| Intelligent robot visual servoing is the fusion of results from many elemental areas including high-speed image processing, robot kinematics, robot dynamics, control theory, computer technology and real-time computation, and is a main subject in the research field of robot visual servoing system.One major classification of robot visual servoing system distinguishes position-based control form image-based control based on the error signal is defined in 3D coordinates or directly in terms of image features. In position-based control, features are extracted from image and used in conjunction with a geometric model of the target and the know camera model to estimate the pose of the target with respect to the camera. The primary disadvantage of position-based control is that it is often highly calibration dependent. The impact of calibration dependency often depends on the situation. In image-based servoing, control values are computed on the basis of image features (F ) directly. The image-based approach may reduce computational delay, eliminate the necessity for image interpretation and eliminate errors due to sensor modeling and camera calibration. However, there are two main disadvantages to image-based visual serving: 1) It is necessary to relate differential changes in the image feature parameters to differential changes in position of the manipulator. So it is necessary to compute inverse feature Jacobian (J). However, the feature Jacobian is the function of the distance from the camera focal center to the point being imaged. For an eye-in-hand system, determining the distance is very difficult. Many reports are based on using a constant image Jacobian, which is computationally efficient, but valid only over a small region of the task space. 2) The disadvantage is presence ofsingularities in the feature mapping function, which reflect themselves as unstable points in the inverse Jacobian control law.In this thesis, on the basis of giving a tutorial introduction to this topic, we present new visual mapping models which combine the robot work space with the image feature space tightly and new visual tracking controllers based on artificial neural network. Manv computer simulations and experiments of the object tracking are carried out in eye-in-hand visual servoing experiment system. The main work is summarized as follows:1. The direct kinematics and inverse kinematics of MOTOMAN-SV3 robot are studied. A new method for inverse kinematics equation of MOTOMAN-SV3 robot is presented. This method needn't change the coordinates of the robot manipulator end effector. A lot of multiplications of inverse matrices are avoided in the process of deduction.2. A novel method of reaching law variable structure control based on fuzzy rules is presented, which is that the reaching law parameters are on-line adjusted by fuzzy rules. This method is used in a digital ac position servo system based on TMS320F240 DSP. The experimental results show that the system designed by this method has both satisfactory quality and very smaller chattering.3. A new image correlation matching method based on the modified genetic algorithm and a new image matching method with the modified moment invariants based on the modified genetic algorithm are presented. Many experiments are carried out. The experimental results show that these methods are very effective for image matching processing.4. Based on the approach of dynamically estimating the image Jacobian, the visual positioning method for MOTOMAN-SV3 robot is studied. Computer simulation is carried out.5. A nonlinear visual mapping model of the 2D eye-in-hand robotic visual tracking problem is proposed and a new visual tracking controller based on artificial neural network with BP-GA mixed algorithms is designed. Simulation results show that thismethod can drive the static tracking error to zero quickly. Moreover, a new nonlinear visual mapping model of the 3DOF eye-in-hand robotic visual tracking problem is studied and a new visual tracking...
|
PDF Full Text Request