Research On Precision Locating Control Algorithm For Assembling Robot Based On Embedded System

The position of the object is often uncertain at robot assembly line and the position and pose of the end-effector can not be previsously fixed. Robot only works based on its position and pose in the absolute coordinate system, which requires a higher absolute precision of pose of robot. To meet the demand of robot pinpoint, research must be carried out on inverse kinematics of robot and it's controlling problem and compensation for location error. The traditional model of inverse kinematics equation and compensation model for location error are nonlinear, they are very difficult to build. For strong nonlinear mapping ability, neural network is researched and used widely. The related pinpoint model of robot is created and researched based on neural network. The main results of research are as follows:(1) The inverse kinematics of assembling robot is derived based on the analysis of inverse kinematics problems, and the kinematics equation is solved using algebraic method. Infuence of kinematic parameter errors on precision of pose of the end-effector is analyzed. The results show that the movement variable errors have great influence on the precision of pose, and the pinpoint model of the robot based on inverse kinematics should have a good solution precision of joint angles.(2) It's difficult for traditional methods, such as algebraic method, geometry method and iteration method, to solve inverse kinematics of robot, so inverse kinematics model is designed based on improved algebraic algorithm neural network. The network is trained and simulated twice based on samples got from algebraic method in the MATLAB. Test results show that the improved algebraic algorithm can satisfy the requirement of inverse kinematics. In addition, influence of kinematic parameter errors on the precision of pose of the end-effector is analyzed. The results show that the model can meet the precision of pose.(3) Many factors affected the precision of pose are nonlinear and it's difficult for the parameter identification method to find the ideal mathematical model. The error compensation model based on improved BP algorithm is proposed, which relies on the improved algebraic algorithm model for inverse kinematics. Simulation results show that the effect is obvious and the model is feasible.(4) Realization of models based on improved algebraic algorithm neural network and improved BP algorithm neural network is researched in 8-bit embedded microcontrollers. Modular programs and simulation experiments are performed. After optimizations and debugs, the last results for program are 4141B occupied within the ROM space and a single run time of 0.273s. Error between the simulation and the expectation is small. The pinpoint model of robot based on neural network can be better realized in 8-bit embedded microcontrollers.