With the rapid development of artificial intelligence technology, the multi-UAV cooperative search has wide applications in the field of Internet of Things, such as resource exploration, emergency rescue, intelligent transportation, etc. However, the communication network in an unknown environment may be inaccessible, and the real-time information sharing among UAVs cannot be guaranteed, resulting in the failure of cooperative search. Aiming at this issue, this article is devoted to the design of the multi-UAV flight strategy to improve the cooperative search capability in an uncertain communication environment. Specifically, a new cooperative architecture oriented to a local communication network is devised to control the observation locations of multiple UAVs in the search process, and some local communication networks are established based on the distance among UAVs to meet the requirements of the search task. On this foundation, we develop a multi-UAV cooperative search model (MCSM) with communication cost and formation benefit as an optimization function to ensure the effectiveness of multi-UAV search. Moreover, in the process of model solving, an improved sparrow search algorithm (ISSA) is presented with some different search strategies to enhance the optimization capability. To verify the superiority of the proposed method, we designed several groups of simulation experiments to analyze the performance of MCSM. Experimental results illustrate that our method can not only maintain high cooperative search accuracy but also has high stability and convergence speed.

The soaring demand for intelligent mobile applications calls for deploying powerful deep neural networks (DNNs) on mobile devices. However, the outstanding performance of DNNs notoriously relies on increasingly complex models, which in turn is associated with an increase in computational expense far surpassing mobile devices’ capacity. What is worse, app service providers need to collect and utilize a large volume of users’ data, which contain sensitive information, to build the sophisticated DNN models. Directly deploying these models on public mobile devices presents prohibitive privacy risk. To benefit from the on-device deep learning without the capacity and privacy concerns, we design a private model compression framework RONA. Following the knowledge distillation paradigm, we jointly use hint learning, distillation learning, and self learning to train a compact and fast neural network. The knowledge distilled from the cumbersome model is adaptively bounded and carefully perturbed to enforce differential privacy. We further propose an elegant query sample selection method to reduce the number of queries and control the privacy loss. A series of empirical evaluations as well as the implementation on an Android mobile device show that RONA can not only compress cumbersome models efficiently but also provide a strong privacy guarantee. For example, on SVHN, when a meaningful (9.83,10−6)-differential privacy is guaranteed, the compact model trained by RONA can obtain 20× compression ratio and 19× speed-up with merely 0.97% accuracy loss.

Research on multiple unmanned aerial vehicles (UAVs) cooperative surveillance systems serving Internet of Things (IoT) applications, such as smart cities, precision logistics, etc., has become a hot topic. However, the target movement is unpredictable in an uncertain environment, and multiple UAVs are affected by obstacles or inaccessible regions, resulting in the decreased surveillance performance and even the loss of the target. This article is dedicated to determine the current surveillance environment through the 2-D laser scanner. At the cost of the energy consumption and the transmission unreliability, a multi-UAV cooperative path optimization (MCPO) model is designed to adjust the surveillance location of each UAV, which improves the target surveillance performance. Specifically, for different types of obstacles or inaccessible regions, we present a novel obstacle-avoidance selection strategy with two mechanisms in mind: 1) when some of UAVs encounter obstacles, but others can accurately monitor the target, a strict constraint mechanism is established to promptly adjust the surveillance location of each UAV, which ensures the accuracy of formation surveillance and 2) when all UAVs have to avoid obstacles, a fuzzy constraint mechanism is presented and combined with Lucas–Kanade (LK) method to expand the search range of the multi-UAV and enhance the flexible adjustment capability of the formation. To verify the superiority of the proposed optimization method, we develop a 3-D simulation experiment environment based on the UE4 platform and design several groups of experiments to analyze the effectiveness of MCPO. The experimental results demonstrate that MCPO can not only maintain the flight stability of multiple UAVs but also has satisfactory formation flexibility and surveillance accuracy.

With the development of smart devices, the computing capabilities of portable end devices such as mobile phones have been greatly enhanced. Meanwhile, traditional cloud computing faces great challenges caused by privacy-leakage and time-delay problems, there is a trend to push models down to edges and end devices. However, due to the limitation of computing resource, it is difficult for end devices to complete complex computing tasks alone. Therefore, this article divides the model into two parts and deploys them on multiple end devices and edges, respectively. Meanwhile, an early exit is set to reduce computing resource overhead, forming a hierarchical distributed architecture. In order to enable the distributed model to continuously evolve by using new data generated by end devices, we comprehensively consider various data distributions on end devices and edges, proposing a hierarchical federated learning framework FLEE , which can realize dynamical updates of models without redeploying them. Through image and sentence classification experiments, we verify that it can improve model performances under all kinds of data distributions, and prove that compared with other frameworks, the models trained by FLEE consume less global computing resource in the inference stage.

UAV surveillance and tracking have attracted great enthusiasm in intelligent transportation, and various approaches have been reported up to now. However, these approaches often ignored the uncertainties in the urban environment, such as occlusion, view change, and background clutter. Ignoring these uncertain factors often leads to a reduction in surveillance performance and tracking quality. This study devotes to improving the cooperative surveillance capability of multi-UAV formation by designing different cooperative strategies in the urban environment. To be specific, a novel cooperative architecture is designed to control the observation locations of multiple UAVs throughout the formation process. For different types of interference, we introduce a novel target recognition rate of each UAV as the decision factor and design corresponding cooperative strategies to guarantee the accuracy of cooperative surveillance. Based on this architecture, we develop a vision-based method of cooperative surveillance and tracking by multiple UAVs (SMART) whose objective function is the motion cost and flight reliability of UAVs to ensure that each UAV can be in the optimal surveillance location for the target. The proposed SMART skillfully integrates the strict, elastic, and flight constraint strategies. During the execution of the multi-UAV formation, the inherent safety constraints of multiple UAVs and the designed strategies are used to solve the quadratic optimization model to adjust the locations of these UAVs. To demonstrate the superiority of our method, we conduct a 3D simulation urban environment and devise several experiments to analyze the performance of SMART on it. The experimental results demonstrate that SMART can not only maintain the high cooperative flight capability, but also provide high flexibility and fault tolerance.

Background: Clinical notes record the health status, clinical manifestations and other detailed information of each patient. The International Classification of Diseases (ICD) codes are important labels for electronic health records. Automatic medical codes assignment to clinical notes through the deep learning model can not only improve work efficiency and accelerate the development of medical informatization but also facilitate the resolution of many issues related to medical insurance. Recently, neural network-based methods have been proposed for the automatic medical code assignment. However, in the medical field, clinical notes are usually long documents and contain many complex sentences, most of the current methods cannot effective in learning the representation of potential features from document text. Methods: In this paper, we propose a hybrid capsule network model. Specifically, we use bi-directional LSTM (Bi-LSTM) with forwarding and backward directions to merge the information from both sides of the sequence. The label embedding framework embeds the text and labels together to leverage the label information. We then use a dynamic routing algorithm in the capsule network to extract valuable features for medical code prediction task. Results: We applied our model to the task of automatic medical codes assignment to clinical notes and conducted a series of experiments based on MIMIC-III data. The experimental results show that our method achieves a micro F1-score of 67.5% on MIMIC-III dataset, which outperforms the other state-of-the-art methods. Conclusions: The proposed model employed the dynamic routing algorithm and label embedding framework can effectively capture the important features across sentences. Both Capsule networks and domain knowledge are helpful for medical code prediction task.