地理空间信息2025，Vol.23Issue(12)：29-32,64,5.DOI:10.3969/j.issn.1672-4623.2025.12.006
基于神经网络及卡尔曼滤波的机器人组合导航定位分析
Robot Integrated Navigation and Positioning Analysis Based on Neural Network and Kalman Filter
摘要
Abstract
To address the issue of significant positioning errors in the integrated navigation system of robots caused by unstable global position-ing system signals and inertial navigation system(INS)failures during operation,we proposed an robot integrated navigation and positioning analysis method based on neural networks and Kalman filtering.Taking the navigation and positioning data under normal conditions as training samples,the convolutional recurrent neural network(CRNN)model learns the characteristic relationships within the positioning data to complete the training process.When GPS signals are unstable or the INS malfunctions,the CRNN model extracts feature maps from the robot's integrat-ed navigation data through its convolutional layers and processes the contextual feature sequences of navigation data through its recurrent layers to capture temporal dimension features,which are then sent to the feed forward layer for identifying errors in integrated navigation positioning.An extended Kalman filtering(EKF)is employed in a dual-layer processing format,with a primary and a secondary filter.Initially,the sub-filter model is used to preliminarily correct low-level positioning errors under the condition of integrated navigation positioning errors.Subsequently,the main filter model performs a deep correction of high-level positioning errors,achieving error correction for the robot's integrated navigation positioning.Research result indicates that after applying the proposed method to the robot's integrated navigation system,the positioning accu-racy of system is significantly improved even when GPS signals are lost,with high matching degree and strong interactive performance,can com-plete the navigation and positioning of running trajectory,and enhance the adaptability and stability in complex and changeable environment.关键词
神经网络/卡尔曼滤波/机器人/组合导航/定位误差识别/误差校正Key words
neural network/Kalman filtering/robot/integrated navigation/positioning error identification/error correction分类
天文与地球科学引用本文复制引用
张辉,胡宗强..基于神经网络及卡尔曼滤波的机器人组合导航定位分析[J].地理空间信息,2025,23(12):29-32,64,5.基金项目
浙江省自然科学基金资助项目(LTGS23D010003). （LTGS23D010003）
