Autonomous Legged Robot Navigation with Environment Awareness System in Complex Outdoor Environments

Abstract

In order to solve the navigation of legged robot in outdoors, this paper constructs the environment awareness system. Firstly, the Graph Simultaneous Localization and Mapping (Graph-SLAM) is used to establish the 3D environment modeling and the hybrid filtering algorithm is proposed to reduce the high frequency noise of the model. Secondly, the 5 Dimension (5D) Spatial Descriptor is designed to analyze the terrain traversability region of the legged robot. After that, Variant Step Size Rapidly Exploring Random Trees (VSSRRT) is introduced for path planning. Finally, the experiment is carried out with the Yobogo legged robot which verifies that the system can satisfy the navigation requirements of the legged robot.