Despite advancements in language-controlled reinforcement learning (LC-RL) for basic domains and straightforward commands (e.g., object manipulation and navigation), effectively extending LC-RL to comprehend and execute high-level or abstract instructions in complex, multi-agent environments, such as football games, remains a significant challenge. To address this gap, we introduce Language-Controlled Diverse Style Policies (LCDSP), a novel LC-RL paradigm specifically designed for complex scenarios. LCDSP comprises two key components: a Diverse Style Training (DST) method and a Style Interpreter (SI). The DST method efficiently trains a single policy capable of exhibiting a wide range of diverse behaviors by modulating agent actions through style parameters (SP). The SI is designed to accurately and rapidly translate high-level language instructions into these corresponding SP. Through extensive experiments in a complex 5v5 football environment, we demonstrate that LCDSP effectively comprehends abstract tactical instructions and accurately executes the desired diverse behavioral styles, showcasing its potential for complex, real-world applications.

In the era of digital transformation, the large-scale deployment of sensors has led to the collection of highly complex and diverse data, posing significant challenges for multivariate time series forecasting (MTSF). Traditional forecasting approaches, often based on linear assumptions, are limited in their ability to capture the nonlinear temporal dynamics prevalent in real-world scenarios. To address these challenges, this study proposes an innovative multivariate prediction framework that integrates deep learning with traditional machine learning techniques. The framework incorporates an asymmetric transfer entropy coefficient (ATC) to identify genuine causal relationships among features, constructing a directed graph for feature importance ranking. This mechanism enhances feature selection by capturing both dynamic and static relationships among variables. An enhanced fuzzy C-means clustering algorithm, SCFCM, is introduced, which incorporates cosine similarity and Euclidean distance to improve sample discriminability in high-dimensional spaces and enhance clustering accuracy. Bayesian optimization is employed to dynamically determine the kernel sizes and numbers of the CNN-Transformer (Convolutional neural network-Transformer) prediction network, thereby improving feature extraction efficiency. The unified architecture integrates feature selection, clustering, and forecasting, achieving superior predictive performance. This comprehensive prediction model is referred to as ATC-SCFCM-DKCNT. Experiments on six real-world datasets demonstrate that ATC-SCFCM-DKCNT consistently outperforms state-of-the-art methods in terms of prediction accuracy and computational efficiency, highlighting its strong generalization ability and robustness in handling complex, high-dimensional data.

In this paper, we propose an optimisation framework for deployment of multiple reconfigurable intelligent surfaces (RISs) to meet the wireless coverage demands for smart warehouses. Specifically, we are the first to formulate a unified network optimisation task that jointly considers RIS placement and beamforming to maximize overall network coverage with a deterministic channel model to accurately describe the multipath effect for the warehouse. To address this problem, we design a hybrid optimisation framework composed of three synergistic modules. (1) A Large Language Model (LLM) acts as a semantic planner that generates physically feasible multi-RIS configurations, jointly determining the placement and beamforming directions guided by structured prompts and environment-aware embeddings. (2) A Genetic Algorithm (GA) module performs local numerical refinements to enhance the precision of LLMgenerated solutions under physical constraints. (3) A Diversity Reflection and Correction (DiRect) module evaluates structural similarity among candidate configurations and triggers additional semantic regeneration to maintain exploration diversity. These three modules form an alternating iterative process in which LLM reasoning, GA-based evolution, and DiRect-driven regeneration collectively guide the optimisation toward high-coverage configurations. Extensive simulations validate the effectiveness and robustness of the proposed framework. Compared with traditional heuristics, reinforcement learning methods, and LLMguided baselines, our hybrid framework achieves 10%-15% higher coverage within 10-20 iterations. The performance consistently scales with the number of RISs and element sizes, and remains stable under varying transmitter positions, demonstrating strong adaptability to complex smart warehouse layouts. Overall, the proposed hybrid optimisation framework provides a scalable and physically grounded solution for RIS-assisted network deployment optimisation in realistic in.

For the Internet of things (IoT) network, the integrated localization and communication (ILAC) is expected to provide high localization and communication performance simultaneously. However, the existing research to evaluate the performance of ILAC systems fails to reveal the fundamental performance of ILAC systems in practical IoT network topology analytically. In this paper, we develop a unified analytical ILAC framework using stochastic geometry. We then validate the theoretical results obtained from the proposed analytical framework with the simulation results via extensive Monte Carlo simulations. We further analyse the communication coverage and localization coverage probability with respect to the network density, time-frequency-power domain resource allocation, and communication throughout and localization threshold. Finally, based on the ILAC simulation results, we reveal design guidance for ILAC systems. Specifically, we observe the fundamental trade-off between localization and communication performance attributed to the time-frequency-power domain resource allocation. Network density positively affects the ILAC performance, while power control is much less effective due to the dense network topology. The major observations are that time-domain (TD) resource allocation is preferred in dense networks with low localization CRB thresholds, while frequency-domain (FD) resource allocation dominates in sparse networks with large localization CRB thresholds.

In the coexisting new radio-based access to unlicensed spectrum (NR-U) and WiGig networks (CNWNs), directional-sensing-based listen-before-talk (LBT) mechanisms, i.e., directional LBT (dirLBT) and paired LBT (pairLBT), have been proposed to address the exposed node problem caused by traditional omnidirectional LBT (omniLBT) mechanism. In this paper, we are the first to leverage the stochastic geometry to analyze the large-scale CNWN performance when NR-U base stations (NBSs) adopt the directional-sensing-based LBT mechanisms. The analytical expressions for the downlink successful transmission probabilities (STPs) of CNWNs are derived and validated by Monte Carlo simulations. Based on these STPs, the area spectral efficiency (ASE) of CNWNs is derived. Equipped with these results, the effect of NBS sensing threshold, density and sensing beamwidth on the STP and ASE performance are analyzed numerically. Moreover, the STP and ASE performance are compared when NBSs adopt dirLBT, pairLBT and omniLBT mechanisms. Furthermore, the asymptotic ASE of CNWNs when NBS density approaches infinity is derived and validated. The results show that directional-sensing-based LBT mechanisms outperform the omniLBT mechanism in terms of ASE in the CNWNs, especially in ultra-densely deployed scenarios. Under our simulation environment, the dirLBT mechanism can improve the ASE by up to 82.5% as compared with the omniLBT mechanism. Additionally, the NBS sensing threshold for directional-sensing-based LBT should be higher than −73 dBm to achieve a better STP and ASE as compared with that without adopting LBT in NBSs. Besides, there exists an optimal NBS density to maximize the STP and ASE of CNWNs, and when NBS density becomes larger than$200,000$NBSs per$\text {km}^{2}$, deploying more NBS has limited enhancement on the ASE. These results indicate that directional-sensing-based LBT mechanisms should be employed in the ultra-densely deployed CNWNs, and the NBS sensing threshold and sensing beamwidth should be carefully chosen to ensure the superiority of directional-sensing-based LBT mechanisms.

The 5G New Radio-based in unlicensed spectrum (NR-U) has been proposed to harmoniously coexist with the Wireless Gigabyte (WiGig) network in the 60 GHz unlicensed spectrum. It employs the directional listen-before-talk (dirLBT) mechanism to improve the throughput of the coexisting NR-U and WiGig networks (CNWNs). In this paper, we are the first to leverage the stochastic geometry to analyze the performance of the large-scale CNWNs. The medium access probabilities (MAPs) of NR-U base station (NBS) and WiGig access point (WAP) are both derived in closed-form. Based on these MAPs, the downlink successful transmission probabilities (STPs) of NR-U and WiGig networks, which is determined by the retaining probability of the serving NBS/WAP and the downlink coverage probability of NR-U/WiGig network, are given in analytical expressions. All these MAPs and STPs are validated by Monte Carlo simulations to verify the correctness of our proposed model. Moreover, the effect of NBS and WAP density on the mean STP of the large-scale CNWNs are analyzed numerically. The results show that the dirLBT adopted by NR-U outperforms omnidirectional LBT in terms of the mean STP, especially in ultra-densely deployed CNWNs scenario. Furthermore, there exists an optimal NBS density to maximize the mean STP. The results indicate that the dirLBT mechanism should be adopted in the densely deployed CNWNs with proper chosen of NBS density.

The Age of Information (AoI), which evaluates the information freshness, and the energy efficiency (EE) play key roles in cellular IoT networks. This is due to that outdated data can hardly provide any useful information for delay-sensitive applications and the IoT devices usually have limited battery life. In particular, the AoI can be significantly affected by the transmission latency, which becomes the bottleneck for the AoI performance in ultradensely deployed cellular IoT networks. Moreover, it is desirable for cellular IoT networks to achieve low AoI with high EE. The data compression (DC) can decrease the AoI and improve the EE by reducing the transmission latency. Therefore, in this work, the AoI and EE performance in a large-scale densely deployed uplink cellular IoT network are jointly analyzed with the DC technology. Specifically, the closed-form results of AoI are derived and validated by Monte Carlo simulations. Based on these results, the AoI–EE ratio is defined to evaluate the tradeoff between the AoI and the EE. Equipped with these results, the effects of compression ratio (CR) and status update packet generation rate (SUPGR) on both the AoI and the AoI–EE ratio are analyzed numerically. The results show that by jointly optimizing the CR and SUPGR, the AoI can be decreased by up to 82% and the AoI–EE ratio can be reduced by up to 83% as compared with the case that only adjusts the SUPGR without the DC. It indicates that the DC should be widely adopted in IoT devices, which can improve the information freshness with low-energy consumption, especially in an ultradensely deployed scenario.

The fog-radio-access-network (F-RAN) has been proposed to address the strict latency requirements, which offloads computation tasks generated in the user equipment (UE) to the edge to reduce the processing latency. However, it incorporates the task transmission latency, which may become the bottleneck of latency requirements. Data compression (DC) has been considered as one of the promising techniques to reduce the transmission latency. By compressing the computation tasks before transmitting, the transmission delay is reduced due to the shrink transmitted data size, and the original computing task can be retrieved by employing data decompressing (DD) at the edge nodes or the centre cloud. Nevertheless, the DC and DD incorporate extra processing latency. For the F-RAN system, the latency performance has not been investigated considering the DC and DD processes. Therefore, in this work, the successful data compression probability (SDCP), i.e., the probability of the task execution latency being smaller than a target latency and the signal to interference ratio (SIR) of the received signal being higher than a threshold, is defined to analyse the latency performance of the DC-enabled F-RAN. Moreover, to analyse the impact of compression offloading ratio (COR), which determines the proportion of tasks being compressed at the edge, on the SDCP of the F-RAN, a novel hybrid compression mode is proposed based on the queueing theory. Based on this, the closed-form result of SDCP in the large-scale DC-enabled F-RAN is derived by combining the Matern cluster process and M/G/1 queueing model, and validated by the Monte-Carlo simulation. Based on the derived SDCP results, the effects of COR on the SDCP is analysed numerically. The results show that the SDCP with the optimal COR can be enhanced with a maximum value of 0.3 and 0.55 as compared with the cases of compressing all computing tasks at the edge and at the UE, respectively. Moreover, for the system requiring the minimal latency, the proposed hybrid compression mode can alleviate the requirement on the backhaul capacity.

Reconfigurable intelligent surface (RIS), which provides indirect line-of-sight (LoS) transmission paths between the receiver and its blocked transmitter, has been proposed as one of the promising technologies for network performance enhancement. Recent works indicate that both the numbers of RISs and unit cells per RIS have a significant impact on network performance. However, with a given total number of unit cells, the joint analysis of these two factors considering blockage effect has not been investigated. In this paper, the coverage probability of a three-dimensional downlink millimeter wave RIS-assisted network is analyzed. We first derive the acceptable area where a LoS RIS is capable to satisfy the signal-to-noise ratio threshold at the receiver. Next, we model the centers of building blockages and human-body blockages as two independent Poisson point processes and derive the probability that indirect LoS transmissions exist. Then, we derive and validate the analytical upper and lower bounds of the coverage probability as the functions of network parameters and blockage densities. We also derive and validate the closed-form coverage probability when RISs are much closer to UE than BS. Finally, we propose a general network cost model for RIS-assisted network. Results show that in terms of coverage enhancement, densely deployed small-scale RISs outperform sparsely deployed large-scale RISs in scenarios with dense blockages or short transmission distances, while sparsely deployed large-scale RISs are preferable in scenarios with sparse blockages or long transmission distances.

Computing trajectory similarities is a critical and fundamental task for various spatial-temporal applications, such as clustering, prediction, and anomaly detection. Traditional similarity metrics, i.e. DTW and Hausdorff, suffer from quadratic computation complexity, leading to their inability on large-scale data. To solve this problem, many trajectory representation learning techniques are proposed to approximate the metric space while reducing the complexity of similarity computation. Nevertheless, these works are designed based on RNN backend, resulting in a serious performance decline on long trajectories. In this paper, we propose a novel graph-based method, namely TrajGAT, to explicitly model the hierarchical spatial structure and improve the performance of long trajectory similarity computation. TrajGAT consists of two main modules, i.e. , graph construction and trajectory encoding. For graph construction, TrajGAT first employs PR quadtree to build the hierarchical structure of the whole spatial area, and then constructs a graph for each trajectory based on the original records and the leaf nodes of the quadtree. For trajectory encoding, we replace the self-attention in Transformer with graph attention and design an encoder to represent the generated graph trajectory. With these two modules, TrajGAT can capture the long-term dependencies of trajectories while reducing the GPU memory usage of Transformer. Our experiments on two real-life datasets show that TrajGAT not only improves the performance on long trajectories but also outperforms the state-of-the-art methods on mixture trajectories significantly.

Three-dimensional (3-D) massive multiple-input-and-multiple-output (MIMO) systems, which explore degrees of freedom in both the vertical and the horizontal dimensions, are a promising technology to enhance spectral efficiency in the next-generation communication systems. As an emerging modulation technology with millimetre wave (mm-wave) communication in massive MIMO systems with limited radio-frequency (RF) chains, the spatial scattering modulation (SSM) makes use of beamspace domain resources to further improve the spectral efficiency. However, most published works on the SSM only focus on two-dimensional (2-D) MIMO systems. In this paper, we generalize the SSM system to 3-D space. First, we design a novel generalized 3-D SSM system by considering both vertical and horizontal angles to determine scattering paths, which are selected to convey information bits. Then, we propose a whitening filter based optimal detection algorithm to detect the received symbols with correlated noise, where the correlation is generated by the combination of the received signals from the large-scale planar receiving antenna array. Next, we design a 2-D fast Fourier transform (FFT) based transceiving approach to improve the hardware friendliness. After that, we propose a low-complexity detector based on the linear minimum mean square error (MMSE) detection algorithm. Moreover, we derive the union upper bound on average bit error probability (ABEP) in a closed form and the asymptotic performance expression for the generalized 3-D SSM system. Finally, we analyse the impact of transmission environment on the ABEP performance. Numerical results show that the generalized 3-D SSM system outperforms the conventional 2-D SSM system, which reduces the ABEP by$10\times $with the same signal-to-noise ratio (SNR) level under the typical indoor environment.

With the rapid development of mobile devices, limited licensed spectrum resources can no longer meet the explosive growth of data traffic demand. Therefore, the industry and academia have set their sights on utilizing the unlicensed spectrum in the cellular network which is mainly used by the Wi-Fi network. As a result, it is crucial for the cellular network to harmoniously coexist with the Wi-Fi networks. Recently, the new radio in unlicensed spectrum (NR-U) network has been proposed to address the coexistence between the cellular and Wi-Fi networks from the perspective of beam or space. However, the spatial resource in cellular base stations (BSs), especially in the small-cell BSs, may be insufficient. In this work, we propose a novel joint spatial-temporal domain based scheme to address this. The joint resource allocation problem is formulated to maximize the total throughput of the coexisting NR-U and Wi-Fi networks. By using the mixed integer quadratic programming (MIQP), a joint spatial-temporal resource allocation scheme is obtained. Simulation results show that the joint spatial-temporal domain coexistence scheme can achieve a maximum of 30% and 58.3% gain in terms of the total throughput performance as compared with the sole carrier sense adaptive transmission (CSAT) and interference nulling scheme, respectively, under insufficient spatial resources.

Uplink transmission delay has been considered as a main component of the end-to-end delay in the fog radio access networks (F-RAN). However, existing analysis of the transmission delay ignore the packet retransmission delay, i.e., mean local delay (MLD), which is the main component of transmission delay. In addition, the MLD has not been investigated in a clustered F-RAN. Therefore, in this work, we leverage the Matern cluster process (MCP) to analyse MLD in a large-scale F-RAN. To derive the MLD, we obtain the uplink coverage probability (CP) firstly. The MLD can be derived by the moment result of the CP, which is defined as the probability of received signal-to-interference-ratio (SIR) is larger than a threshold. To obtain the moment result of the CP, the analytical result of the CP and its approximation with a lower computational complexity are derived. However, the moment result of CP is difficult to be validated via simulations. Therefore, we derive the meta distribution, which is calculated directly from the moment result of the CP, and validate its correctness by Monte Carlo simulations. Equipped with this, the analytical results of MLD in closed-form are derived. Based on these results, the effect of UE activity factor and the uplink power control (PC) factor on the MLD are analysed. The results show that the fog access points (FAP) density has no effect on the MLD.

The 1000 fold capacity enhancement is one of the key requirements in the future 5G networks, stimulating the interest in jointly adopting several advanced techniques (e.g. multiple antennas and heterogeneous networks (HetNets)). Analysis of the performance of the HetNets jointly with multiple antennas becomes crucial. In this paper, we analyse the K-tier multi-antenna HetNets from a downlink coverage perspective. The coverage probability is derived using the Gil-Pelaez inversion theorem under the stochastic geometry framework. Moreover, a closed form approximated result is obtained for observing the influence of the normalized range bias (NRB). The result shows that our proposed result closely match the Monte Carlo simulation, and the approximation result is effective for searching the optimal NRB.

The Long Term Evolution-Unlicensed (LTE-U) scheme has been proposed to exploit the unlicensed spectrum, especially the 5 GHz band, for cellular networks to further increase capacity. Since the 5 GHz band has already been used by WiFi networks, the carrier sense adaptive transmission (CSAT) scheme has been proposed LTE-U access points (LAPs) to harmoniously coexist with WiFi access points (WAPs), where duty cycles are used by LAPs to leave certain time slots that only allow WAPs to access the unlicensed band. However, the performance of the CSAT scheme has not been sufficiently analyzed for multiple unlicensed channels (UCs) in a large-scale network. In this work, we derive the explicit expressions of downlink successful transmission probabilities (STPs) of LAP users and WAP users for a large-scale multi-UC network using stochastic geometry tools. Based on the derived STPs, the fairness between the LTE-U network and the WiFi network, which is defined as the minimum throughput of LTE-U and WiFi users, are analysed versus the duty cycle and the LAP density. Furthermore, the optimal duty cycle is obtained based on the derived STPs in the duty-cycle and non-duty-cycle durations. Our results show that for a given number of UCs and WAP density, and with the optimal duty cycle, the optimal LAP density increases with the increasing number of UCs for maximizing the fairness, but leads to a poorer minimum throughput performance.

In indoor environments, the performance of ultra-dense cellular networks is significantly affected by blockages, especially for ultra-high frequencies. However, previous works either ignore or simplify such effects for analysing the networks. On the basis of stochastic geometry, this paper proposes a mathematically tractable approach to analyse ultra-dense networks in indoors, which considers both the effects of wall blockages and the distance-based path loss. The effects of wall blockages are firstly investigated by modelling the walls as a Boolean scheme of straight lines on a finite plane. Then a path loss model incorporating both the blockage-based and distance-based path loss is applied to analyse the performance of indoor networks. Finally, the analytical result is validated by comparing it with Monte Carlo simulations. The simulation results also show that the optimum transmitter density is finite for indoor ultra-dense networks with blockages, although the coverage probability benefits from the increase of transmitter density.

The Reduced Power Subframes (RPS) are encouraged to be applied in the LTE-Advanced Heterogeneous Networks (HetNets), to reduce the capacity loss caused by the Almost Blank Subframes (ABS). However, the RPS are supposed to be used in the macrocells only. In fact, the RPS can also be used in femtocells to mitigate the interference that the macrocell edge users suffers, but its performance is not investigated yet. In this paper, we introduce the RPS both in the macrocells and the femtocells. The results of the Signal to Interference Ratio (SIR) coverage probability under the stochastic geometry framework are derived in a closed-form which is verified through Monte Carlo simulation. Based on these results, the macrocell edge users' SIR coverage and the average rate coverage of the network (the average fraction of users achieving a target rate) are analysed numerically. Our proposed scheme enhanced both the SIR of the macro edge users and the average rate coverage probabilities.

Although the almost blank subframes (ABSFs) proposed in heterogeneous cellular networks can enhance the performance of the cell range expansion (CRE) user equipments (UEs), it significantly degrades the macro-cell total throughput. To address this problem, the low power subframes (LPSFs) are encouraged to be applied in the macro-cell center region by the further-enhanced inter-cell interference coordination. However, the residual power of the LPSF, which interferes with the CRE UEs, and the proportion of the LPSF affect the downlink throughput together. To achieve a better rate coverage probability, the appropriate LPSF power and the proportion are required. In this paper, the analytical results of the overall signal to interference and noise ratio coverage probability and the rate coverage probability are derived under the stochastic geometric framework. The optimal region bias ranges for maximizing the rate coverage probability are also analyzed. The results show that the ABSF still outperform the LPSF in terms of rate with the optimal range expansion bias, but lead to a heavier burden on the backhaul of the pico-cell. However, with a static range expansion bias, the LPSF provide better rate coverage than the ABSF. In addition, in a low-range expansion scenario, the reduced power of the LPSF has negligible effect on the rate coverage with the optimal resource partitioning.