In the development of unmanned agricultural machinery, efficiently and accurately detecting field obstacles is crucial for ensuring both operational safety and efficiency. However, due to the complex agricultural environments, existing obstacle detection methods still suffer from low detection accuracy and large model parameters. To address these issues, this study presents CFD-YOLOv8, an application-oriented adaptation of YOLOv8 tailored to complex farmland environments that integrates complementary architectural modules and a practical IoU-based loss to improve detection performance. First, we design a Task-Aligned Dynamic Detection Head to improve the model’s adaptability to challenging environments while achieving lightweight optimization. Second, we incorporate RFCAConv into the C2f module to expand the receptive field and strengthen the model’s focus on crucial target regions. Finally, we introduce the Powerful-IoU loss function to optimize bounding box handling, thereby accelerating convergence and enhancing localization accuracy. Experiments conducted on our custom-built field obstacle dataset demonstrated that CFD-YOLOv8 improves average detection precision by 1.9%, with precision and recall rates increasing by 3% and 0.2%, respectively, while reducing model parameters by 18.9%. These results significantly outperform current mainstream obstacle detection methods. The findings of this study offer robust technical support for autonomous obstacle avoidance and path planning in unmanned agricultural machinery operating in complex environments, laying a foundation for the further advancement of agricultural mechanization and intelligence.

The rapid development of the global shipping industry and the changes in complex marine environments have put forward higher requirements for ship formation and route optimization. The purpose of the research is to improve the efficiency and accuracy of ship formation and route planning through improved algorithms. Based on this, a ship formation model combining improved particle swarm optimization algorithm and a generative route optimization method based on improved Douglas-Peucker algorithm are proposed. The particle swarm algorithm introduces dynamic adaptive parameter adjustment and the cross mutation strategy of genetic algorithm, while the Douglas-Peucker algorithm integrates density-based noise application spatial clustering algorithm to improve model performance. The test results show that the total navigation distance of the allocation path generated by the ship formation model is 605.3 meters, the calculation time is 31.8 seconds, and all ships can be accurately allocated to the target point. When the number of iterations is 1000, the route optimization model has a route coverage rate of 95.9% on the training set, an average error of 30.5 meters, and a computation time of 45.9 seconds, achieving zero collisions. The experimental results show that the improved algorithm outperforms traditional methods in accuracy and stability of formation target allocation and route planning, especially under complex sea conditions, and can significantly reduce computation time and errors. The research provides a new technological means for optimizing ship formations and routes, which has certain application potential and practical value.

The current motion interaction model has the problems of insufficient motion fidelity and lack of self-adaptation to complex environments. To address this problem, this study proposed to construct a human motion control model based on the muscle force model and stage particle swarm, and based on this, this study utilized the deep deterministic gradient strategy algorithm to construct a motion interaction control model based on the muscle force model and the deep reinforcement strategy. Empirical analysis of the human motion control model proposed in this study revealed that the joint trajectory correlation and muscle activity correlation of the model were higher than those of other comparative models, and its joint trajectory correlation was up to 0.90, and its muscle activity correlation was up to 0.84. In addition, this study validated the effectiveness of the motion interaction control model using the depth reinforcement strategy and found that in the mixed-obstacle environment, the model's desired results were obtained by training 1.1 × 10³ times, and the walking distance was 423 m, which was better than other models. In summary, the proposed motor interaction control model using the muscle force model and deep reinforcement strategy has higher motion fidelity and can realize autonomous decision making and adaptive control in the face of complex environments. It can provide a theoretical reference for improving the effect of motion control and realizing intelligent motion interaction.

The health of crops is of great consequence to the security of the global food supply and the sustainability of agricultural economies. To address the shortcomings of traditional monitoring methods, which are characterized by high labor intensity, low efficiency, and insufficient timeliness, this paper proposes an innovative intelligent analysis system. The system uses remote sensing data from drones and convolutional neural network technology to achieve efficient crop classification and accurate identification of pests and diseases. Specifically, the system adopts a multi-scale attention convolutional network to optimize crop classification, improves the cyclic consistent adversarial network for remote sensing image translation to enhance the dataset, and improves the lightweight MobileNet V2 for disease and pest recognition. The proposed method achieved an average F1 and intersection to union ratio of 94.67% and 89.14% for crop classification and recognition, respectively. When translating crop remote sensing images, the Frechette distance and kernel distance were 98.73 and 3.08, respectively. The translated image enhanced the dataset and improved recognition accuracy and convergence speed. When identifying crop pests and diseases, the accuracy and recall were 97.14% and 97.18%, respectively. The parameter count was reduced to 2.01MB, indicating superiority. This study solves the uncertainty in remote sensing data and the complexity and limited number of training samples for convolutional neural networks. This provides technical support for the transformation of agriculture towards intelligence and sustainability.

This study investigates the factors influencing the permeability and water storage capabilities of Recycled Graded Crushed Stone Layer (RGCSL) materials, which are crucial for constructing sustainable "sponge cities". The research focuses on how aggregate characteristics, such as particle size and filling sequence, affect the porosity structure of RGCSL and, consequently, its permeability and water storage performance. The findings reveal significant impacts of these factors on material performance, leading to the development of a performance prediction model based on the principle of superposition and backpropagation neural networks. The model's efficacy was validated through simulation experiments, indicating that the water storage capacity of recycled sand is significantly higher than that of coarse aggregates, with the model achieving an accuracy of 89.1%. This study is vital for advancing environmental restoration and sustainable urban development.

In this study, a thin-film artificial intelligence transistor device based on the integration of memory and computing was designed and assessed. For this device, skyrmions were used to construct an integrated memory and computing architecture that combined content-addressable memory functions and logic operational functions. First, a dual memory cell under vertical current control was designed; based on this, a current-driven skyrmion track memory was constructed. Then, the logic gate components of the skyrmion track were constructed as the operation module. Subsequently, simulates were conducted on the device, and the results showed that when Db = 0 mJ/m2 and 3 mJ/m2, Wb = 0 nm, and the position of the temporal lateral displacement curve was highest. When Wb = 18 nm, the position of the temporal lateral displacement curve was lowest. In the logic gate performance simulate, the moving speed of the skyrmion reflected a positive proportional relationship with driving-current density. Finally, in the skyrmion moving simulate, the upper part of the system stably displayed NOR operational functions, and the lower end stably displayed NAND operational functions. This research lays the foundation for designing a fifth-generation AI processor in which storage and computing are integrated.

Artificial intelligence has made people’s demands for computer computing efficiency increasingly high. The traditional hardware circuit simulation method for neural morphology computation has problems of unstable performance and excessive power consumption. This research will use non-volatile flash memory cells that are easy to read and write to build a convolutional neural network structure to improve the performance of neural morphological computing. In the experiment, floating-gate transistors were used to simulate neural network synapses to design core cross-array circuits. A voltage subtractor, voltage follower and ReLU activation function are designed based on a differential amplifier. An Iris dataset was introduced in this experiment to conduct simulation experiments on the research circuit. The IMC circuit designed for this experiment has high performance, with an accuracy rate of 96.2% and a recall rate of 60.2%. The overall current power consumption of the hardware circuit is small, and the current power consumption of the subtractor circuit and ReLU circuit does not exceed 100 µA, while the power consumption of the negative feedback circuit is about 440 mA. The accuracy of analog circuits under the IMC architecture is above 93%, the energy consumption is only about 360 nJ, and the recognition rate is about 12 μs. Compared with the classic von Neumann architecture, it reduces the circuit recognition rate and power consumption while meeting accuracy requirements.

In the development of unmanned agricultural machinery, efficiently and accurately detecting field obstacles is crucial for ensuring both operational safety and efficiency. However, due to the complex agricultural environments, existing obstacle detection methods still suffer from low detection accuracy and large model parameters. To address these issues, this study presents CFD-YOLOv8, an application-oriented adaptation of YOLOv8 tailored to complex farmland environments that integrates complementary architectural modules and a practical IoU-based loss to improve detection performance. First, we design a Task-Aligned Dynamic Detection Head to improve the model’s adaptability to challenging environments while achieving lightweight optimization. Second, we incorporate RFCAConv into the C2f module to expand the receptive field and strengthen the model’s focus on crucial target regions. Finally, we introduce the Powerful-IoU loss function to optimize bounding box handling, thereby accelerating convergence and enhancing localization accuracy. Experiments conducted on our custom-built field obstacle dataset demonstrated that CFD-YOLOv8 improves average detection precision by 1.9%, with precision and recall rates increasing by 3% and 0.2%, respectively, while reducing model parameters by 18.9%. These results significantly outperform current mainstream obstacle detection methods. The findings of this study offer robust technical support for autonomous obstacle avoidance and path planning in unmanned agricultural machinery operating in complex environments, laying a foundation for the further advancement of agricultural mechanization and intelligence.

The rapid development of the global shipping industry and the changes in complex marine environments have put forward higher requirements for ship formation and route optimization. The purpose of the research is to improve the efficiency and accuracy of ship formation and route planning through improved algorithms. Based on this, a ship formation model combining improved particle swarm optimization algorithm and a generative route optimization method based on improved Douglas-Peucker algorithm are proposed. The particle swarm algorithm introduces dynamic adaptive parameter adjustment and the cross mutation strategy of genetic algorithm, while the Douglas-Peucker algorithm integrates density-based noise application spatial clustering algorithm to improve model performance. The test results show that the total navigation distance of the allocation path generated by the ship formation model is 605.3 meters, the calculation time is 31.8 seconds, and all ships can be accurately allocated to the target point. When the number of iterations is 1000, the route optimization model has a route coverage rate of 95.9% on the training set, an average error of 30.5 meters, and a computation time of 45.9 seconds, achieving zero collisions. The experimental results show that the improved algorithm outperforms traditional methods in accuracy and stability of formation target allocation and route planning, especially under complex sea conditions, and can significantly reduce computation time and errors. The research provides a new technological means for optimizing ship formations and routes, which has certain application potential and practical value.