Gait Synthesis For Biped Robot Based On CPG

Gait synthesis for the biped robot has always been an difficult but important issue in the robotics field. Many methods have been developed based on different theories, like dynamic model method, passive dynamics, virtual model control,walking data reference trajectory, etc. But these gaits can't adapt the enviromnet well, or they can't even withstand the slight disturbance. The discovery of the CPG (Central Pattern Generator) theory brings new vision to the gait synthesis for the biped robot, because of CPG's rhythmicity and its capacity of resisting disturbance. The aim of this research is to apply CPG to gait synthesis for the biped robot, to gain some experience on CPG application through the research. The main work of the dissertation are as follows:First, description of the links in robotics and the Denavit-Hartenberg method are introduced. Despite of its disadvantage of error accumulation, D-H method can easily meet most accuracy requirment, and it is easy to caculate and easy to understand. A 7-link biped robot model is built, and its kinematic model is built with D-H method.Secondly, the formation of the CPG conception, the process of its application on robot control and the development of its mathmatical model are introduced. The detailed mathmatical model of the Mastuoka oscillator is given as well as the meaning of its parameters, and a CPG network is designed to control the angles of the ankles.Thirdly, the trajectories of the robot's ankles and hips are planned with spline. In order to use CPG in ankles control, the constraints of the hip's height and the ankle's swing angle are canceled. The angle curves of the knees and hips are gained through reverse kinematics, using the robot's 4 trajectories of its ankles and hips. We use 3 key gestures in the robot's walking process and the outputs of the CPG to produce the"leading curve", and realize the continuity of the gait by adding the outputs of the CPG to the"leading curves"to control the angles of the ankles.Finally, we achieve the angle curves of the joints, and the stick figure of the walking movement in Matlab. The kniematic simulation results confirm the proposed method. The gait gained by complete trajectory planning and the gait with CPG control are compared and analyzed, the analyzation of the comparision has directive function for further CPG application in robot control.