Large neural network models are hard to deploy on lightweight edge devices demanding large network bandwidth. In this article, we propose a novel deep learning (DL) model compression method. Specifically, we present a dual-model training strategy with an iterative and adaptive rank reduction (RR) in tensor decomposition. Our method regularizes the DL models while preserving model accuracy. With adaptive RR, the hyperparameter search space is significantly reduced. We provide a theoretical analysis of the convergence and complexity of the proposed method. Testing our method for the LeNet, VGG, ResNet, EfficientNet, and RevCol over MNIST, CIFAR-10/100, and ImageNet datasets, our method outperforms the baseline compression methods in both model compression and accuracy preservation. The experimental results validate our theoretical findings. For the VGG-16 on CIFAR-10 dataset, our compressed model has shown a 0.88% accuracy gain with 10.41 times storage reduction and 6.29 times speedup. For the ResNet-50 on ImageNet dataset, our compressed model results in 2.36 times storage reduction and 2.17 times speedup. In federated learning (FL) applications, our scheme reduces 13.96 times the communication overhead. In summary, our compressed DL method can improve the image understanding and pattern recognition processes significantly.

Most researchers focus on designing accurate crowd counting models with heavy parameters and computations but ignore the resource burden during the model deployment. A real-world scenario demands an efficient counting model with low-latency and high-performance. Knowledge distillation provides an elegant way to transfer knowledge from a complicated teacher model to a compact student model while maintaining accuracy. However, the student model receives the wrong guidance with the supervision of the teacher model due to the inaccurate information understood by the teacher in some cases. In this paper, we propose a dual-knowledge distillation (DKD) framework, which aims to reduce the side effects of the teacher model and transfer hierarchical knowledge to obtain a more efficient counting model. First, the student model is initialized with global information transferred by the teacher model via adaptive perspectives. Then, the self-knowledge distillation forces the student model to learn the knowledge by itself, based on intermediate feature maps and target map. Specifically, the optimal transport distance is utilized to measure the difference of feature maps between the teacher and the student to perform the distribution alignment of the counting area. Extensive experiments are conducted on four challenging datasets, demonstrating the superiority of DKD. When there are only approximately 6% of the parameters and computations from the original models, the student model achieves a faster and more accurate counting performance as the teacher model even surpasses it.

This paper proposes a new global model aggregation method based on using zero-knowledge federated learning (ZKFL). The purpose is to secure horizontal or P2P federated machine learning systems with shorter aggregation times, higher model accuracy, and lower system costs. We use a model parameter-sharing Chord overlay network among all client hosts. The overlay guarantees a trusted sharing of zero-knowledge proofs for aggregation integrity, even under malicious Byzantine attacks. We tested over popular datasets, Fashion-MNIST and CIFAR10, to prove the new system protection concept. Our benchmark experiments validate the claimed advantages of the ZKFL scheme in all objective functions. Our aggregation method can be applied to secure both rank-based and similarity-based aggregation schemes. For a large system with over 200 clients, our system takes only 3 seconds to yield high-precision global machine models under the ALIE attacks with the Fashion-MNIST dataset. We have achieved up to 85% model accuracy, compared to only 3%$\sim$45% accuracy observed with federated schemes without protection. Moreover, our method demands a low memory overhead for handling zero-knowledge proofs as the system scales greatly to a larger number of client nodes.

Recently, the advancements in edge computing have boosted the deployment of video analysis systems based on deep learning, which breaks the limitation of the constrained communication and computing resources of local devices. However, processing multi-scene high-resolution video streams in crowd surveillance remains a significant challenge since it is difficult to formulate dynamic video content and communication environments to support offloading decisions. To bridge the gap between applications and modeling, this paper presents aReal-TimeCloud-edge-deviceCollaboration framework, which enables fast and accurateCrowd counting (RT3C) on the real dataset. RT3C comprises key frame detection, adaptive patch partition, patch encoder and decoder and computation offloading decision, designed to divide key frames into a minimum number of patches and determine the offloading location of patches. A Real-Time Multi-Agent Actor-Critic (RTMAAC) algorithm based on multi-agent reinforcement learning is proposed to decide whether to compute patches with a lightweight model on edge or a large model on cloud. Unlike traditional approaches ignoring the contents, RTMAAC is a dynamic online decision algorithm based on context of the network and video. Extensive experiments demonstrate that RT3C effectively discriminate the valid frames and optimizes offloading decisions in complex environments, outperforming other baseline algorithms on the two crowd counting datasets. In summary, RT3C provides a promising framework for multi-scene video streams, which can be extended to other applications to realize video computation based on deep models.

In the big data era, resource-constrained mobile devices generate an overwhelmingly large amount of data with complex tasks that demand distributed execution. Offloading computation-intensive tasks to nearby edge clouds is promising to solve this problem. However, mobile end devices cannot handle heterogeneous or delay-sensitive tasks. These end devices are also energy constrained with weak adaptability to environment changes. To address and tackle these problems, we present a two-moduletransfer reinforcement learning(TRL) framework for adaptive task offloading. A domain adaptation module is used to align heterogeneous characteristics of mobile devices. The TRL makes offloading decisions with adeep reinforcement learning(DRL) module. We evaluate the performance of TRL through real-world experiments on edge clouds. Our experiment results show that TRL reduces the task processing time by a factor of 20% from using three well known DRL methods. Our method achieved (15.4$\sim$40)% reduction in task drop rate over these methods. With domain adaptation, the TRL results in (50$\sim$80)% reduction in model convergence time. These advantages in using the TRL framework make it appealing in real-life edge computing applications.

Drone-swarm-assisted mobile edge computing (MEC) provides extra computation and storage capacity for smart city applications and the Industrial Internet of Things. To solve the problems of traditional fixed base stations in a complex terrain, including cost of deployment, transmission loss of telecommunication, and limited coverage, this article brings forward the unmanned aerial vehicles (UAVs) as MEC nodes in the air. For the purpose of matching the dynamic mobile devices and UAV trajectory, this article raises a multi-UAVs-assisted MEC offloading algorithm based on global and local path planning controlled by ground station and onboard computer. Firstly, this article considers a drone swarm scheduling and allocation strategy based on the priority of monitoring areas, UAVs residual energy and distance to target points, so as to minimize the global flight length and energy consumption. Secondly, based on user mobility, this article calculates the optimal communication coverage of a UAV, and jointly optimizes the local path planning and computing offloading, so as to maximize the number of offloading services and minimize the total latency in completing the computation task. Finally, based on the total latency and energy consumption of path planning and computation offloading, a UAV cluster computation offloading strategy with optimized energy efficiency is realized. Experimental results prove that the proposed algorithm can provide more offloading services while obtaining shorter path length and greater energy efficiency.

The future ubiquitous network, which is mainly characterized by full coverage communication, air-ground integration, multidimensional fusion, network reconfiguration and sensing-communication-computing integration, has become the development trend of 6G technology. The realization of ubiquitous coverage and perceptive fusion of IoT-UAV-Edge is an urgent problem to be solved for complex fusion services. Therefore, this paper proposes a drone-enabled smart air agent in 6G edge fusion system. Firstly, the energy efficient dynamic routing strategy based on joint air-ground control optimization is designed to improve the fusion sensing performance and prolong the service time of drone swarm. Then, the system integration of user-IoT-UAV-Edge is realized to achieve the functionalities of perception, transmission, computing and analysis. Finally, an airborne data fusion mechanism based on multi-source sensing is designed to solve the associated cognitive optimization problem for multi-modal information. The experimental results invalidate the effectiveness and practicability of our system on autonomous path planning, effective computing offloading and accurate airborne fusion.

With the widespread application of artificial intelligence and the Internet of Things, the intellectualization of the industrial Internet of Things (IIoT) has received more and more attention. However, in the application scenario with numerous sensors, the contradiction between massive requests of computing tasks and high requirements of inference quality affects the operation efficiency and service reliability. Moreover, due to the heterogeneity of computing resources and the randomness of communication environments of the cloud-edge system, how to compute and deploy deep learning models in a cloud-edge collaborative environment has also become a challenging problem. Therefore, this article presents a collaborative cloud-edge service cognitive framework for deep neural network (DNN) model service configuration to provide dynamic and flexible computing services. In order to adapt to different service requirements, we explored the tradeoffs between accuracy, latency, and energy consumption indicators, and a revenue target is established, which considers the quality of service experience and the system energy consumption to improve resource utilization efficiency. By transforming the optimization of the revenue target into a partially observable DNN configuration reinforcement learning problem, a dueling deep Q-learning network-based self-adaptive DNN configuration algorithm is proposed. Experimental results show that the proposed mechanism can effectively learn from external experience, adapt to the dynamic network environment, and reduce delay and energy consumption while meeting the service requirements.

The outbreak of the corona virus disease 2019 (COVID-19) has caused serious harm to people’s physical and mental health. Due to the serious situation of the epidemic, a lot of negative energy information increases people’s psychological burden. However, effective interventions against mental health problems are not in abundance. To address such challenges, in this article, we propose the concept of negative information to describe information that has a negative impact on people’s mental health. To achieve the measurement of negative information, the level of mental health inversely measures the degree of negative information. Specifically, we design a system to measure the negative information used to monitor the mental health state of the user under the impact of negative information. The cognition of mental health is realized based on the intelligent algorithm deployed on the edge cloud, and the needs of users can be responded to in real time in practical applications. Finally, we use real collected dataset to verify the influence of negative information. The experiments show that the system can achieve negative information measurement and provide an effective countermeasure for solving mental health problems during a pandemic situation.

Unmanned aerial vehicles (UAVs) with communication, computing, and storage capabilities have high mobility. Based on this advantage, it can push the service closer to the user. Our research group is concerned with implementing the Internet of Things (IoT) enabled massive crowd management platform that employs 5G to facilitate network connectivity among the UAV and sensory networks. In such a highly dynamic environment, IoT devices, users, and UAVs are the key factors to determine the caching strategies. Due to the limitations of drone batteries and changes in UAV cluster density, the environment is characterized as highly dynamic. However, the existing UAV caching strategy does not consider both the changes of the users and UAVs. Therefore, this paper proposes a three-layer UAV cache architecture in 5G network to achieve hierarchical adaptation to the dynamic changes of users and UAVs. Based on this architecture, we propose a dual dynamic adaptive caching(DDAC) algorithm. The DDAC algorithm is divided into two parts: user adaptation and UAV adaptation. For user adaptation, we designed a user-adaptive UAV trajectory model, which ensures the transmission efficiency of the UAV. For UAV adaptation, we designed and deployed a UAV-adaptive cache model based on a greedy algorithm in the cognitive center layer. The UAV can dynamically adjust the caching strategy according to the cluster density. Finally, the results of the experiment prove that our proposed UAV adaptive cache model has better performance in the cache hit ratio compared with the existing UAV cache model.

This article proposes a novel ultralarge-scale crowd monitoring system, namely, the ULCM system. The ULCM system enables advanced sensing and networking technologies aimed at collecting and processing multimodal, multiperspective, and real-time crowding data relevant to crowd management. This data will be further analyzed to provide a global realization of evolving events over a large geographical area as they occur in real time. The ULCM is the infrastructure component of an intelligent platform that is being developed by our research group to provide crowd intelligence to decision makers through an interactive digitized visual environment. In order to achieve a full comprehensive scene overview, the ULCM deployment utilizes a multiplicity of unmanned aerial vehicle (UAV) agents in different operational scenarios. The aerial deployment and control are realized by custom multiple UAV networks and airborne LiDAR sensors. The deployment and control on the ground sensory agents are based on multiple subnetworks, including closed-circuit television (CCTV) and infrared gas and ultrasonic sensors networks. Eventually, ULCM employs the software-defined network (SDN) and edge cloud technologies to optimize the networking and data analytics performance from the perspective of infrastructure.

With the increase in natural gas consumption, distributed natural gas supply and transaction have become new development goals of the industrial Internet of Things (IoT) for natural gas. However, there are obvious disadvantages of the existing natural gas pipeline network in aspects of infrastructure warning, multilevel data transmission, automatic transaction, and security. Emerging technologies, such as blockchain, edge computing, and AI have been introduced to address these shortcomings. This article proposes Smart Micro-GaS, i.e., the concept of a cognitive micro natural gas industrial ecosystem based on mixed blockchain and edge computing. Three aspects, multilevel, multiview, and multidimension, are put forward for its design and deployment. Then, based on the most important smart contract algorithm in blockchain, a mixed transaction model for natural gas is established. Finally, a case analysis is conducted on a smart natural gas testbed for data prediction and the proposed smart contract algorithm. The framework proposed in this article makes the natural gas data have multilevel liquidity and realizes diversified transactions.

With the rapid development of wireless communication and flight control technologies, the unmanned aerial vehicles (UAVs) have been widely used in multiple application scenarios. A typical scenario is massive crowd management of the multi-millions annual Hajj Pilgrimage to Mecca where UAVs are widely utilized to conduct crowd monitoring by carrying sensory devices. The safe flight of a UAV is crucial for ensuring the successful execution of missions. With the aim to overcome the disadvantage caused by the ground station intrusion detection, the combination of UAV and airborne LiDAR has been widely studied in the field of UAV obstacle recognition. This article studies the UAV network architecture under a common scenario and proposes an obstacle recognition and intrusion detection algorithm for UAV based on an airborne LiDAR (ALORID). First, the preprocessing of the data obtained by a LiDAR, i.e., the coordinate conversion of LiDAR data in combination with UAV motion parameters, is completed. Then, the LiDAR data graph at the current moment is generated by the image noisy point filtering algorithm. After that, the improved density-based spatial clustering of applications with noise (DBSCAN) algorithm is used for image clustering of intrusions to obtain the LiDAR time-domain cumulative graph in a certain detection time. Finally, the motion recognition and location detection of each cluster are completed. The experiment results verify the effectiveness of the proposed algorithm in identifying the moving state of the intrusions.

The drone is applied extensively, and the usage shows the rapidly growing trend. Due to technical problems or external malicious interference, the crash problem of drone is increasingly serious. The location and recycle of crashed drone is a key problem to be solved. We use the machine learning algorithm to assess the accuracy of drone crash site in the complex scene, and make the data statistics and test in combination with the true scene. After the further summary, this paper puts forward a complete drone location, search and recycling program, to provide the guidance scheme for the reuse of the crashed drone.

This paper sets up the real-time user information search system under the home environment by integrating the robot technology, intelligent wearable technology, speech recognition technology, speech synthesis technology, facial expression recognition technology, and cloud computing technology, and builds the user emotion perception means based on big data clouds, to make more convenient human-machine interaction and more accurate and reasonable feedback. RoCoSense system built in this paper can greatly enrich the intelligent system interactive means by the real-time perception of user emotion, and has a wide application prospect in the fields of consumption, entertainment, and health care.

As a new type of low power wide area (LPWA) technology, the narrowband Internet of Things (NB-IoT) technology supports wide coverage and low bitrate services, thus it has a great potential to be the future commercial technology of LPWA network. Therefore, it has attracted attention of both academia and industry. In this paper, we present the NB-IoT development, and main characteristics and design objectives of NB-IoT according to 3GPP R13. In addition, we provide the review of related literatures about NB-IoT modeling and algorithm analysis. And we explain current problems of NB-IoT system-level modeling based on visualized simulation platform. Moreover, this paper is devoted to the construction of the NB-IoT model based on OPNET and the verification of its characteristics, such as wide coverage and high channel utilization. This paper mainly considers NB-IoT model design and realization in terms of NB-IoT physical layer characteristics. We summarize the correlated characteristics of NB-IoT uplink and downlink. Then we design and construct the NB-IoT model based on Long Term Evolution (LTE) network. Lastly, we use the constructed NB-IoT model for simulations and conduct an experiment on it using the LTE network with channel bandwidths of 3 MHz, 5 MHz, 10 MHz, 15 MHz, and 20 MHz. The simulation results have verified the performance of NB-IoT, wherein uplink time delay is lower than 10 s, channel utilization is higher than that of LTE network, and coverage area is larger than LTE network.

With the development of LPWA (Low Power Wide Area) technology, the emerging NB‐IoT (Narrowband Internet of Things) technology is becoming popular with wide area and low‐data‐rate services. In order to achieve objectives such as huge amount of connection and wide area coverage within NB‐IoT, the problem of network congestion generated by random access of numerous devices should be solved. In this paper, we first introduce the background of NB‐IoT and investigate the research on random access optimization algorithm. Then we summarize relevant features of NB‐IoT uplink and narrowband physical random access channel and design random access with differentiated barring (RADB), which can improve the insufficiency of traditional dynamic access class barring method. At last, the algorithms proposed in this paper are realized with established NB‐IoT model using OPNET Modeler platform, and simulations are conducted. The simulation results show that RADB is able to effectively solve preamble request conflict generated by random access of numerous devices and preferentially provide efficient and reliable random access for latency‐sensitive devices.

Key Management protocol is one of the most important mechanisms for communication security, whereas its security analysis is critical to evaluate the information security. In this paper, we study a kind of group key management schemes which use key calculation in rekeying. At first, the security vulnerability in Code for Key Calculation (CKC) is analyzed. The codes in the key tree can be exposed to the user who should not get them. Thus, the user can get additional key information in group key updating process. Moreover, the user can continue to get the communication contents after he/she leaves the group. Sequentially, we construct two effective attacks and discuss the condition of successful attack. We analyze similar problems in other schemes. Finally, we propose an improved scheme to CKC, named Non-Code-aided Key Calculation (NCKC). Performance analysis and simulation results show that NCKC can fulfill forward and backward security at the cost of a little increase in communication overhead.