Research On Walking Control And Path Planning Of A Humanoid Robot

With the improvement of living standards and more serious problem of aging population, people is eager to bring a kind of intelligent robot to their home for taking care of the olders. Since the structure and action of humanoid robots are similar to human being's, they can easily be accepted by human beings and can adapt themselves to all kinds of facilities in human society. Therefore humanoid robot is the first choice among all kinds of robots that can enter human's world. The research of humanoid robots has absorbed more and more attentions and become the focus in robotics.This dissertation discusses the important techniches about the walking system of humanoid robot. Based on the analysis of research achievement about humanoid robot in different contries we focus on the work of humanoid robot's walking control and path planning in the digital simulating circumstance. Because unstable structure of link rod is adopted in humanoid robot, the gait planning, walking control, map building and path planning are discussed on the precondition of stability as following:1. Since humanoid robot has a head, a trunk and two arms as compared with the biped robot, so generation of walking pattern is more complex. we discuss a method of real-time and stable generation of the motion of legs, arms and trunk. The motion of legs in the sagittal plane must satisfy that x-axis position of ZMP has the largest x-axis stability margin, similarly, the motion of legs in the lateral plane must satisfy that y-axis position of ZMP has the largest y-axis stability margin. The motion of arms and trunk are generated based on equilibrium of z-axis moment. The effectiveness of the proposed method is confirmed by the results of simulation.2. According to the freedom degree configuration, during single leg support and double leg support Lagrange dynamics is build. Then we design the layered control system based on the controlling demand of humanoid robot. Aimed at the problem of inertia fluctuation in dynamics we specially set a controller for adjucting inertia fluctuation in front of joint controllers. In designing the part of dynamics that cannot be expressed precisely will be transformed into disturb and perturbation. Through solving the problem by H∞state feedback,a robust controller is devised,which made the biped robot system not sensitive to disturb and keep stable with perturbation. This controller greatly reduces the errors. Throrgh analyzing the control performance of single joint, feed-forward controller of joint can be obtained based on classical control theory3. A scheme of building the map of circumstance is presented based on sensor equipment and walking trait. In order to make the humanoid robot walk freely in complicated circumstance, the reliable capabilities for obtaining plane information from its surroundings are demanded. In this work we present a system for extracting planes from data taken by stereo vision. After the depth image has been obtained, the pixels of each line are scanned and split into straight line segments. The neighbouring relation of line segments is kept in link structure. The groups of three line segments are selected as seed regions. A queue is maintained for storing seed regions, and then the plane region is expanded around the seed region. The process of region growing continues until the queue of seed regions is empty. After trimmed the edges of the planes become smooth. In the end we get extracted planes. The results of experiment testify that the speed and precision of algorithm satisfy the demands of humanoid robot's navigation. Then a method for building a grid map of the robot's surroundings is proposed. Each grid in the map is flagged with floor or obstacle and contains a value telling the height of the floor or obstacle, and the height information of floor is precise and that of obstacle is coarse. The system maintains two grids, one is 3D Occupancy grid that supplies probability support and the other is floor grid that tells the information of plane. Our approach has been verified on the platform of simulation and the result proves that the method is able to generate the map for path planning and collision avoidance.4. Considering about the characteristic of humanoid robot walking we proposed a scheme of path planning. The humanoid robot's actions are discretized into a set of given actions and configuration space is discretized into grids. The environment map is built by use of stereo vision. The shape of humanoid robot is simplified into double-cylinders model for the sake of checking obstacles. In the end a sequence of actions, which waste the least cost, is found out from the environment map as a path. The results of simulation test the validity of the method.