Grasp planning with high-performance and efficiency in unstructured environments is a critical problem that intelligent robots need to solve urgently to complete picking tasks. To solve this problem, a quantized grasp quality generative neural network is proposed to generate pixel-level grasps. A light-weight convolutional neural network is first used to generate initial grasp configurations. Aiming at the degradation of grasping performance caused by the light-weight design, a decoupled grasp quality network is designed to generate a pixel-wise grasp quality. To improve the robustness and generalization of the model, an adaptive filtering method is proposed to filter the grasp configurations. Then, an ellipse fitting-based grasp pose optimization method is proposed to obtain the final grasp configuration. On the public Cornell dataset, the state-of-the-art grasp accuracy of 98.9% and the efficiency of 15 ms are achieved, with approximately 564× fewer parameters compared to comparable performance. In the grasping experiments on the robotic platform, grasp success rates of 99% and 93.4% can be achieved in single object and cluttered scenes, respectively. Finally, the model is deployed on an embedded artificial intelligence (AI) computing device to verify its practicality.

In order to reduce the torque ripple for switched reluctance motor (SRM), the learning error preprocessing‐based torque–flux linkage recurrent neural network adaptive inversion control (TFRNNAIC) for SRM torque is proposed by the filter preprocessing and the non‐linear mechanism characteristics for SRM. In TFRNNAIC with the advantages of parallel and series–parallel structure, the torque feedback error learning method is employed to update the weights of the torque–flux linkage recurrent neural network. To suppress the ripple effectively in torque error used for the weight learning of the torque–flux linkage recurrent neural network and obtain an accurate TFRNNAIC, namely the torque–flux linkage model for SRM, the low‐pass filter preprocessing for the torque error is used. Moreover, the other low‐pass filter is executed to reduce the ripple in the output for the PD torque control. The superposition of outputs for TFRNNAIC and PD torque control is taken as the reference flux linkage. Compared with other control strategies, such as the classical parallel neural network control, the classical series–parallel neural network control and TFRNNAIC, the simulation results show that the learning error preprocessing‐based TFRNNAIC for SRM torque is capable of effectively reducing the torque ripple for SRM with the good recurrent performance.

High torque ripple dramatically affects the switched reluctance motor (SRM) application. To reduce the torque ripple, a reference torque neural network (RTNN) is proposed to adjust the reference torque online. Firstly, the RTNN is built on the torque sharing function (TSF) method. Furthermore, the RTNN is designed as a single-input and -output network. As the periodic relationship between the torque ripple and the rotor angle, the rotor angle constitutes the central node parameter of the implicit function in RTNN. Therefore, one-step adjustment of the RTNN can perform well at restraining reference torque. Lastly, the torque error is used to adjust the parameters of RTNN to reduce the torque ripple. In the MATLAB environment, through the simulation comparison with fuzzy torque and PD current compensation method, the effectiveness of RTNN at torque ripple suppression is proven with different loads and speeds.