Research On Motion Mapping Of Humanoid Manipulator Based On Dynamic Gesture And Motion Coordination

As the end effector of humanoid robot,humanoid manipulator has the characteristics of multi degree of freedom and multi finger coordination.With higher performance,the corresponding control method is also more complex.In the control study of humanoid manipulator,the method of collecting human hand information is often used to control humanoid As the end effector of humanoid robot,humanoid manipulator has the characteristics of multi degree of freedom and multi finger coordination.With higher performance,the corresponding control method is also more complex.In the control study of humanoid manipulator,the method of collecting human hand information is often used to control humanoid robot.In this kind of research,the existing hardware equipment for collecting human hand information has certain limitations.Visual information collection depends on the environment and is disturbed by factors such as illumination,which is limited in practical operation.Limited by the sensor entity,the data glove can not meet the dexterity and the sufficiency of collected information at the same time.To solve this problem,this thesis studies the improvement of the control effect of lightweight data gloves from the perspective of motion mapping,and studies a motion mapping based on dynamic gesture and motion coordination.The main work of this thesis is as follows :Firstly,the motion coordination element is studied based on principal component analysis.The principal component analysis method is used to extract the features of human hand motion samples,and the motion coordination element is constructed by extracting the features: the spatial coordinate information of 33 kinds of human hand bone points is collected by using the depth information sensor,and the spatial coordinate information is transformed into joint angle information as the research sample of human hand motion law.For the motion samples,the principal component is extracted by taking the finger as the unit,and the motion information is projected into the plane formed by the principal component features.A fitting curve representing the motion law of the finger is found in the plane,and the motion coordination element is constructed according to multiple fitting curves.Experiments verify the effectiveness of the motion coordination element based on principal component analysis,which provides theoretical support for the follow-up research of motion mapping.Secondly,based on the improved DTW algorithm,the real-time recognition of dynamic gestures is realized.In order to judge the category of dynamic gesture in hand motion,the DTW algorithm used for dynamic gesture recognition is improved :Based on the traditional DTW algorithm,the calculation mode and some constraints are improved＿Keogh’s pruning strategy was optimized.Based on the improved algorithm,a method for real-time recognition of dynamic gestures is designed.The experimental results show that based on LB＿Keogh’s pruning strategy is better than the traditional lower bound pruning in DTW path optimization calculation.The recognition method based on the improved algorithm can realize the real-time recognition of dynamic gestures.Thirdly,a motion mapping method based on dynamic gesture and motion coordination is proposed.The gestures with similar motion laws are clustered,the collected information is distinguished according to the clustering category through dynamic gesture recognition,and the corresponding motion coordination is used to solve the human hand information to form the control signal.The specific research work includes: minimizing the Eros distance through the Hungarian algorithm,and clustering the grasping actions with similar motion characteristics based on the hierarchical clustering of MEros distance.Combined with the research results of the previous chapters,the motor coordination element and gesture recognition corresponding to motion classification are studied.Combining dynamic gesture recognition and motion coordination for motion mapping.The experimental results show that gesture division can promote the performance of motion coordination elements.The clustering of gestures with similar motion laws can be realized based on MEros distance.The effectiveness of motion mapping based on dynamic gesture and motion coordination is verified.This thesis studies on improving the control effect of lightweight data gloves.A motion mapping method combining dynamic gesture and hand motion coordination is proposed.This motion mapping can supplement the control information according to the law of human hand motion,so as to achieve the purpose of controlling the humanoid manipulator with a small amount of information,which has a certain significance for the control of the humanoid manipulator.