility management framework that develops and employs a digital twin (DT) as the training environment. To develop a cellular network DT for mobility management and HO optimization, we present a tri-pronged approach including realistic network deployment, realistic user mobility and 3GPP HO events. Results show that the proposed DT-trained RL solution for the multi-objective optimization can converge 7x faster than the brute force method with negligible loss in the value of the objective function. An analysis of the individual KPI values reveal a strong trade-off between HO signaling overhead and radio link failures.

The conventional network dimensioning and optimization approaches prioritize coverage and capacity as the most vital components. However, handover signaling overhead has emerged as a critical concern in the emerging cellular networks. This is particularly evident with the proliferation of network densification leading to a higher number of handovers. Hence, an optimal cell overlap is vital to ensure retainability and service continuity for the ever-growing fraction of mobile users and the expected cell densification. It is also crucial because the unprecedented signaling overhead can clog both the core network and air interface. To address this challenge, this paper presents an analytical model built on the control data separation architecture (CDSA) to quantify the handover signaling overhead as a function of cell overlap, user speed and cell density. We first compute probabilities for handover failures and successes and model the handover signaling overhead as a Markov chain. Numerical results demonstrate that for a given cell density and user velocity, a suitable cell overlap yields substantial reductions in handover signaling by improving handover success rate. The proposed model has the potential to become an integral element in the network planning process for emerging cellular networks.

Data-driven Machine Learning (ML) based propagation models are essential for modern wireless network planning and optimization. However, their effectiveness is limited by scarse data conditions. Generative Adversarial Networks (GANs) often considered as a viable approach for data augmentation, struggle in these conditions because they also require large datasets for effective training. To address this challenge, we propose a novel approach that incorporates domain knowledge directly into GAN training. Using an analytical propagation equation based on 3GPP recommendations, we generate pseudo-random data to train a neural network, which then initializes the GAN generator network. This initialization improves the GAN's learning ability in extreme data scarcity. The framework enhances data generation quality by up to 52% and machine learning applicability by 60%, providing a robust solution to the scarse data problem in wireless network modeling with demonstrating the potential of integrating domain knowledge within ML methodologies.

This paper presents and evaluates a simple but effective approach for substantially reducing inter-frequency handover (HO) failure rate. We build a machine learning model to forecast inter-frequency HO failures. For improved accuracy compared to the state-of-the-art models, we use domain knowledge to identify and leverage the model input features. These features include reference signal received power (RSRP) of the source and target base stations as well as the RSRP of the interferers for both the source and the target layers. Six machine learning classifiers are tested with the highest accuracy of 93% observed for the XGBoost classifier. The novel idea to include the RSRP of the interferes improved the accuracy of XGBoost by 10%.

The performance of user-centric ultra-dense networks (UCUDNs) hinges on the Service zone (Szone) radius, which is an elastic parameter that balances the area spectral efficiency (ASE) and energy efficiency (EE) of the network. Accurately determining the Szone radius requires the precise location of the user equipment (UE) and data base stations (DBSs). Even a slight error in reported positions of DBSs or UE will lead to an incorrect determination of Szone radius and UE- D BS pairing, leading to degradation of the UE-DBS communication link. To compensate for the positioning error impact and improve the ASE and EE of the UCUDN, this work proposes a data-driven optimization and error compensation (DD-OEC) framework. The framework comprises an additional machine learning model that assesses the impact of residual errors and regulates the erroneous data-driven optimization to output Szone radius, transmit power, and DBS density values which improve network ASE and EE. The performance of the framework is compared to a baseline scheme, which does not employ the residual, and results demonstrate that the DD-OEC framework outperforms the baseline, achieving up to a 23% improvement in performance.

The fifth generation of cellular network (5G) can facilitate in-ambulance patient monitoring, diagnosis, and treatment by a remote specialist. However, 5G coverage and link quality can vary in time and location. The ambulance route selection can help meet the communication requirements of the in-ambulance applications. In this paper, we propose an innovative ambulance route selection framework which combines the communication requirements along with the network coverage and resources. The framework leverages the minimization of drive test (MDT) data to estimate the network coverage along the ambulance routes. To address the uneven distribution of location-based user-generated MDT data, we examine the performance and trustworthiness of several interpolation techniques to enrich the global MDT map for route selection. A simulated analysis shows that the proposed framework can dynamically adapt to varying application requirements as well as rapidly changing network conditions such as outages. Results also reveal that nearest neighbor and kriging interpolation techniques help complement the proposed framework by addressing the data sparsity problem.

In the wake of network densification and multi-band operation in emerging cellular networks, mobility and handover management is becoming a major bottleneck. The problem is further aggravated by the fact that holistic mobility management solutions for different types of handovers, namely inter-frequency and intra-frequency handovers, remain scarce. This paper presents a first mobility management solution that concurrently optimizes inter-frequency related A5 parameters and intra-frequency related A3 parameters. We analyze and optimize five parameters namely A5-time to trigger (TTT), A5-threshold1, A5-threshold2, A3-TTT, and A3-offset to jointly maximize three critical key performance indicators (KPIs): edge user reference signal received power (RSRP), handover success rate (HOSR) and load between frequency bands. In the absence of tractable analytical models due to system level complexity, we leverage machine learning to quantify the KPIs as a function of the mobility parameters. An XGBoost based model has the best performance for edge RSRP and HOSR while random forest outperforms others for load prediction. An analysis of the mobility parameters provides several insights: 1) there exists a strong coupling between A3 and A5 parameters; 2) an optimal set of parameters exists for each KPI; and 3) the optimal parameters vary for different KPIs. We also perform a SHAP based sensitivity to help resolve the parametric conflict between the KPIs. Finally, we formulate a maximization problem, show it is non-convex, and solve it utilizing simulated annealing (SA). Results indicate that ML-based SA-aided solution is more than 14x faster than the brute force approach with a slight loss in optimality.

Densification and multi-band operation in 5G and beyond pose an unprecedented challenge for mobility management, particularly for inter-frequency handovers. The challenge is aggravated by the fact that the impact of key inter-frequency mobility parameters, namely A5 time to trigger (TTT), A5 threshold1 and A5 threshold2 on the system's performance is not fully understood. These parameters are fixed to a gold standard value or adjusted through hit and trial. This paper presents a first study to analyze and optimize A5 parameters for jointly maximizing two key performance indicators (KPIs): Reference signal received power (RSRP) and handover success rate (HOSR). As analytical modeling cannot capture the system-level complexity, a data driven approach is used. By developing XGBoost based model, that outperforms other models in terms of accuracy, we first analyze the concurrent impact of the three parameters on the two KPIs. The results reveal three key insights: 1) there exist optimal parameter values for each KPI; 2) these optimal values do not necessarily belong to the current gold standard; 3) the optimal parameter values for the two KPIs do not overlap. We then leverage the Sobol variance-based sensitivity analysis to draw some insights which can be used to avoid the parametric conflict while jointly maximizing both KPIs. We formulate the joint RSRP and HOSR optimization problem, show that it is non-convex and solve it using the genetic algorithm (GA). Comparison with the brute force-based results show that the proposed data driven GA-aided solution is 48x faster with negligible loss in optimality.

5G is expected to serve diverse applications and users due to the popularity of Internet of Things (IoT), big data and industrial applications. Many of these IoT and industrial applications have inherent loss tolerance that can be used to enable energy efficient uplink communication. The uplink energy efficient system will increase the battery life of devices enabling new use cases in industrial IoT. In this paper, we map the effects of application loss tolerance to the rate requirements of the user. We then mathematically model an energy minimization problem for the uplink user association and resource allocation in heterogeneous networks. We aim to provide acceptable quality of service (QoS) with improved energy efficiency by exploiting the loss tolerance and bandwidth expansion simultaneously. A distributed uplink joint user association and resource allocation strategy for uplink energy per bit minimization is presented. We conduct extensive simulation based study for a heterogeneous network to evaluate the performance of our proposed schemes. Average energy per bit consumption in the proposed scheme is -74 dB compared to -53 dB in state-of-the-art channel individual offset (CIO) scheme.

The popularity of cellular internet of things (IoT) is increasing day by day and billions of IoT devices will be connected to the internet. Many of these devices have limited battery life with constraints on transmit power. High user power consumption in cellular networks restricts the deployment of many IoT devices in 5G. To enable the inclusion of these devices, 5G should be supplemented with strategies and schemes to reduce user power consumption. Therefore, we present a novel joint uplink user association and resource allocation scheme for minimizing user transmit power while meeting the quality of service. We analyze our scheme for two-tier heterogeneous network (HetNet) and show an average transmit power of -2.8 dBm and 8.2 dBm for our algorithms compared to 20 dBm in state-of-the-art Max reference signal received power (RSRP) and channel individual offset (CIO) based association schemes.

We consider layered decode-forward (DF) cooperation in a relay-aided wireless multicast network. Splitting the source-message into two equal layers, we provide unequal power allocation to the individual layers through a simple mapping operation on a QAM constellation. The layering process thus allows the destinations to partially recover the message from the source's transmissions. At the relay, we propose a multiplexed-coded approach that, with a single transmission, caters for the disparity in the number of layers decoded at different destinations. In addition to simulations, we validate the performance gains of the proposed strategy through a system-level implementation using software-defined radios. Catering for real- world factors such as carrier and timing synchronization, we conduct over-the-air experiments in an indoor office environment and find that the proposed scheme can achieve a frame- error-rate that is 25% of that with conventional two-hop DF relaying.