Step Prediction And Online Gait Planning For Lower Extremity Exoskeleton

Gait planning provides the desired trajectories,so that it plays an important role in lower extremity exoskeleton.At present,two types of gait planning method are mainly used.One is based on the gait features of human to adapt to different subjects’walking patterns,the other is based on the gait theory of bipedal robot to get good balance performance.However,there is no methods to achieve both goals simultaneously.So this thesis proposes an online gait planning method by combining the universal step prediction model with the inverse kinematics.The main work is as follows:（1）We analyze the gait features of different subjects,and find that the inter-subjects similarities in task space are higher than in joint space.Then,we propose a gait planning method by transferring gait features from task space to joint space with inverse kinematics.The individual adaptability is improved by 34%compared to traditional joint space gait planning methods.（2）The remained voluntary trunk movements of patients have a bad influence on exoskeleton balance.So we propose an improved instantaneous capture point（ICP）model by adding trunk dynamics for accurate prediction of step position.The experimental results of 7 subjects show that the step prediction error of improved ICP model is only 10.4%,which is much lower than 78.9%using traditional ICP model.It means that the accuracy of the prediction has been significantly increased.（3）Finally,we propose an online gait planning method by combining the step prediction model with inverse kinematics,while we switch the gait phase by a finite state machine.Compared to the traditional ICP gait planning method,the average value and the maximum value of|D_ft|（the distance from center of mass to the interfoot line）are decreased by 0.78cm（19%）and 2.27cm（31%）respectively in simulation;the average value and the maximum value of|D_ft|are decreased by 0.35cm（10%）and 1.77cm（23%）respectively in the exoskeleton experiments of three subjects.In the walking simulation with interference,the maximum ofΔ|D_ft|in our method is lower than traditional method.These results indicate that our method can improve the balance performance of exoskeleton.