Quadruped Robot Oriented Terrain Recognition And Motion Planning Method Study

Legged robots offer a simple promise: the potential to navigate areasinaccessible to their wheeled counterparts. While wheeled vehicles may excel withregard to speed and fuel, legged locomotion is able to access virtually100%of theearth’s land surface, thus, caught wide attention of scholars at home and abroad.Research on legged robots has a long history, in terms of walking, the mainstreamtrend now focuses on developing new methods to further enhance the robotadaptability for complex, unknown environment. To make legged robot better adaptto walking challenges, this paper presents a terrain recognition and motion planningmethod for quadruped robot.Firstly, starting from obtaining the depth information from terrain, combiningresearch status at home and abroad, this paper deals with terrain depth informationusing tensor voting algorithm. Recognition results of typical terrain are also given.In order to associate recognition result with motion planning, evaluation function ofterrain is also established.Secondly, a motion planning method devised from terrain recognition result isdeveloped, including robot self-positioning method during motion, COG (center ofgravity) trajectory generation method, gait and foot target selection method and endeffector trajectory generation method. In order to strengthen the stability of thequadruped robot while crossing the terrain, some motion planning principles undercertain conditions are also proposed.Finally, a multi-terrain simulation platform is designed and based on LabVIEWrobotics environment, a multi-layer control architecture for quadruped robot isdeveloped to control the robot to cross a variety of terrain. The simulation resultsshow that the terrain recognition and motion planning method proposed in thispaper can effectively guide the quadruped robot through complex, unknown terrain.