Generation Of Biped Robot’s Walking Pattern And Its Experiment Research Based On Human Motion Capturing

With the increase of human-robot interaction environment, the need of personification for humanoid robots is becoming higher and higher. This is not only reflected in the proximity of the human in appearance, but also reflected in the similarity of robots’ and human’s gaits. To make the robot walk with the natural gait which is similar to human’s under the premise of ensuring the stability of walking. On the one hand, human is the imitated object of robots, the analysis of its motion law can provide reference for the study of the walking stability of robots. On the other hand, to generate walking pattern which is similar to human’s can improve the proximity between robots and human, which will be applied in the entertainment industry better. In this paper, the 3D motion capture technique is used to obtain the human’s walking patterns. We study the generation of the robots walking pattern from the perspective of "bionics".We first compute the real joint center of the human with the least square principle by planning the specific motion. After capturing human motion using the PhaseSpace three-dimensional capture system, we adopt the Lagrange interpolation to solve the problem of data loss. We apply the kinematic matching to realize the transformation of the walking pattern when considering the difference between their kinematic parameters. In the transformation of the walking pattern, the geometric scaling principle is used to reconstruct the trajectory of the hip joint and the ankle joint. Then we compute the joint angles which meet the need of the actual robot limb length and proportion based on the inverse kinematics analysis of the robot. According to the principle that the center of mass is fine tuned in the horizontal direction, we adopt the geometric relation of pitch angles to deal with the problem of beyond the limitation of the joint angles. To make the robot walk the whole sole ouches the ground, we compare the safety margin of the angles relative to the limitation and use proportional distribution principle to redistribute the compensation angle. To meet the requirements of walking stability, we use sine function to smooth joint angles. Finally we get the walking pattern. To evaluate the walking similarity between the robot and the human, we choose the similarity of the parameters in Cartesian space and joint space as the evaluation indexes. As for parameters, in Cartesian space, we choose the stride length, the step height, the walking period, single/double support phase ratio and in joint space, we choose the joint angle and the angular velocity.On the basis of getting the robot walking pattern adopting the above kinematic matching method, we study the online switching walking patterns to realize the follow walking between the robot and human. We put forward the method that first build the robot walking pattern library on the basis of successful walking patterns. Then we retrieve a walking pattern with highest similarity from the library after extracting the human gait parameters on line. For the stability of the walking pattern joint, we put forward the concept of “the joint angle distance” as the evaluation index to choose proper joint. To realize the smooth transition of the two patterns, the three spine interpolation is used, and finally we achieve the goal of the robot switching walking patterns with the change of the walking speed of human.Finally we establish the virtual prototype system of the robot in ADAMS and make the walking simulation based on the capturing human data and the walking pattern switching simulation. Later the robot walking experiment is carried out. The simulation and experimental results show that the method of generation of biped robot’s walking pattern based on human motion capturing can realize stable walking for robots, and improve the similarity of robots’ and human’s gaits.