A Real-Time Precise Positioning Method for Indoor Robots Based on Multi-Sensor Fusion

Abstract

Achieving real-time, precise, and reliable localization for robots in indoor environments is a critical prerequisite for autonomous navigation and the execution of complex tasks. Single sensors are often limited by their inherent physical characteristics and susceptibility to environmental interference, making it difficult to maintain stable performance in dynamic scenes. To address this issue, this paper investigates a real-time precise localization method based on multi-sensor information fusion. By systematically analyzing the data characteristics of heterogeneous sensors, including LiDAR, visual sensors, inertial measurement units (IMUs), and wheel encoders, the method establishes accurate noise models and designs spatiotemporal synchronization and anomaly detection mechanisms, thereby laying a solid data foundation for fusion. An adaptive fusion localization model based on state space is constructed, and an algorithm that dynamically adjusts sensor weights according to online data quality is proposed. Furthermore, multi-scale feature fusion and a robust graph optimization framework are employed for pose estimation, enhancing the system's accuracy and fault tolerance. Finally, an embedded real-time localization system is designed, which ensures the efficient execution of the algorithm from the perspectives of hardware architecture, computational optimization, and resource scheduling. Experimental validation demonstrates that this method achieves centimeter-level localization accuracy in both static and dynamic indoor environments, while meeting the requirements for high real-time performance, strong stability, and fast convergence, thus providing a reliable pose estimation solution for indoor mobile robots.