The rapid development of artificial intelligence and remote sensing technologies is indispensable for modern agriculture. In orchard environments, challenges such as varying light conditions and shading complicate the tasks of intelligent picking robots. To enhance the recognition accuracy and efficiency of apple-picking robots, this study aimed to achieve high detection accuracy in complex orchard environments while reducing model computation and time consumption. This study utilized the CenterNet neural network as the detection framework, introducing gray-centered RGB color space vertical decomposition maps and employing grouped convolutions and depth-separable convolutions to design a lightweight feature extraction network, Light-Weight Net, comprising eight bottleneck structures. Based on the recognition results, the 3D coordinates of the picking point were determined within the camera coordinate system by using the transformation relationship between the image’s physical coordinate system and the camera coordinate system, along with depth map distance information of the depth map. Experimental results obtained using a testbed with an orchard-picking robot indicated that the proposed model achieved an average precision (AP) of 96.80% on the test set, with real-time performance of 18.91 frames per second (FPS) and a model size of only 17.56 MB. In addition, the root-mean-square error of positioning accuracy in the orchard test was 4.405 mm, satisfying the high-precision positioning requirements of the picking robot vision system in complex orchard environments.

Convolution is a crucial component of convolution neural networks (CNNs). However, the standard static convolution has two primary defects: data independence and the weak ability to integrate global and local features. This paper proposes a novel and efficient dynamic convolution method with global and local attention to address these issues. A building block called the Global and Local Attention Unit (GLAU) is designed, in which a weighted fusion of global channel attention kernels and local spatial attention kernels generates the proposed dynamic convolution kernels. The GLAU is data-dependent and has better adaptability and the ability to integrate global and local features into each layer. We refer to such modified CNNs with GLAUs as “GLAUNets”. Extensive evaluation experiments for image classification compared to classical CNNs and the state-of-the-art dynamic convolution neural networks were conducted on the popular benchmark datasets. In terms of classification accuracy, the number of parameters, and computational complexity, the experimental results demonstrate the outstanding performance of GLAUNets.

Addressing key challenges in unstructured environments, including local optimum traps, limited real-time interaction, and convergence difficulties, this research pioneers a hybrid reinforcement learning approach that combines simulated annealing (SA) with proximal policy optimization (PPO) for robotic arm trajectory planning. The framework enables the accurate, collision-free grasping of randomly appearing objects in dynamic obstacles through three key innovations: a probabilistically enhanced simulation environment with a 20% obstacle generation rate; an optimized state-action space featuring 12-dimensional environment coding and 6-DoF joint control; and an SA-PPO algorithm that dynamically adjusts the learning rate to balance exploration and convergence. Experimental results show a 6.52% increase in success rate (98% vs. 92%) and a 7.14% reduction in steps per set compared to the baseline PPO. A real deployment on the AUBO-i5 robotic arm enables real machine grasping, validating a robust transfer from simulation to reality. This work establishes a new paradigm for adaptive robot manipulation in industrial scenarios requiring a real-time response to environmental uncertainty.

The critical component of the vision transformer (ViT) architecture is multi-head self-attention (MSA), which enables the encoding of long-range dependencies and heterogeneous interactions. However, MSA has two significant limitations: its limited ability to capture local features and its high computational costs. To address these challenges, this paper proposes an integrated multi-head self-attention approach with a bottleneck enhancement structure, named WMSA–WBS, which mitigates the aforementioned shortcomings of conventional MSA. Different from existing wavelet-enhanced ViT variants that mainly focus on the isolated wavelet decomposition in the attention layer, WMSA–WBS introduces a co-design of wavelet-based frequency processing and bottleneck optimization, achieving more efficient and comprehensive feature learning. Within WMSA–WBS, the proposed wavelet multi-head self-attention (WMSA) approach is combined with a novel wavelet bottleneck structure to capture both global and local information across the spatial, frequency, and channel domains. Specifically, this module achieves these capabilities while maintaining low computational complexity and memory consumption. Extensive experiments demonstrate that ViT models equipped with WMSA–WBS achieve superior trade-offs between accuracy and model complexity across various vision tasks, including image classification, object detection, and semantic segmentation.

Driven by orchard labor shortages and rising demand for intelligent harvesting, automated apple picking requires a balance between conformal enveloping and slip-resistant stability. To reduce damage and slippage caused by fragile skins, variable morphologies, and motion disturbances, this study proposes a multi-joint pneumatic flexible apple-picking hand with adjustable circumferential configuration. Based on structural configuration determining grasping stability, six apple-morphology-based finger-base supports were designed. Parametric analysis of soft gripper cavities identified an isosceles trapezoidal profile as the best configuration. Using the Yeoh constitutive model, an equivalent joint model for conformal gripping was developed, and genetic algorithm (GA) optimization selected the four-joint design as the preferred configuration. Static finite element simulations determined an operating pressure of 20.32 kPa. Grasping stability was quantified by relative slip displacement in rigid–flexible coupled dynamic simulations. Among the tested support configurations within 60–110°, the 90° bracket produced the most stable slip response under vertical and horizontal disturbances. Thin-film pressure tests showed an asymmetric but stable three-finger load-sharing pattern. Field trials in a high-density dwarf spindle orchard achieved an 83.98% harvesting success rate. After 72 h of cold storage, no obvious surface browning, epidermal abrasion, or compression marks were observed during visual inspection. This assessment was limited to visible external damage and did not include quantitative evaluation of internal bruising, firmness degradation, flesh browning, or long-term storage quality. These results demonstrate stable grasping performance and low visible external damage under the tested conditions.

1. Non-destructive testing (NDT) is a core technology for ensuring the integrity of materials and structures in industrial manufacturing and equipment maintenance. Traditional methods (such as ultr...