Rice leaf disease severely affects crop yield and food security. Traditional deep learning methods often require large-scale labeled data and lack generalization in complex environments. To address these limitations, we propose RiceViM, an efficient framework for rice leaf disease classification that integrates an Efficient Multi-Scale Attention (EMA) module with the Vision Mamba state space model. EMA enhances the extraction of local and global features, while Vision Mamba captures long-range dependencies with low computational cost. Experimental results on a benchmark dataset show that RiceViM achieves a classification accuracy of 95.87%, outperforming baseline models in both accuracy and efficiency. More notably, the model only requires 9.98M parameters and 8.96ms of single-graph inference time, achieving a balance between high precision and low power consumption. The proposed method demonstrates strong potential for practical applications in intelligent agricultural systems.

Recently, a motor vector control system has been a hot spot for robot control. Whether it is a UAV or a household robot, the accuracy of motor vector control is becoming increasingly important. The traditional motor magnetic field vector control system (FOC) has some limitations. In the small brushless motor control system, the non-encoder type position sensor has a specific nonlinearity. At the same time, in the FOC control system, the PI controller in the FOC is also difficult to solve the nonlinear problem of the control system. Artificial intelligence is developing rapidly, and we can solve nonlinear problems through neural networks. Therefore, this problem presents an intelligent motor magnetic field vector control system (IFOC) based on MLP. Firstly, we design a new position sensor structure. Then, after obtaining the data of the linear position sensor, we use MLP to solve the nonlinear problem of the sensor. With the precise position sensor, we can transform from the static coordinate system to the rotating coordinate system (The dp coordinate system) through the Clarke-Park transformation and then introduce MLP to replace the original PI controller. Finally, we built an experimental platform to verify the feasibility of the IFOC control system.