Tightly-coupled lidar-inertial localization with visual landmarks in operational scenarios

Purpose Simultaneous localization and mapping (SLAM) is widely used in autonomous robotics. Although LiDAR and vision-based methods have been widely deployed in large-scale environments, their submeter accuracy is insufficient for operational scenarios. This paper aims to present a tightly integrated SLAM framework combining point clouds, visual landmarks and IMU data to enhance localization accuracy. Design/methodology/approach A novel method to resolve planar marker pose ambiguity is proposed. Edge constraints between the point cloud and marker correct marker’ pose and distortions. A tightly coupled coarse-to-fine optimization approach integrates point cloud features and marker corners during tracking and localization. Pose constraints are added to the objective function to improve accuracy, and bundle adjustment is applied to key frame and marker poses in the common view to prevent registration errors. Findings The proposed method is evaluated in real-world environments. Experimental results demonstrate that it achieves superior localization and mapping accuracy compared to existing methods, thus validating the effectiveness of the framework. Originality/value This paper introduces a tightly integrated framework that combines point clouds, visual landmarks and IMU measurements to improve small-scale localization accuracy in operational scenario.