The introduction of unmanned aerial vehicle (UAV) will facilitate data collection of wireless sensor networks effectively. However, since the communication range setting of existing data collection methods does not take into account the difference between air-to-ground (AG) and ground-to-ground (GG) links, the same communication range is usually used. This results in unsatisfactory performance of existing methods in minimizing the task completion time. In this paper, by considering the different characteristics of AG and GG links, a more efficient data collection method is proposed, in which clustering of the ground nodes, UAV trajectory and data collection mode are jointly optimized. Simulation results show that the difference between AG and GG links will affect the task completion time significantly. Thanks to the more efficient data transmission and shorter trajectory length, task completion time of the proposed method reduces by 8.86%~74.38% in different scenarios compared with existing methods.

Spectrum sensing is of crucial importance for improving spectrum efficiency and realizing immersive communication. Deep learning (DL) has been introduced for spectrum sensing, with test statistics generated directly from signal samples in an automatic manner. However, most of the existing data-driven spectrum sensing methods are based on supervised learning and they usually require a massive amount of labeled training data to achieve high detection performance. It is difficult to obtain sufficient labeled training data in practice. To address this issue, a pseudo labeling multi-scale network (PLMSNet) for semi-supervised spectrum sensing is proposed to make the best use of a majority of unlabeled samples and achieves well detection performance with only a few of labeled training samples. Moreover, the proposed scheme is implemented in a real-world software defined radio (SDR) communication system. Both simulation and real-world experiments demonstrate that our proposed method achieves superior detection performance compared with the benchmark methods.

In wireless communications, unauthorized broadcasting within licensed spectrum bands disrupts legitimate signals and interferes with adjacent frequencies, risking critical systems. Existing methods for detecting unauthorized broadcasting often underutilize known signal data, reducing their effectiveness in dynamic, open-set scenarios. To address this, we propose a novel framework combining a temporal convolutional autoencoder (TCAE) with boundary-guided support vector data description (BGSVDD) for accurate detection of unauthorized signals in the radio frequency spectrum. The TCAE captures temporal signal features effectively with an adaptive temporal convolutional network (ATCN), while the BGSVDD uses a small set of known unauthorized samples to create robust decision boundaries with our proposed adaptive misclassification penalty (AMP) loss. Moreover, a global-local support vector (GLSV) strategy enables efficient online model updates, maintaining detection performance in evolving wireless environments with minimal resource overhead. Experiments with real-world broadcast signals show our method outperforms state-of-the-art techniques, especially under challenging interference conditions. Tests on public datasets further confirm its strong generalization across diverse spectrum protection applications.

Data driven unmanned aerial vehicle (UAV) position prediction methods have been widely used in flying ad hoc networks, which are mainly based on classical and machine learning based algorithms. These prediction methods use historical position series to predict the future positions. However, effectiveness of these methods in actual scenarios has not been fully verified, especially for high speed UAVs. This paper evaluates typical position prediction methods using real flight trajectories of fixed wing UAVs. The results show that accuracy of position prediction depends on the update speed of historical positions. The higher the update speed, the smaller the prediction error. However, the impact of update speed on different methods varies. Some methods have advantages at low update speeds, while others are superior at high update speeds. Moreover, characteristics of flight trajectory also affect the prediction performance, no matter which method is used.

Semantic communication (SemCom) is envisioned as a promising technique to break through the Shannon limit. However, semantic inference and semantic error correction have not been well studied. Moreover, error correction methods of existing SemCom frameworks are inexplicable and inflexible, which limits the achievable performance. In this paper, to tackle this issue, a knowledge graph (KG) is exploited to develop SemCom systems. Two cognitive semantic communication frameworks are proposed for the single-user and multiple-user communication scenarios. Moreover, a simple, general, and interpretable semantic alignment algorithm for semantic information detection is proposed. Furthermore, an effective semantic correction algorithm is proposed by mining the inference rule from the KG. Additionally, the pre-trained model is fine-tuned to recover semantic information. For the multi-user cognitive SemCom system, a message recovery algorithm is proposed to distinguish the messages of different users by matching the knowledge level and the context at the destination. Extensive simulation results conducted on a public dataset demonstrate that our proposed single-user and multi-user cognitive SemCom systems are superior to benchmark communication systems in terms of the data compression rate and communication reliability. Finally, we present realistic single-user and multi-user cognitive SemCom systems results by building a software-defined radio prototype system.

Implementing an resilient, efficient, and reliable network structure is crucial for highly dynamic unmanned aerial vehicle (UAV) swarms, for which flying ad hoc network (FANET) is the most suitable form. Similar to traditional ad hoc networks, the performance of FANET largely depends on the efficiency, reliability, and stability of routing schemes. However, unique characteristics of UAV make the routing design of FANET face more challenges. In order to better understand the development of FANET routing schemes, this paper attempts to clarify the current research status and grasp the future development trend of FANET routing by reviewing and analyzing relevant literatures in the past decade. Results show that geographic routing, delay tolerant network, and opportunity forward are possible countermeasures to the challenges of FANET routing.

Existing works on UAV network protocols generally use software simulators for performance evaluation, which makes the analysis results often differ significantly from the test results in actual environments. In order to make the analysis of UAV network protocols more realistic, it is necessary to introduce actual UAV nodes into the simulation. By drawing on the idea of hardware-in-the-loop (HIL) simulation, this paper proposes a simulation framework with real UAV nodes in the loop. To show the potentials of the proposed simulation framework, an instance for HIL simulation of routing protocols is implemented. Preliminary results indicate that using a small number of actual flying UAV nodes as an organic component of simulation could reduce the gap between simulation results and actual situations.

The introduction of unmanned aerial vehicle (UAV) makes low-power links face many new challenges in UAV-assisted wireless-sensor network (WSN). In this article, the spatial characteristics of these low-power links are analyzed comprehensively, and the applicability of traditional link quality metrics and models is evaluated. In particular, the differences between link characteristics of UAV-assisted WSN and those of traditional WSN are discussed, and possible reasons for these differences are analyzed. To achieve this goal, two UAVs working on different bands are used and three different flight modes are designed. Experimental results show that link characteristics of UAV-assisted WSN become more complicated due to the 3-D movement of UAV. Link quality fluctuates significantly when the altitude or horizontal distance changes, and it is difficult to demarcate a clear bound between the connected and transitional region. Traditional models for spatial characteristic are only applicable when there is no UAV communication interference and the direction between the air node and ground node remains unchanged. Meanwhile, existing models between physical-layer metrics and packet reception ratio are only applicable when there is no UAV communication interference. For links with the same quality, stability of UAV-assisted WSN is not significantly different from that of traditional WSN. Finally, there is more obvious overall asymmetry between uplinks and downlinks in UAV-assisted WSN, and the quality of downlinks is significantly better than that of uplinks. The discovery of these new features would have great significance on the design of UAV-assisted WSN.

Spectrum prediction is of crucial importance for realizing the cognitive Internet of Things to tackle the spectrum scarcity problem. Deep-learning-based spectrum prediction methods have attracted extensive attention due to their superior accuracy. However, the training speed of deep networks is low and the architecture of traditional networks is uninterpretable. In order to tackle these problems, a radio frequency machine-learning-driven spectrum prediction scheme is proposed by exploiting a novel model-enabled autoregressive (AR) network. A cell with only two parameters is exploited in each layer of the AR, which accelerates the network training. Moreover, the domain knowledge of the AR structure enables our proposed scheme to be explainable. Simulation results show that our proposed scheme has the best prediction accuracy than the long short-term memory (LSTM)-based scheme and the AR scheme. It is also shown that its convergence speed is higher than that of the LSTM-based scheme.

Wireless technique classification (WTC) is of crucial importance in Internet of Things for realizing efficient spectrum sharing and interference management. However, the existing deep-learning-based methods have low classification accuracy, especially at low signal-to-noise ratio levels. In this article, a multiscale convolutional neural network framework is proposed for WTC. A multiscale module is exploited to capture the higher abstraction features. Simulation results demonstrate that our proposed scheme can achieve a better classification performance and a higher convergence speed compared to the state-of-the-art schemes.

Parameters of low power radio transceiver such as transmit power are essential for the performance of link quality estimation, channel modeling, and node localization. In this paper, characteristics of channel frequency, transmit power, and receive power of a typical radio transceiver are evaluated through measurements. The results show that changes of transmit power and receive power are both different from ideal characteristics, especially the transmit power. The change of receive power is piece-wise linear, and there are obvious non-linear regions. Although the change of transmit power is approximately linear, there are obvious offsets among different channels of a single node and among the same channels of different nodes. It means that the calibration model of transmit power is device dependent. Two calibration models are proposed for both transmit power and receive power, respectively. Lookup table model produces better accuracy but more additional memory overhead for both the calibration of transmit power and receive power. Linear fitting model for receive power calibration can significantly reduce the memory overhead. However, its average error is 0.49dBm higher than that of lookup table model. Although some work believes that the calibration for receive power is device independent, our measurement results show that this assumption will bring significant errors. Lookup table with offset model for transmit power calibration could reduce the memory overhead obviously while only having an average error of 0.08dBm.

Most existing unmanned aerial vehicle (UAV) assisted wireless sensor network (WSN) studies usually assume that using a large SNR or a short distance can ensure good link quality. However, they don't consider the actual characteristics of UAV assisted WSN. In this paper, spatial characteristics of wireless links, link quality indicating capabilities of physical layer metrics and symmetry of uplink and downlink are analyzed by comparing the measured data of UAV assisted WSN and traditional WSN. The results show that link characteristics of UAV assisted WSN have changed significantly compared with those of traditional WSN. Firstly, connected region of the three traditional communication regions (connected region, transitional region, and disconnected region) is significantly compressed and even no longer exists, which makes that short distance can no longer guarantee reliable communication. Secondly, traditional physical layer metrics such as RSSI, SNR and LQI are no longer valid and even become deceptive, which may lead to wrong judgment of link quality. Finally, there is obvious overall asymmetry between uplink and downlink of UAV assisted WSN, and the quality of downlink is significantly higher than that of the uplink. The discovery of these new features would have great significance on guiding the design of UAV assisted WSN.

Cross-Technology Interference (CTI) greatly affects the performance of low power sensor networks, especially under severe WiFi interference. Fast and accurate packet reception rate estimation under CTI is crucial for improving network efficiency and reducing packet retransmissions. However, there are many drawbacks in existing approaches such as low accuracy, high overhead, and requirement of offline training. In this paper, a theoretical packet reception rate estimation approach is proposed, which combines bit error rate model of IEEE 802.15.4 with noise distribution ingeniously. This approach is fully distributed that could effectively eliminate the overheads caused by transmitting measurement packets or collecting statistics of data packets. In addition, offline data collection and training are no longer needed. More importantly, the proposed approach is almost unaffected by the CTI level. Compared with state-of-the-art approaches, estimate error of the proposed one is reduced by at least 4.83%~79.54% under different CTI levels.

Real-time and accurate packet reception rate estimation is crucial for wireless sensor networks. However, existing approaches usually rely on offline data collection and training, which limits their generality. Specifically, models fitted by the test data acquired under specific conditions cannot be used in all environments and for arbitrary packet sizes. In this paper, a simplification method for the theoretical bit error rate model of IEEE 802.15.4 2.4 GHz physical layer is proposed, which is about 18 to 38 times faster than the original one. Then, with the simplified model, a lightweight packet reception rate estimation approach is designed, which is self-adaptive to different environments and arbitrary packet sizes. With the proposed approach, offline data collection and training are no longer needed, which will reduce deployment cost effectively. Compared with state-of-the-art approaches, estimate error of the proposed one is reduced by 2.46%~74.97% in different environments, and by 2.46%~62.00% for different packet sizes.