Search results

Robot automatic grasping has important application value in industrial applications. Recent works have explored on the performance of deep learning for robotic grasp detection. They usually use oriented anchor boxes (OABs) as detection prior and achieve better performance than previous works. However, the parameters of their loss belong to different coordinates, this may affect the regression accuracy. This paper aims to propose an oriented regression loss to solve the problem of inconsistency among the loss parameters.

In the oriented loss, the center coordinates errors between the ground truth grasp rectangle and the predicted grasp rectangle rotate to the vertical and horizontal of the OAB. And then the direction error is used as an orientation factor, combining with the errors of the rotated center coordinates, width and height of the predicted grasp rectangle.

The proposed oriented regression loss is evaluated on the YOLO-v3 framework to the grasp detection task. It yields state-of-the-art performance with an accuracy of 98.8% and a speed of 71 frames per second with GTX 1080Ti on Cornell datasets.

This paper proposes an oriented loss to improve the regression accuracy of deep learning for grasp detection. The authors apply the proposed deep grasp network to the visual servo intelligent crane. The experimental result indicates that the approach is accurate and robust enough for real-time grasping applications.

The purpose of this paper is to detect damages in steel structures with actual connections, i.e. semi-rigid connections. The method will detect the damages by tracking the changes in the stiffness of structural members using only a limited number of dynamic responses and without knowing the type or time history of the dynamic force applied on the structure.

The paper proposes a technique that combines the iterative least-square and unscented Kalman filter (UKF) methods to identify the stiffness of beams and columns in typical two-dimensional steel-framed structures with semi-rigid connections. The detection of damages is by using nonlinear time-domain structural health monitoring method.

The technique is verified by using numerical examples using noise-free and noise-included dynamic responses from two different types of dynamic forces: harmonic and blast loads. The results showed that the UKF method with iterative least-square is a powerful approach to identify and detect damages in structures that have nonlinear behavior and the method was able to detect the damages in beams with a very high accuracy for noise-free and noise-included dynamic responses. In addition, the optimum number and locations of dynamic responses (accelerometer sensors) required for damage detection were determined.

This paper fulfills an identified need to detect damages in steel structures using only a limited number of accelerometer sensors.