7 March 2024 in IEEE Transactions on Instrumentation and Measurement
Deep-learning-based detection methods have been widely applied to industrial defect inspection. However, directly using vanilla detection methods fails to achieve satisfying performance due to the lack of identifiable features. In this paper, a novel attention-based multi-scale feature fusion method (AMFF) is proposed, aiming to enhance defect features and improve defect identification by leveraging attention mechanism in the feature fusion. AMFF includes self-enhanced attention module (SEAM) and cross-enhanced attention module (CEAM). SEAM is performed on a single feature map, which first adopts multiple dilation convolutions to enrich contextual information without compromising resolution and then utilizes the intra-layer attention on the current feature map. CEAM takes both the current feature map and the adjacent feature map as input to perform cross-layer attention. The adjacent feature map is modulated with the guidance of the current feature map, which is then combined with the current feature map and the output of SEAM for final prediction. AMFF is utilized in current feature fusion networks, e.g., FPN and PAFPN, and is further integrated into prevalent detectors to guide them to pay more attention to defects rather than the background. Extensive experiments are conducted on two real industrial datasets released by Tianchi platform, i.e., fabric and aluminum defect datasets. For each dataset, 500 images are randomly selected for test and the rest for training. The proposed AMFF is demonstrated to significantly boost defect detection accuracy with acceptable computational cost, and the real-time performance could fully satisfy practical requirements.
4 November 2021 in International Journal of Automation and Computing
5 January 2024 in International Journal of Control, Automation, and Systems
23 May 2025 in IEEE/ASME Transactions on Mechatronics
This study presents a novel flapping-wing vehicle (FWV) with an active wing twisting mechanism, inspired by avian feather adaptations, to improve aerodynamic efficiency and stability. The mechanism, implemented via a conical rocket mechanism and crank-slider system, reduces drag, and optimizes thrust generation by aligning the wings with airflow during the downstroke and actively twisting them during the upstroke. A coupled flapping-twisting dynamic model is developed to guide the design, and a lightweight prototype is constructed with a bearing-supported cantilever system enabling controlled torsional motion. A series of aerodynamic force measurements and flight tests validate the design's effectiveness, showing a 22% reduction in power consumption and a significant increase in lift compared to conventional nontwisting wings. These findings highlight the potential of active wing twisting to enhance FWV performance in real-world applications, including urban surveillance, disaster response, and ecological monitoring.
14 November 2023 in IEEE Robotics & Automation Magazine
Disturbances caused by unknown dynamics and environmental factors render the automatic control of underwater vehicles extremely challenging. These effects are complex, time varying, and difficult to model accurately, leading to possible instability in the control process. This article focuses on the disturbance rejection problem of an underactuated bionic underwater vehicle (BUV). A novel control approach that combines the robustness of a model predictive controller with the modeling capabilities of a learning-based observer is proposed. More specifically, a practical Gaussian process (GP)-based disturbance observer is designed for online disturbance estimation. The interactive prediction results are integrated into a model predictive controller to obtain a refined dynamics model. The control framework optimally solves for the optimal control output to achieve control at the dynamics level. The proposed approach realizes disturbance rejection control while ensuring real-time application, constraint satisfaction, and recursive feasibility under complicated disturbances. The feasibility and efficiency of the proposed approach are verified through simulations and real-world experiments.
2 September 2021 in International Journal of Control, Automation, and Systems