Research On Autonomous Control Of Soft Robotic Hand Based On Tactile Perception

The soft robotic hand(SRH)is a new type of mechanical device made of flexible material.As an important execution component of soft robot interacting with target object,SRH is of great significance for the improvement of working level and intelligent level of robot.Although the existing SRHs can complete certain grasping actions,most of them lack or have weak tactile perception ability,which makes SRH's autonomy poor.Based on this,from the aspects of SRH's structural design,fabrication,pneumatic control,tactile perception,and autonomous control ability,this paper designs an SRH based on tactile perception,and proposes three kinds of pneumatic control schemes and realizes multi-finger collaborative control of the SRH,and uses Machine Learning to fuse the light/pressure bimodal information for realizing the SRH's tactile perception,and designs a multi-positioned tactile sensor system and designs a perceptual-control algorithm based on the autonomous selection strategy to realize the autonomous control of the SRH.The main research work and results of this paper are as follows:Firstly,completing the structural design and production of the tactile perception SRH.According to the structure of human hand and the principle of joint motion,the structural design and model of the SRH are proposed.And a fiber optic sensor,based on optical principle,is fabricated as a tactile sensor of the SRH.And then,the physical production of the tactile perception SRH is completed and the prototype is assembled.Secondly,the multi-finger collaborative control of the SRH is realized.Three kinds of pneumatic control schemes are proposed,including 2/2-way proportional valve scheme,3/2-way PWM-solenoid valve scheme and 2/2-way PWM-solenoid valve scheme,and the performance of these schemes is explored and compared by a designed experiment.And the multi-finger collaborative control of the SRH is achieved by constructing an SRH hardware platform and selecting a suitable pneumatic control scheme.Thirdly,the tactile perception of the SRH is realized by using Machine Learning algorithm with light/pressure bimodal information.The KNN,SVM,and Logistic Regression(LR)algorithm is used to realize the static and continuous motion recognition of twelve SRH's gestures.And the K-means clustering algorithm is used to successfully identify the size,shape and weight of objects.Finally,the autonomous control of the SRH is realized based on tactile perception.By performing a silicone coating on the capacitor/resistance pressure sensors and attaching them to the surface of the SRH,a multi-positioned tactile sensor system is prepared.The perceptualcontrol algorithm based on the autonomous selection strategy is designed and implemented.And then,according to the algorithm and triggered state,the autonomous control of the SRH is realized.This paper deeply studies the structural design,manufacture,pneumatic control,tactile perception and autonomous control of the SRH,and proposes pneumatic control methods,a bimodal tactile sensing method based on machine learning,and an autonomous control method based on autonomous selection strategy.In a word,the work of this paper has important theoretical significance and engineering value.