A Bio-inspired SLAM System for a Legged Robot

Abstract

Legged robots have strong adaptability to the environment, but the robots usually rely on high-precision sensors to complete simultaneous localization and mapping (SLAM) for autonomy. This paper proposes a bio-inspired SLAM system for a legged robot, using an industrial monocular camera and a single-threaded lidar instead of expensive sensors. This method fuses information from multiple sensors based on RatSLAM, a navigation system imitating the neural processes in the hippocampus of rodent. Experiment platform is based on a quadruped robot dog, demonstrating how the system works in real-world situations. This work is also compared with pure visual SLAM and pure laser SLAM algorithms, in order to exhibit the robustness and superiority of the bionic SLAM algorithms for legged robots. This system provides a reference for the use of bionic algorithms on bionic robots.