The Design Of Motion Capture System And The Study Of Gait Pattern Based On Motion Captured Data

Humanoid robot uses humanoid motion mode as discrete supporting legs and touching the ground alternately. With high flexibility, humanoid robot can be well integrated into people's lives and provide us with services. The existing humanoid robots cannot well realize the stability and flexibility due to various problems of gait planning. Human gait features good stability, flexibility and efficiency. Therefore, the collection and analysis of the law of motion of each joint during human walking are significant for gait planning of humanoid robot.Firstly, in this thesis, focusing on the data collection of each human joint in the course of walking, a set of wearable human motion capture system is designed to collect the motion data of human waist and lower limb joints according to the characteristics of the skeleton of human joint. Then, the waist's influence on the process of human walking is discussed by collecting the motion data of human walking under free state and bound state. Thus, a late-model robot model is built—eight-link model. Finally, by similarity comparison with the database of human walking of CMU, the accuracy of the capture system in data collection of human motion is testified. In addition, in the purpose of testifying the availability of the data, the eight-bar linkage model is built in Motion Builder for simulation. The main research achievements obtained in this thesis are as follows:1) Wearable human motion capture systemIn this thesis, a set of wearable human motion capture system is designed according to skeleton structure of human joint. The most important structural characteristic is using back-hip difference value method to collect the motion data of waist, which provides data support for testifying the function of waist in the course of human walking. The designed wearable human motion capture system uses STM32 as the main control chip. Firstly, the motion data measured by heading and attitude system, which is installed on the back and hip of human body, is collected by common synchronization/asynchronization transceiver of STM32; then, analog-digital converter of STM32 is used to collect the voltage signals from the potentiometers installed on each joint of lower limbs and the signals are converted to angle data of each joint; finally, by Bluetooth module, which is connected with synchronization/asynchronization transceiver, STM32 sends the collected data to PC and the collected data will be displayed on the computer in a real-time manner.2) The establishment of humanoid robot eight-link modelFirstly, during the practical collection experiment, whether the waist of the subject is bound or not is taken as the single variable to set contrast experiment. By collecting the walking data respectively when the subject is under two kinds of state and analyzing the difference between the two data sets, the importance of waist for human walking. Then, improving the classical seven-link model of humanoid robot by adding the free degree of waist joint to its body. Thus, the link of the trunk is divided into upper and lower links. Thus, the new-style humanoid robot model is constructed—eight-link model. The last D-H identification is used to build the forward kinematics model of humanoid robot and geometrical analysis method is adopted to establish inverse kinematics model of humanoid robot.3) The collection and analysis of the human motion dataFirstly, the following things are completed: taking a walking period from walking data of different subjects and CMU human walking capture data respectively; using cubic spline interpolation method to unify the sampling frame number of each walking period and substituting the data into the similar function and compare the similarity of the walking data collected by capture system and the data in CMU database; testifying the validity of data collection by capture system. Then, we need to build ASF skeleton model in Motion Builder according to the established eight-link model of humanoid model and converting the gait pattern data of human to AMC document and importing it into the skeleton model to make the model move along with the data. Thus, the availability of eight-link model and gait pattern can be proved.