Study on Humanoid Gait Planning and Stability Control
|School||National University of Defense Science and Technology|
|Course||Control Science and Engineering|
|Keywords||Humanoid Robot ZMP Stability Criterion Force/Torque Sensor Gait Planning Piecewise Controllable Stability|
Humanoid robot is always one of the most active fields in intelligent robotics due to its unique walking motion pattern. The precondition for the use of the humanoid robot in practice is to improve the ability of the humanoid robot to move and adapt to the environment. This paper mainly focuses on the research into gait optimizing planning and stability control of the humanoid robot which is supposed to play pingpong ball.Firstly, the mechanical structure, control system and sensor system of the humanoid robot“KDW-5”are introduced. The kinematics model of the robot is built based on the improved Denavit-Hartenberg(D-H) coordinates representation. The dynamic model of the robot in the single support phase is built based on the Lagrange function.Then a feasible method is proposed to detect real ZMP and the phase of supporting using the force/torque sensors. Then the data are used to judge the stability of the robot based on the ZMP stability criterion.The whole walking phase of the robot is divided into three phases: the start walking phase, the cycle walking phase and the stop walking phase in this paper because that the initial states, walking parameters and final states of the robot are different during these different phases. The gaits of these three phases are planned using different planning methods. The validity of this algorithm was proved by the experiment.It is possible that the robot can recover to stable state by using its actuated joints while it is unstable. The definition of controllable stability of the robot is proposed in this paper. A method to judge the controllable stability is also proposed. Finally the method is used in the stability control of the robot which is disturbed by external force. The robot can recover to stable state by changing the step length. The validity of this algorithm was proved by the simulation.