Research On A Soft Robotic Tongue With Pneumatic Actuation

Soft robotic tongue plays an important role of verisimilitude improvement in pronunciation teaching device and humanoid robot.However,due to the insufficient understanding of the role played by robotic tongue and the existing technological difficulties,there has been no product to meet such kind of demand,which causes the lack of convenience to the deaf during pronunciation training as well as verisimilitude of humanoid robot when speaking.For the purpose of solving the problem mentioned above,theoretical and experimental research of soft robotic tongue is carried out based on practical demand.A pneumatically actuated soft robotic tongue with multi-layer and array-chamber structure is proposed.In regard to the flat structure of the robotic tongue,the interlayer separates the upper and lower deformable tongue surfaces and restrains the deformation of the interlayer itself,which can realize the roll-up and roll-down deformation of the robotic tongue compliantly.By using the pneumatically actuated array-chamber structure,not only the controllable composite deformations can be realized,but also the uncontrollable inflation of the soft tongue surface with single-chamber structure can be alleviated.Five basic types of deformation are realized including roll-up,roll-down,elongation,groove and twist through the composite control of the robotic tongue.In order to realize compliant actuation of the robotic tongue,a pneumatic actuation system is designed based on PC control.The deformation pattern and essential parameters of the novel pneumatically actuated soft robotic tongue are analyzed by means of FEM simulation,which lays the foundation of the fabrication and function realization of the soft robotic tongue.In order to control the deformation of the soft tongue surface caused by the composite deformation units with array-chambers,the mechanical model of the deformation of the soft robotic tongue base is proposed.Through the mechanical model,the relationship between the pressure in chambers and deformation is obtained when the robotic tongue is pressurized and deformed.Specifically,that is the relationship between pressure and bending angel in roll-up and roll-down deformation,and the relationship between pressure and elongation in elongating deformation.On the basis of the mechanical model,the establishment of the geometrical shape of the tongue surface as well as the control of the positional parameters is realized using surface interpolation and construction through the control of the position of key geometrical points.A control model for the soft robotic tongue based on the mechanical model,surface interpolation method and double-layer feedforward ANN is studied and realized,which can be used to control the deformation of the robotic tongue.Furthermore,the working process of the robotic tongue when demonstrating pronunciation is simulated by means of FEM simulation and 3D plotting of parametric surface in software MATLAB.By means of casting in room temperature with 3D printed mold,the prototype of the robotic tongue is fabricated.The pneumatic system and control system are integrated with solenoid proportional valves and DAQ cards forming a whole system used for actuating and controlling the motion of the robotic tongue.Based on 3D laser scan and OptiTrack motion tracking system,the experimental platform applied in the characteristic and functional verification experiments for the soft robotic tongue is set up,and through which a series of related experiments are carried out.The results of characteristic experiment show that the robotic tongue is able to realize five basic types of deformation including roll-up,roll-down,elongation,groove and twist.The experimental results of tongue shape test reveal 3 disciplines.First of all,both of the prediction modules used for acquiring the input air pressure according to the target tongue shape are able to predict the input pressure of the robotic tongue precisely.For the module based on the deformation model,the relative error between the theoretical predicted data and experimental data is only 9.5%(for rolling deformation)and 1.6%(for elongation deformation)when the deformation of the robotic tongue reaches the steady state.For the module based on the double-layer feedforward ANN trained with experimental data,the prediction curve is consistent with the actual experimental curve and the difference can be ignored.Secondly,the largest deformation is over 20 mm and the shortest response time is only 0.12 s when the robotic tongue is demonstrating pronunciation,which reaches or is close to the related technological demand.Finally,the fitted tongue surface using interpolation matches well with the real tongue surface measured in the test.Through the experiments of repeatability accuracy,the robotic tongue is verified to be with the ability of tongue shape retention in short term and high deformation repeatability in long term.During the long-term test,the minimum deformation fluctuation is only around 0.1 mm which can be ignored compared with the deformation itself.