Design Of Manipulator’s Visual Servo System Based On Davinci Platform

Robot intelligence is the core field of current robotic development processand robot visual servo technology is an important direction in this process. Relyingon visual sensors, robot visual servo technology adds visual information into thecontrol system so as to control the robot to perform high-precision servo motionand achieves the effective integration of robotics and visual information. Thispaper carries on the development of visual applications based on TMS320DM6446DaVinci platform and builds the visual servo control system of PUMA560manipulator and finally achieves the servo motion of the manipulator.In the motion control section, the mathematical model of PUMA560manipulator is established and the forward kinematics and inverse kinematics ofthe established model are verified by the application of robot toolbox. In this paper,photoelectric encoder is adopted for the feedback information of position andvelocity. Double-loop control systems of manipulator’s four joints are also built insimulink environment and the system optimization and multi-joint movement arecompleted.In the visual processing part, DaVinci platform is chosen for the video andimage processing. Firstly, the hardware and software environment of DaVinciplatform are built, which means that the ARM and DSP dual-core communicationand the communication of dual embedded systems are achieved. Secondly, manypre-processing operations are performed on the captured images such as gray,binary and alpha median filtering. On this basis, the feature extraction and centerpositioning of moving target are finished by using the LoG edge detection operator.Finally，the construction of the visual servo system is finished by introducingposition coordinates into the established control system as the overall controlcommand of the entire visual servo system.At last, a visual controller of BP neural network model based on conjugategradient algorithm is designed. The rationality of the system is tested and theoverall performance of the system is analyzed by conducting simulations andSemi-physical experiments of overall system in the MATLAB environment.