Manipulator Perception And Control For Slender Peg-Hole Assembly

Manipulator autonomous assembly is of significant importance in industrial production.Although it has been applied in several industries by traditional force control,it starts to reveal limitations with increasing demand of flexibility,especially in a type of slender peg-hole assembly tasks.Concerning with the characteristics that slender peg-hole assembly possess,this study focuses on two prominent issues in both peg-hole alignment and peg-hole insertion.By introducing non-contact fixed camera system,this study proposes new sensing and controlling methods for slender peg-hole assembly.In the peg-hole alignment process,the slender beam is likely to deform which changes the tool parameters.The time-invariant plastic deformation and time-variant elastic deformation caused by gravity pose a huge difficulty in control of tool tip.The method this study proposes takes advantage of depth-independent visual servoing.Image features as two points and two lines from the projection of the tool tip are selected and modeled.The deformation of tool tip is also modeled and the differential kinematics is deducted which builds up the relationship of joint motion and feature motion on image plane.With the methodology of depth-independent interaction matrix,an adaptive controller is designed and stability is proved.This slender deformbale peg-hole alignment method has use for neither parameters of material and length of the peg nor knownledge of direction of gravity.As long as the image features converge to desired ones,the alignment task is accomplished.In the peg-hole insertion process,this study proposes a hybrid controller of force and vision to simultaneously control the compliance and insertion depth under the circumstance of uncertainty in environment and tool parameters.The six-axis force/torque sensor is installed on the end-effector of the manipulator and the camera is fixed on the top.The information from contact and motion are both adopted for online estimation of local constraint and the direction of hole is estimated recursively.To keep safety,force control gets higher priority and the force/vision subtasks are divided based on this idea.Controllers in both subtasks are designed.This method reduces dependencies on the model of constraint environment and tool parameters.To verify the feasibility of the proposed methods,simulations and experiments are carried out.One simulation is to examine the stability of adaptive visual servoing controller for peg-hole alignment on Matlab.Another is to prove the effectiveness of online estimation algorithm of constraint environment.An experiment on KUKA LBR iiwa 7 is conducted and the performance of proposed methods are tested.