融合CRNN与扩展卡尔曼滤波的机器人组合导航定位方法OA

To address the issue of positioning errors exceeding the limit in robot integrated navigation caused by positioning sig-nal fluctuations and occasional failures of inertial components and to enhance system reliability in complex dynamic environ-ments,this paper proposed an integrated navigation and positioning method for robots combining the convolutional recurrent neural network(CRNN)and extended Kalman filter(EKF).The CRNN model was trained using normal operational data to mine the spatiotemporal features and underlying patterns within the navigation data to realize error estimation.A dual-layer fil-tering framework based on the EKF was constructed to perform hierarchical recursive corrections on the error estimates output by the CRNN through the synergy of sub-filters and the main filter.Experimental results show that even under extreme condi-tions where the positioning system signals are completely lost,the robot navigation system using this method can still maintain high positioning accuracy.Through the synergy of deep learning and filtering techniques,the proposed fusion method of CRNN and the EKF can significantly enhance the positioning accuracy,fault tolerance,and environmental adaptability of the robot integrated navigation system,providing an effective solution for reliable navigation in complex scenarios.