Key points are not available for this paper at this time.
Quadruped robots, an important class of unmanned aerial vehicles, have broad potential for applications in education, service, industry, military, and other fields. Their independent positioning plays a key role for completing assigned tasks in a complex environment. However, positioning based on global navigation satellite systems (GNSS) may result in GNSS jamming and quadruped robots not operating properly in environments sheltered by buildings. In this paper, a tightly coupled LiDAR vision inertial odometer (LVIO) is proposed to address the positioning inaccuracy of quadruped robots, which have poor mileage information obtained though legs and feet structures only. With this optimization method, the point cloud data obtained by 3D LiDAR, the image feature information obtained by binocular vision, and the IMU inertial data are combined to improve the precise indoor and outdoor positioning of a quadruped robot. This method reduces the errors caused by the uniform motion model in laser odometer as well as the image blur caused by rapid movements of the robot, which can lead to error-matching in a dynamic scene; at the same time, it alleviates the impact of drift on inertial measurements. Finally, the quadruped robot in the laboratory is used to build a physical platform for verification. The experimental results show that the designed LVIO effectively realizes the positioning of four groups of robots with high precision and strong robustness, both indoors or outdoors, which verify the feasibility and effectiveness of the proposed method.
Gao et al. (Mon,) studied this question.