Zhenhua Wang

The beetle, of the order Coleoptera, possesses outstanding flight capabilities. After completing flight, they can fold their hindwings under the elytra and swiftly unfold them again when they take off. This sophisticated hindwing structure is a result of biological evolution, showcasing the strong environmental adaptability of this species. The beetle’s hindwings can provide biomimetic inspiration for the design of flapping-wing micro air vehicles (FWMAVs). In this study, the Asian ladybird (Harmonia axyridis Pallas) was chosen as the bionic research object. Various kinematic parameters of its flapping flight were analyzed, including the flight characteristics of the hindwings, wing tip motion trajectories, and aerodynamic characteristics. Based on these results, a flapping kinematic model of the Asian ladybird was established. Then, three bionic deployable wing models were designed and their structural mechanical properties were analyzed. The results show that the structure of wing vein bars determined the mechanical properties of the bionic wing. This study can provide a theoretical basis and technical reference for further bionic wing design.

Humans tend to instinctively integrate information from varioussenses such as vision and touch to accomplish dynamic adjustmentswhen grasping and manipulating objects.

Monocular depth estimation has been actively studied in fields such as robot vision, autonomous driving, and 3D scene understanding. Given a sequence of color images, unsupervised learning methods based on the framework of Structure-From-Motion (SfM) simultaneously predict depth and camera relative pose. However, dynamically moving objects in the scene violate the static world assumption, resulting in inaccurate depths of dynamic objects. In this work, we propose a new method to address such dynamic object movements through monocular 3D object detection. Specifically, we first detect 3D objects in the images and build the per-pixel correspondence of the dynamic pixels with the detected object pose while leaving the static pixels corresponding to the rigid background to be modeled with camera motion. In this way, the depth of every pixel can be learned via a meaningful geometry model. Besides, objects are detected as cuboids with absolute scale, which is used to eliminate the scale ambiguity problem inherent in monocular vision. Experiments on the KITTI depth dataset show that our method achieves State-of-The-Art performance for depth estimation. Furthermore, joint training of depth, camera motion and object pose also improves monocular 3D object detection performance. To the best of our knowledge, this is the first work that allows a monocular 3D object detection network to be fine-tuned in a self-supervised manner.

Semantic simultaneous localization and mapping (SLAM) with a monocular camera is particularly attractive because of the deployment simplicity and economic availability. Data association problem which assigns unique identities for objects shown in multiple frames plays a fundamental role in semantic SLAM. Previous prevalent methods which mainly focused on associating geometric KeyPoints are no longer suitable. Some naive methods that rely on object distance or 2-D/3-D Intersection over Union are also vulnerable when occlusions happen. In this article, we propose a novel data association method for cuboid landmarks based on Dirichlet process mixture model. By jointly considering object class, position, and size, our method can perform data association robustly. We evaluated our method in simulated datasets, public benchmark KITTI, and on a real robot in an office environment. Experimental results show that our method not only associates cuboids robustly but also achieves SOTA pose estimation accuracy in monocular SLAMs.

Coal is an indispensable energy source for humans. As an important part of the mining industry, intelligent separation of coal and gangue will promote development. The traditional methods of recognition do not consider the interference created by a dynamic environment. There are many problems such as noise, complex backgrounds and occlusion, which lead to low accuracy and cannot satisfy real-time requirements in mining. Aiming at dynamic environments, a real-time multilevel fusion recognition system was built in this paper. First, we introduced a near-infrared camera into the field of separation, which was used to form a binocular system with a visible light camera. The SVM classifier was obtained by feature selection and fusion training of the binocular system, which overcomes the interference of environmental factors. Then, we proposed a new deep learning training method of two-sample fusion to improve the recognition network performance by expanding the number of samples and features. Finally, the SVM and deep learning algorithms were combined to establish a fast detection strategy. In addition, the length suppression algorithm was added to solve the occlusion problem. The accuracy of the fusion algorithm was 0.923 and the detection speed was increased to 26 fps. The experimental results indicated that the sorting system satisfied the requirements of real-time and robust of the coal industry.

As an important part of a factory’s automated production line, industrial robots can perform a variety of tasks by integrating external sensors. Among these tasks, grasping scattered workpieces on the industrial assembly line has always been a prominent and difficult point in robot manipulation research. By using RGB-D (color and depth) information, we propose an efficient and practical solution that fuses the approaches of semantic segmentation and point cloud registration to perform object recognition and pose estimation. Different from objects in an indoor environment, the characteristics of the workpiece are relatively simple; thus, we create and label an RGB image dataset from a variety of industrial scenarios and train the modified FCN (Fully Convolutional Network) on a homemade dataset to infer the semantic segmentation results of the input images. Then, we determine the point cloud of the workpieces by incorporating the depth information to estimate the real-time pose of the workpieces. To evaluate the accuracy of the solution, we propose a novel pose error evaluation method based on the robot vision system. This method does not rely on expensive measuring equipment and can also obtain accurate evaluation results. In an industrial scenario, our solution has a rotation error less than two degrees and a translation error < 10 mm.

The redundant humanoid manipulator has characteristics of multiple degrees of freedom and complex joint structure, and it is not easy to obtain its inverse