Few-shot aerial image object detection aims to rapidly detect object instances of novel category in aerial images by using few labeled samples. However, due to the complex background of aerial images, few labeled samples of novel categories, and the model trained with the few-shot learning paradigm is biased towards the base categories, it greatly increases the difficulty of identifying foreground objects of novel categories. In addition to this, tiny object detection is always a hot potato in aerial image object detection, and it is even more difficult for few-shot object detection. To this end, we propose a Few-shot aerial image object detection with Confidence-Iou collaborative proposal filtration and Tiny object constraint loss (FsCIT). Specifically, we first introduce a new confidence-iou collaborative proposal filtration scheme to RPN, which combines the unbiased IoU scores between the two bounding boxes with foreground-background confidence scores to filter redundant region proposals and rescue more foreground proposals for the novel categories in RPN. Then, we design a new tiny object loss constraint term to attempt at overcoming the challenge of tiny object detection in few-shot aerial image object detection. This term considers the central point distance, the size of ground-truth bounding boxes, and the distances between the four edges of the ground-truth bounding box and the predicted bounding box. Experiments on DIOR, AI-TOD and HRRSD datasets show that FsCIT is effective and can improve the performance of few-shot aerial image object detection.

Unmanned system-of-systems (USoS) often operates in dangerous, dirty, and open environments, which causes them to be vulnerable to disruptions. Uncertainties and unexpected shocks can severely interfere with the USoS operation and adversely affect the overall USoS performance. Therefore, we propose a resilience metric and the corresponding dynamic assessment methods for homogeneous and heterogeneous USoS, and emphasize on establishing cooperative reconfiguration strategies for resilience enhancement. First, a multilevel analysis framework and mission-oriented resilience metric for USoS are established to comprehensively consider the variable performance and mission baseline. Subsequently, a theoretical analysis method for homogeneous USoS based on queuing theory is developed under a random attack. Furthermore, a heterogeneous USoS assessment algorithm is presented to determine how three different attacks and collaborative reconfiguration strategies affect the resilience of networked USoS. Finally, as a key demonstration, a 50-swarm USoS serves as a pivotal example to validate the efficacy of the proposed method, which can serve as a reference for improving the USoS resilience.

Practical manufacturing system operates in highly dynamic and uncertain environments, where stochastic disturbances disrupt the execution of the production schedule as originally developed. Previous dynamic scheduling mainly focuses on the constructing predictive models for machine unavailability, with little studies on the adaptive and self-learning capacities for changing scheduling environments. Therefore, a digital twin (DT) driven scheduling with a dynamic feedback mechanism is proposed, in which a reinforcement learning (RL) based adaptive scheduling is developed in DT to make corrective decisions for the disturbances during production runs. In the proposed architecture, the happening disturbance is first detected in the virtual layer by the status continuously updating in accordance with the physical workshop. Furthermore, the reschedule triggering condition is determined in real-time through the calculation of the progress deviations resulting from disturbances. For the scheduling approach, the distributed RL (DRL) based adaptive scheduling method is built to perceive the dynamic production status from virtual environment and implement corrective strategies to hedge against the occurred disturbances. Finally, the proposed method is verified by a practical job shop case and the corresponding DT system is developed to show the effectiveness and advantages after a practical implementation.

The dependency matrix is widely used in fault diagnosis as a valid diagnostic model. However, the conventional dependency matrix-based diagnosis method pays more attention to the extension and improvement of the dependency matrix, and the timeliness is rarely discussed. To solve the problem and improve the diagnosis efficiency, fast diagnosis method using two search algorithms is proposed. The hash function is introduced, and the dependency matrix is transformed into the key-value pairs. The binary search algorithm is brought in to locate the desired failure mode by splitting the dependency matrix. Thus, the fault diagnosis is converted into a search process, which can improve the diagnosis efficiency by making full use of the advantages of the search algorithms. Finally, a numerical simulation is utilized to verify the proposed method. The proposed methods are applied to matrices of different sizes. The results show that the hash-based diagnosis method has significantly better efficiency, making it more suitable for onboard diagnosis. On the other hand, the binary search-based method is advantageous from a memory perspective only when the matrix size is small.