Research On Gait Planning And Walking Control For Humanoid Robot

Humanoid robot is one of the frontier research topics in the field of robot research, which is a combination of many disciplines, such as machinery, electronics, computer, material, sensor, control technology, artificial intelligence, and so on. It is a reflection of the intellectualization and automatization of a country and an important symbol of high-tech level of a country. Stable walking is the basic behavior of humanoid robot, and a foundation of its other behaviors. Humanoid robot has great application prospect because of its flexible walking ability. However, humanoid robot is a nonlinear system with the features of many degree of freedom, complex structure and strong coupling. So the realization of stable walking is becoming a hot and difficult issue in the field of humanoid robot research. The research of walking control can not only provide theoretical guidance for the machine design and optimization of humanoid robot, but also promote the research and application of bionics. Based on this research, people can better decode the secret of life, understand the walking characteristics of human and make good use of these characteristics to improve our lives. Therefore, the research has an important scientific significance and great application value.Gait planning and walking control are the main research contents of achieving stably walking for humanoid robot. The gait planning is giving all joints locomotion trajectories in order to achieve desired walking for humanoid robot in ideal environment, and the main goal of walking control is to correct the planning gait according to real-time walking state of humanoid robot aiming to guarantee the stable walking of robot. In this dissertation, the research on gait planning and walking control has been deeply done based on collection, collation and analysis of abundant domestic and foreign relevant papers and literatures. The main research work are as follows:(1) Modeling of kinematics and kinetics of humanoid robot and its stability criterionAfter analyzing the structure of humanoid robot and simplifying each joint of the the robot, the link model of humanoid robot, the world coordinate and the local coordinate of each joint are established. The forward and inverse kinematics models of humanoid robot are deduced based on the D-H method and the geometrical relationships of the leg joints, and at the same time its kinetics model is deduced according to Lagrange method. Moreover, the definition of walking stability criterion-zero moment point(shorted for ZMP) for humanoid robot is introduced, the expression of ZMP is deduced and constraint conditions of satifying ZMP while humanoid robot achieving stably walking is studied.(2) Gait planning based on CSI-PSO In view of the problems of humanoid robot gait planned by traditional planning method, of which the order is too high, the trajectory often oscillates, ZMP margin is not large enough, and the parameters is difficult to optimize, so a gait planning method based on CSI-PSO is proposed. This method has two steps. The first step is using cubic spline interpolation(shorted for CSI) algorithm to plan the gaits of the start walking, periodic walking and end walking in saggital and lateral planes, which containing the distance parameters between center of two hip joints(shorted for CTHJ) and the ankle of support leg at each key moment. The second step is constructing objective function, which is used to estimate the gait stability, using the distance between ZMP and the center of stable region at each moment of both single and double support phase as parameter, and then optimizing the distance parameters between CTHJ and the ankle of support leg of gait which is planned by CSI using particle swarm optimization(shorted for PSO) algorithm, in order to keep the ZMP of gait as close as possible to the center of stable region. The simulation results show that the gait planned by this method has bigger ZMP stability margin than the optimization gait based on genetic algorithm, and the gait is monotonous during each period, of which the trajectories of joints location, angle, angular velocity and angular acceleration are smooth.(3) Walking control based on all joints angles correction The factors such as the tracking errors of joints angles causes the ZMP error and affects the stability of walking. Because all joints angles errors affects ZMP, it is difficult to achieve ideal walking control performance if only adjusting one angle or two of the ankle and hip of the support leg. Further, the relationship between each joint angle and ZMP is nonlinear, so it is difficult to calculate the angle correction of each joint based on ZMP error directly. For those problems, a walking control method based on all joints angles correction is proposed. This method calculates the CoM(Center of Mass) position correction of humanoid robot tracking desired ZMP by fuzzy algorithm based on ZMP error, and then corrections of all joints angles to match the CoM position correction are computed by using equality constraint quadratic programming Lagrange method and Jacobian matrix between CoM and all joints angles of robot. The simulation results show that the desired ZMP trajectory can be tracked better and more stable walking can be achieved than the walking control method which only adjusts the ankle or hip joint.(4) Walking control using two arms swinging methodIn view of the problem of unstable walking and even falling down caused by the yaw moment generated during the humanoid robot walks, a new method of walking control based on two arms swinging is proposed. The parametric angle trajectories of arms swinging are planned by cubic spline interpolation method. Through the exhaustive search algorithm, the swinging angle parameters are traversed to guarantee the yaw moment to be mostly counteracted by the arms swinging moment, and then angles trajectories of two arms swinging is obtained. After that, besides by controlling the joints of leg in walking control of humanoid robot, it is done that controlling two arms tracking planned swinging angle trajectories to counteract the yaw moment generated during walking. The simulation result verifies that the method can preferably counteract the yaw moment of walking, which makes the yaw moment smaller than the maximal moment of friction between the ground and the support feet at each moment. Furthermore, compared with the swing angle trajectory generated by moment equivalency method, the yaw moment can be better counteracted by tracking the swing angle trajectory generated by this method, and the method can guarantee the monotonicity, smoothness and periodicity of the angle trajectories of arms swinging, minimizing the risk of sliding during the humanoid robot walking.At the end of this dissertation, the main research is summarized. It makes out the main innovations and research achievements, and also points out the problems and issues which need to further research.