Time-variant Gait Planning And Control For Under-actucated Biped Robot

Biped robot has attracted researcher’s extensive attention because of its broad application. Abundant achievements have been got in this field. Under-actuated biped robot is a kind of robot with excellent performances, because it takes full advantage of human’s walking character and possesses the property of high speed, high energy efficiency and strong adaptability. This paper aims at time-variant gait planning and walking control of an under-actuated biped robot with five links and four actuators. It mainly includes the following contents:Firstly, according to the Lagrange principle, the hybrid dynamics model, which includes the continuous model of single support phase and the discrete model of double support phase, is obtained.Secondly, time-variant gait planning of biped robot is done in this paper. Most traditional gait planning methods are time-invariant. However, time-variant gait planning, by which method time-variant trajectories can be obtained before the finish of control, is adopted in this paper. Through gait designing and some deduction, gait planning is transformed to a parameter optimization problem. In simulation, diverse reference gaits are obtained by modified genetic algorithm.Thirdly, the stability of time-variant gait control is analyzed and control strategy is designed. Usually, the Poincare map is adopted to analyze the stability of biped robot. By setting a proper linear transformation of the controlled variables, the stability of the system is greatly improved. Neural network direct inverse control is designed for the walking control. A neural network is used to simulate the inverse model of the robot, and is in cascade connection with the robot. The tracking to reference trajectories is achieved in simulation.Finally, the physical prototype and a mini-robot are introduced. About the physical prototype, its mechanical structure, hardware system and software system are briefly introduced. About the mini-robot, the mini-robot suit is introduced and a real-time control software of the mini-robot is developed.The main innovation points of this paper includes:first, time-variant reference trajectories are designed and diverse gaits are obtained by optimization; second, the stability of the system is greatly improved by a proper linear transformation of the controlled variables; third, neural network direct inverse control is designed to solve the problem of walking control.