Research On Fault Tolerance Algorithm For Motion Planning Of Redundant Manipulators

Since robots play an increasing role in the various fields,China has paid more and more attention to the study and application of robotics.In the practical application of the redundant manipulator,the working accuracy of the manipulator is one of the important performance indices.Due to the various unfavorable factors during the production,processing,installation,and work of the manipulator,the end-effector position error is inevitably generated,which greatly decreases the execution accuracy of the task.And thus,the precision and reliability of the application of the manipulator are degraded.This thesis summarizes the reasons of occurrence of the position error,and then analyzes and studies the different stages of the position error during the execution of the task and the different impacts.Sequentially,the corresponding solutions are proposed,and different position-error tolerance methods are used according to the task complexity.In order to verify the effectiveness of the above-mentioned methods,the computer simulation studies were carried out with the corresponding fault-tolerant schemes.The research content of this thesis can be divided as follows:(1)After the initial state adjustment of the manipulator,the position of the end-effector may deviate from the desired position,resulting in the initial position-error of the end-effector.Therefore,an initial position-error tolerant motion planning scheme is proposed based on the neural-dynamics design method in this thesis.The scheme is resolved on the velocity-level by the pseudo-inverse method,which can quickly and smoothly reduce the initial position error during the task execution,and thus,improve the working accuracy of the robot manipulator.(2)Considering the limitation of the pseudo-inverse method in tackling with the physical limit constraints,this thesis presents an initial position-error tolerance scheme based on the quadratic programming method.In this scheme,a quadratic performance index is designed to minimize the position error.Moreover,the physical limits of the joints are incorporated into the fault-tolerant scheme.The fault-tolerant scheme is then solved in real-time by the primal-dual neural network solver,and the initial position error of the manipulator end-effector is eliminated.(3)Due to the movement of the mobile base,the position error is continuously generated while the food ingredient preparation task is being performed.By defining the real-time position error,the neural-dynamics method is introduced into the position error elimination,and the position error occurring in the task process can thus be eliminated by using the pseudo-inverse method.(4)It is difficult for the manipulator to cut apart a large block of wood plate in the wood cutting process due to the limitation of the working space of a robot.This thesis proposes a method by mounting the robot base onto a moving rail to enlarge the working space of the robot manipulator.The method will rebuild the Cartesian velocity of the end-effector by coordinating the desired velocity of the end-effector and the moving velocity of the base.Therefore,the manipulator can achieve trajectory-tracking smoothly and accurately with the method.(5)The robotic joint velocity jump will happen at the moment of joint failure occurrence.For elimination of the joint velocity jump at the joint-failure moment,a fault-tolerant motion planning scheme is proposed without joint velocity jump in this thesis.The joint velocity jump can be eliminated by replacing the original scheme with the degradation one at the joint-failure moment,and a neural-dynamics method is introduced to eliminate the position error of the end-effector.Furthermore,for the comparison and improvement purposes,the proposed scheme is resolved by the pseudo-inverse and inverse-free methods in real time,respectively.Moreover,the fault-tolerant motion stability and working accuracy of the manipulator can thus be improved.