Programming By Demonstration With Virtual Fixture Assistance For Robotic Machining

Complex workpieces represented by blades are widely used in aerospace,transportation,energy and other important fields.Robot has the advantages of intelligence,large operating space and good flexibility,so it has become a research hotspot to use robot to process complex workpieces.However,for multi-variety,small-batch and non-structured complex workpieces such as Aeroengine Blades and gas turbine blades,it is difficult to program off-line and the preparation takes a long time.Robot kinesthetic teaching combined with virtual fixture can realize the migration of manual operation trajectory to the robot,and overcome the disadvatanges of low efficiency of off-line programming and poor teaching accuracy.However,there are some problems in virtual fixture assisted robot kinesthetic teaching,such as the lack of operating comfort,the difficulty of constructing complex surface virtual fixture,the difficulty of self-adaptive adjustment according to the intention of the operator,and the low accuracy of assistant teaching.To this end,this paper proposes a virtual fixture generation and iterative updating assisted robot kinesthetic teaching technology scheme for complex workpieces,focusing on robot kinesthetic teaching system construction,complex surface virtual fixture construction,virtual fixture control structure design and virtual fixture local iterative correction.The main research work and innovative achievements are as follows:1.Construction of robot kinesthetic teaching system.Based on the gravity compensation of the robot's end effector,the admittance control is used to realize the robot's free traction in the task space.According to the operator's operating habits in the kinesthetic teaching process,a variable admittance control strategy based on standard force is proposed to make the robot and human interaction more compliant.The experiment was carried out on the UR5 robot experimental platform.The results show that the proposed control strategy can maintain the operating force near the standard force in a certain range,and enhance the operating comfort.The average force error and torque error of the measured value and compensation value of the force sensor are 0.024 N and 0.0023N/m,respectively,which verifies that the validity of gravity compensation algorithm.2.Virtual fixture construction based on surface manifold.A method of constructing6D(6-DOF)virtual fixture based on surface manifold and pose spline interpolation is proposed.The teaching data is generated by the local smooth manifold learning algorithm,and the discrete points of the guiding path of the virtual fixture are obtained by the generalized projection characteristics of the surface manifold.The positions of the discrete points are obtained by the 6D virtual fixture obtained by the Akima spline interpolation and the squad spline interpolation algorithm respectively.On the UR5 robot experimental platform,the proposed method is verified.The results show that the manifold of complex surface can be constructed only by several drag teaching,and the average error rate is1.8% after cross validation.The proposed method has generalization characteristics while ensuring accuracy,which makes the construction of virtual fixture of complex surface simple and has generalization characteristics3.Virtual fixture control structure design.Based on the reference direction method,a virtual fixture control structure based on robot contour error compensation is proposed.In this control structure,the definition of robot contour error in 6D task space is proposed for the first time,i.e.teaching precision.On this basis,a local iterative contour error estimation method is proposed.The results show that the definition of teaching accuracy is simple and intuitive.Compared with the traditional control method,the teaching accuracy is improved by 7.5 times.The contour error estimation method can accurately and efficiently obtain the nearest point on the path between the end of the robot and the virtual fixture.The error deviation of the position and the orientation contour is not more than0.004932 ?m and1.05 ?rad,respectively.4.Local iterative correction of virtual fixture.By analyzing and comparing the common representation methods of operator intention,a virtual fixture iterative correction control structure based on operator intention is proposed.Under the action of operator intention,error compensation is carried out for the virtual fixture conforming to the target surface to improve the teaching accuracy.On the contrary,local iterative correction is carried out for the virtual fixture through simple constraint object deformation algorithm(SCODEF),Finally,the teaching path is generated synchronously.The experimental results show that the proposed virtual fixture control structure can not only ensure the high teaching accuracy,but also adaptively modify the virtual fixture guidance path according to the operator's intention to fit the target surface better.Which the average contour error of position and direction is 0.1404 mm and 0.0041 rad,respectively.