CPG Based Motion Control Of A Quadruped Robot Walking On Rough Terrain

For a quadruped robot with multiple DOFs, the traditional model-based motion control method is disadvantageous to produce rapid and steady motion and to adapt to non-structured environment because of the complicate dynamic model. In this thesis, a biomimetic quadruped robot is used as the object to conduct rhythmic motion and environmental adaptability research. Based on the rhythmic motion and biological reflex mechanism of vertebrate, the robot completes rapid and steady locomotion and capable of adapting to ravine and step terrain. The main contents of the research are:Building a complete experiment platform, including a quadruped robot, the master-slave motion control system (includes hardware system, software program) and the human machine interface.Using sinusoidal function to build the central pattern generator model (CPG), which is worked as the trajectory planner to control the basic rhythmic gait of the quadruped robot to walking on plat terrain. Weight allocation and pose adjustment methods are combined to solve the hind legs dragging problem.Imitating the animals' reflex movements, the re-stepping reflex and postural adjustment reflex are modeled to deal with the loss balance caused by losing foothold when the quadruped robot stepping on a ravine or step. Reflexes are triggered by abrupt changes of the body pitch angle and ground touch status. Then the robot changes the movements of the four legs coordinately to rapidly adjust the robot's attitude to counteract balance loss caused by vertical inertia. Thereby the adaptability to unknown ravine and step terrain are achieved.The experimental results show that the robot can complete steady motion on plat terrain with the maximum velocity up to 0.2m/s, equivalent to 1/2 body length per second, and can successfully cross over a 1.2-leg-strid wide trench and walk down a 0.4-leg-length high step.