To address the shortcomings of the DV-Hop algorithm in optimizing node hop count and average hop distance, we propose a multi-mechanism integrated cheetah optimizer to optimize the three-dimensional DV-Hop localization algorithm, namely the IMCO-DV-Hop. By enhancing the positioning accuracy of range-free algorithms in wireless sensor networks, the precision and timeliness of landslide prediction can be improved. Firstly, by analyzing the causes of hop count error and average distance error in the traditional three-dimensional DV-Hop algorithm, the hop count of nodes is corrected using multi-communication radius. Then, collaboration is used to remove anchor nodes that cause large errors, and the average hop distance is optimized through weighted processing, reducing the positioning error caused by the traditional three-dimensional DV-Hop algorithm. Secondly, the Tent chaotic mapping is used to initialize the population, and the Levy flight, dimensionality mechanism, and three-dimensional random tensor are applied to learning and exploration to enhance the global search ability of the improved Cheetah optimizer, thereby improving the positioning accuracy and efficiency of the algorithm. Furthermore, Friedman test and Nemenyi post-hoc analysis are conducted to validate the statistical significance of the performance improvement, confirming that the proposed algorithm significantly outperforms the original DV-Hop and other comparative algorithms.
By integrating the digital and physical worlds, billions of heterogeneous devices are interconnected within the expansive domain of the Internet of Things (IoT). Considering the growing number of connected devices and the lack of sufficient security mechanisms, IoT is susceptible to numerous internal security threats, particularly those aimed at the conventional routing protocol designed for low-power and lossy networks (RPL). Machine learning-based detection methods provide the ability to identify malicious activities in RPL-based IoT environments, before causing significant impacts on the networks’ overall performance. These detection capabilities are achieved through data-driven learning of network behavior patterns, enabling the model to automatically identify malicious activities without relying on predefined attack signatures. Due to the IoT challenges regarding resource limitations, large volume of exchanged data, heterogeneity and scalability issues, along with the privacy risks associated with centralized data collection, recent research initiatives in RPL security have incorporated collaborative learning methodologies. In the current research, a cooperative machine learning-based mitigation approach has been designed by combining federated learning with the inherent ensemble learning capability of XGBoost. Focusing on the timely detection of multiple RPL-specific routing attacks, such as Blackhole attack, Decreased Rank attack, Version Number attack and Flooding attack, we have proposed a federated security scheme based on the Extreme Gradient Boosting (XGBoost) algorithm. By combining the robustness and efficiency of XGBoost with the benefits of federated learning, this research aims to achieve competitive detection performance while improving learning efficiency through collaborative and distributed model training. The proposed model’s performance was assessed across different RPL attack scenarios, and the experimental results demonstrated competitive detection performance while providing the advantages of decentralized learning and privacy-aware collaborative model training compared with conventional centralized machine learning classifiers.
Bluetooth Low Energy (BLE) has become the radio technology of choice for wearables, thanks to its low duty-cycled power consumption, small stack, and ecosystem surrounding. With the transition from simple notification relays to latency-bound services like electrocardiography, photoplethysmography, and electroencephalography streaming, battery life and quality of service (QoS) are the leading design constraints for wearables. Duty cycling and connection interval, slave latency, advertising interval, transmit power, and PHY mode tuning are conventional optimization methods, and they can be used to get predictable savings, but are fundamentally static do not react to non-stationary user activity, interference or bursty traffic. In order to overcome this rigidity, Machine Learning, Deep Learning, and Reinforcement Learning are proposed, all of which attempt to learn context-dependent control policies dynamically (online) to reduce energy consumption, including a Q-learning scheduler, which has been reported to reduce energy consumption by over 30
In Fifth Generation (5G) millimeter Wave (mmWave) and Sixth Generation (6G) Sub Terahertz (Sub-THz) communication systems, there are very vital challenges. Some of them are energy efficiency-latency tradeoff and heat. These challenges fluctuate the Quality of Service (QoS), especially in modern wireless networks with high traffic density. To maintain QoS, the Discontinuous sleep mode is presented which saves much energy, provides a reasonable reduced latency and absorbs heat. This sleep mode can turn off the Radio Frequency (RF) transceiver circuit if data traffic arrival is not existed. Discontinuous Transmission (DTX) and Discontinuous Reception (DRX) sleep modes are adopted for Base Station (BS) and User Equipment (UE), respectively. Due to multiple beam links performed between the BS and the UE, Beam Aware (BA) algorithm is introduced to enable the both to be only aware with the active serving beams. In this paper, BA is carried out through a Serving Beam Cycle (SBC) instead of fixed beam searching durations. Moreover, the Beam Aware Signaling DRX (BASg-DRX) and Beam Aware Standard DTX (BASt-DTX) are proposed to enhance the performance rather than the ordinary system models. Mobility and heat of the UE are taken into account. A MATLAB simulation program was used to check the validity of the proposed system models. The BASg-DRX proposed model outperforms the ordinary model by about (3–35)
To address the challenges of traffic congestion and route inefficiency in Intelligent Transportation System (ITS) environments, an effective Federated Deep Reinforcement Learning (FDRL)-based framework is proposed for lane reassignment and distributed control. Within this framework, each Roadside Unit (RSU) independently learns a local deep reinforcement learning (DRL) policy based on vehicle state information and periodically synchronizes with a global model. This federated approach mitigates local optimization bias and enhances cooperative control performance across the network. A dual-output policy network is designed to jointly optimize lane-level vehicle distribution and speed control. Moreover, a dynamic federated learning strategy adaptively adjusts aggregation weights based on inter-RSU state discrepancies, ensuring robust performance under non-IID conditions such as heterogeneous lane and data imbalance. Experimental analyses demonstrate that the proposed Federated Loss-based Global Allocation (FedLoss-GA) model outperforms independently trained DRL models and simple averaging-based federated models in terms of policy stability and traffic allocation efficiency. Importantly, FedLoss-GA maintains balanced control across RSUs with heterogeneous incoming lanes by aligning local policies to global coordination objectives. The proposed model enables scalable and cooperative traffic control in ITS environments, providing a solid foundation for addressing predictive uncertainties in diverse real-world traffic scenarios.
The emergence of unmanned aerial vehicle (UAV) enabled with deep reinforcement learning (DRL) has the applications in the field of multi-purpose urban development. We introduce a trajectory optimization framework for safety and control by utilizing Lyapunov-based optimization with DRL approach. The operating duration of UAVs and the lifespan of the network, we develop an optimization problem aimed at reducing the overall energy usage of the UAV via coordinated region division across the network. The problem is divided into two sub-problems, First, A Lyapunov function based control program is applied to approximate the control of non-linear multiple UAV system, subsequently, UAV trajectory optimization is formulated as a DRL problem to attain the optimal outcome and to ease the delay with appropriate load balance. The overall energy usage is minimized via coordinated path planning using DRL-LO algorithm. The proposed approach considerably decreases the delay through safety control and optimized trajectory path. The results are effective in comparative with the prevailed approaches. The performance metrics are the data rate and the energy usage of multiple UAVs.
The enormous proliferation of Internet of Things (IoT), cloud computing, and fog computing have opened a mightier opportunities reliable communication and technology. However, the centralization of data issues enables amplified memory usage and security threats. The cloud-fog-IoT resolve the mentioned issues by optimizing the load distribution, delay, and latency respectively. However, the cyber securities in cloud-fog-IoT remains unsolved due to sophisticated attacks targeting the vulnerable areas to took down the complete environment. To this end, we design a Dynamic Distributed Reconfigurable Cognitive-Intrusion Detection System (D2RC-IDS) for cloud-fog-IoT environment using Deep Learning (DL) and Deep Reinforcement Learning (DRL) methodologies. For the designed environment, we have collected the traffic data from three datasets named BoT-IoT, NSL-KDD, and KDDCup-99 datasets respectively. At first, the collected data from the three datasets are normalized using min-max normalization method. Followed by the normalized data are then provided to distributed fog servers for performing signature entrenched IDS using hybrid DL algorithm named Gated Recurrent Unit-Long Short-Term Memory (GRU-LSTM) which classifies traffic into normal, malicious, and suspicious data respectively. The suspicious data are then provided as an input to the Dynamic Reinforced Cognitive-Anomaly IDS (DRC-AIDS) module inspired from the nature of dynamic reinforcement learning algorithm which composed of feature extraction module and classification module named Slide Transformer Network (SlideFormerNet) and Lite-Convolutional Neural Network (Lite-CNN) which classifies the suspicious traffic into malicious and normal. Implementation of projected model is carried out using python 3.11.0 tool and evaluation of propounded model gets validated by metrics such as accuracy, precision, recall, F1-score, Area Under Curve (AUC) and Precision Recall Curve (PRC). The investigational consequences shows that propounded D2RC-IDS model outclasses than state of the art works.
Orthogonal Time Frequency Space (OTFS) modulation is a strong candidate for beyond 5 G and 6 G systems, thanks to its robustness in high mobility, highly dispersive channels. However, the Peak to Average Power Ratio (PAPR) remains a critical bottleneck, as large peaks drive power amplifier nonlinearity, spectral regrowth, and Bit Error Rate (BER) degradation, which is particularly acute in scenarios such as high speed rail. To address this, we integrate the Weighted Fractional Fourier Transform (WFrFT) into the OTFS modulation and optimize its fractional order α via Particle Swarm Optimization (PSO). By adaptively shaping the Time Frequency (TF) energy distribution, the PSO selected WFrFT order lowers Time Domain (TD) peaks and improves amplifier operating efficiency. We further combine this design with classical PAPR reduction schemes: Selected Mapping (SLM) and Partial Transmit Sequences (PTS), and we show that PSO driven phase selection outperforms conventional phase generation, yielding more effective hybrids (PSO–WFrFT+SLM and PSO–WFrFT+PTS). Simulation results indicate consistent PAPR reduction and improved transmission reliability over conventional OTFS baselines, without sacrificing spectral efficiency.
The increasing deployment of the Industrial Internet of Things (IIoT) requires authentication mechanisms with low overhead and predictable latency. Although IO-link wireless (IOLW) supports deterministic industrial communication, it does not define a complete network-layer authentication and key agreement procedure. This paper presents IOLW-CAP, a four-message protocol for the IOLW Master–Slave model. It combines physically unclonable function (PUF)-assisted root-key reconstruction, Chebyshev-based session-dependent nonlinear mixing, and a two-stage key-based key derivation function (KBKDF) instantiated with the cipher-based message authentication code (CMAC), thereby binding the selected enrollment record and current transcript to purpose-specific keys. Security is evaluated through informal analysis, a real-or-random (RoR) proof, and finite coloured Petri Net (CPN) state-space models. The RoR analysis establishes session-key indistinguishability for fresh sessions under the stated assumptions, while none of the modeled tamper, replay, or previous-session-material exposure paths reaches authentication success or session-key establishment in the CPN models. A Bose–Chaudhuri–Hocquenghem (BCH)-based fuzzy extractor achieves reconstruction success rates of 99.99
Mobile Ad-hoc Networks (MANETs) require adaptive routing and scheduling mechanisms to maintain reliable communication, particularly in infrastructure-limited environments such as search-and-rescue operations. The proactive Optimized Link State Routing version 2 (OLSRv2) protocol provides a partial network view through link state dissemination, and the OLSR+SINRT extension improves upon this by filtering out links with poor Signal-to-Interference-and-Noise Ratio (SINR) to compute more stable routes. Time Division Multiple Access (TDMA) eliminates collisions through synchronized time slots, yet its conventional Round-Robin (RR) scheduling allocates resources uniformly and performs poorly at high-burden routing nodes in dynamic topologies. This paper introduces a cross-layer scheduling framework that leverages a Node View Graph (NVG), a localized topological construct derived from OLSR+SINRT, to align TDMA slot allocation with routing demands. Four distributed scheduling schemes are proposed and evaluated: a traffic-oblivious baseline, a traffic-aware approach (TrafficMPR+TLV), an on-demand reservation protocol (SlotRequest), and a reinforcement learning-based scheduler (2hopsMLP). Emulations are conducted using the Extendable MANET Emulator (EMANE) on the realistic AnglovaNet scenario with realistic radio conditions including multipath fading. Results show that traffic-aware scheduling consistently outperforms RR across multiple traffic patterns. SlotRequest achieves the highest completion ratio under sink-centric traffic (63
To address the degradation of service utility caused by location perturbation, the limited real-time performance of online scheduling, and the weak coordination between privacy-preserving allocation and incentive design in mobile crowd sensing (MCS), this paper proposes a sensing-platform-side privacy-preserving task-allocation method based on service-level-driven scheduling and computation over ciphertexts. The proposed method focuses on sensing-platform-side location confidentiality under the semi-honest KDC assumption. Specifically, an adaptive grid-hash indexing method driven by service-level agreements (SLAs) is designed to reduce the candidate search space before controlled encrypted-offset distance scoring while preserving task-relevant spatial semantics. A controlled encrypted-offset distance scoring protocol based on approximate homomorphic encryption and distance grading by the semi-honest KDC is then developed, enabling the sensing platform to obtain only discrete distance scores instead of workers’ plaintext coordinates or exact distances. In the incentive layer, a welfare-driven bipartite matching rule is constructed, and a budget-capped contribution-aware incentive mechanism based on Monte Carlo Shapley estimation is designed to support contribution-aware surplus sharing under budget constraints. Theoretical analysis is provided for budget feasibility and Monte Carlo approximation error. Experiments on real-world urban mobility trajectory data show that the proposed method achieves a practical trade-off among sensing-platform-side leakage reduction, allocation utility, runtime overhead, and budget-aware incentive control under the evaluated settings.
With the rapid growth of massive machine-type communications in non-terrestrial networks, grant-free access has emerged as a key enabler for LEO satellite IoT connectivity under sporadic traffic. This paper concerns uplink grant-free low earth orbit (LEO) satellite communication systems, where ground users can randomly transmit data at any time slot. Consequently, the active users and their channel state information have to be identified by the receiver before data detection. Motivated by the great potential of orthogonal delay-Doppler division multiplexing (ODDM) modulation for high-mobility scenarios, we investigate the problem of joint active user detection and channel estimation for uplink grant-free LEO satellite communication systems with ODDM modulation. In this paper, the joint user activity detection and channel estimation is formulated as a sparse signal recovery problem. Without the prior knowledge of sparsity, we develop a structured compressive sensing (CS) algorithm, specifically a variable step-size sparsity adaptive simultaneous orthogonal matching pursuit (VSS-SASOMP) algorithm to solve the problem. Simulation results demonstrate the effectiveness of the proposed method, showing that VSS-SASOMP consistently outperforms SOMPT and other baselines in both estimation accuracy and activity detection. For example, at SNR = 10 dB, VSS-SASOMP achieves an F1-score of 0.981, providing an improvement of 0.141 over SOMPT. Moreover, the proposed variable step-size strategy reduces the average running time by 72.1
This paper presents a novel neural network model, the Alternating Graph-Regularized Neural Network for Adaptive Beamforming in 5G Millimeter-Wave Massive MIMO Multicellular Networks (AGNN-ABM-MIMO-MCN), designed to enhance signal transmission and spectral utilization in next-generation wireless systems. Beamforming plays a crucial role in optimizing signal directionality and strength across multiple base stations (BSs) in massive MIMO networks. The proposed AGNN-based framework adaptively modifies beam patterns in response to real-time traffic dynamics, ensuring optimal resource allocation and network coverage. By incorporating graph regularization into deep neural architectures, the AGNN effectively captures spatial dependencies among nodes, minimizing interference and improving signal quality. The inclusion of the T-Rex Adaptive Algorithm accelerates convergence and strengthens optimization, yielding notable improvements in Spectral Efficiency (SE), Energy Efficiency (EE), and Bit Error Rate (BER). Simulation results show that the proposed model outperforms existing schemes such as MLAB-MMWM-MIMO, DL-ERN-MCN, and DLF-BSPC-MIMO. This research delivers an intelligent, adaptive, and energy-efficient beamforming solution capable of supporting heterogeneous hardware configurations, paving the way for scalable 5G and beyond-5G wireless communication systems.
Intelligent reflective surfaces (IRSs) have attracted significant attention due to their exceptional capability to enhance the capacity and coverage of wireless networks by intelligently reconfiguring the wireless propagation environment in a controlled manner. Due to their immense potential to revolutionize wireless paradigms, IRSs are envisioned as an emerging technology for sixth-generation (6G) communication networks. To analyze a practical scenario, we study the performance of finite alphabet signals in the context of IRS-assisted wireless communications over Rayleigh fading channels, which is in contrast to the majority of works in the literature that consider Gaussian input. Motivated by the high data rates and enhanced power efficiency of Quadrature Amplitude Modulation (QAM) and its variants, we have employed it as channel input. We analyze the impact of modulation parameters on the achievable rate by implementing a Monte Carlo simulation. We infer from the results that even a small number of IRS elements can provide a substantial rate gain, especially when compared to an additive white Gaussian noise channel, particularly when the signal-to-noise ratio (SNR) is low. Furthermore, we also consider the gaussian multiple access channel (G-MAC) in an IRS-assisted wireless channel for M-QAM inputs and evaluate its achievable rate, which clearly show improved sum rates. Security, being one of the key concerns of 6G wireless networks, has been considered in the proposed work, and we compute the secrecy sum rate for IRS-assisted GMAC with an eavesdropper for the different variants of QAM, which has not been reported in any previous work. Since it is well known that there is no closed-form expression for mutual information with finite constellation, we try to find a bound on the achievable rates. To gain more practical insight, we derive a lower bound of sum rate for IRS-assisted two-user GMAC for M-QAM inputs, and compare it with the simulated results. Next, we have also evaluated the energy efficiency of M-QAM variants in an IRS-assisted environment with and without security and also evaluated the inner bound to determine the performance benchmarks. The numerical results reveal that the bound though not so tight but helps in establishing performance benchmarks. In addition to security, energy efficient communication is also a key concern for next generation communication networks. We compute energy efficiency for various practical scenarios for IRS aided Multiple Access Channel, with and without security constraints. We observe that the SNR requirements for energy efficient communication systems are different from rate-efficient systems. In particular there is tradeoff between rate and energy efficiency, which can guide the system designers while choosing optimum modulation scheme for communication network.
The high reliability and low latency requirements for 5G and beyond wireless networks pose a challenge on the transport infrastructure, especially for ultra-reliable low-latency communication (URLLC) services. Optical transport networks play a crucial role in meeting the requirements of 5G and beyond wireless networks by providing high-capacity, low-latency connectivity between distributed network entities. In this paper, a reliable optical transport framework has been proposed for improving service continuity and quick recovery for 5G and beyond wireless networks for URLLC communications. The uniqueness of the suggested model lies in the implementation of hybrid protection-restoration and its reliability-oriented redundancy allocation. It also integrates CNN-assisted failure detection into a single recovery model. The suggested approach incorporates protection through the creation of k-disjoint paths, dynamic reconfiguration of the path with the use of SDN, and machine learning-enabled failure prediction. A performance metric has been formulated for evaluating network performance based on reliability. The proposed hybrid approach works under various network impairment conditions and is validated via simulation using the Mininet framework on benchmark NSFNET and GEANT2 networks. The proposed coordinated mechanism ensures excellent recovery time of 24 s and 27 s, 98
In this paper, we present a broadband differential-fed dual-polarized dipole antenna with ultra-wideband harmonic suppression for base-station applications. Two orthogonal folded dipoles with slender stubs are incorporated with a double differential feed scheme for the dual-polarized radiation and harmonic suppression. Four T-shaped parasitic elements are added to the dipoles for the broadband operation. To improve the bandwidth of the harmonic suppression, U-shaped slots, open stubs, and shorted strips are applied to the dipoles. The antenna is realized on a single-layer substrate and directly fed by four 50- Ω coaxial cables. The fabricated prototype has an impedance matching bandwidth of 2.49–3.80 GHz, an isolation of ≥ 44 dB, and an ultra-wideband harmonic suppression of 4.0–12.6 GHz for reflection coefficients ≥ − 2.5 dB. Moreover, the broadside dual-polarized radiation is validated by far-field measurements, showing highly symmetrical pattern and low cross-polarization within the operational bandwidth.
