On Constrained Optimal Control Of Visual Servo Systems And Its Experients

The ability of the robot percept to environments is quite limited, so we need to add various external sensors to improve the intelligent level and autonomous ability. In all kinds of sensors, vision sensors are one of the most important sensors of robot, because they include large amount of information. The study and application of combining vision system with a robotic system now becomes a hot research area in intelligent robot.The main purpose of our work is to quickly grab object in a2D plane. The research content includes detection and extraction of target feature information, camera calibration, designing visual controller and the optimization of visual servo controller. The main work and achievements are as follows:1. In a complex environment, the RGB images will change dramatically because of the light, so it is not suitable to the visual servo system. In view of this, it proposes the HSI model segmentation, and uses FFT filtering algorithm to eliminate noise, retains the feature information of the target which we needs. The experimental results show the useful of this method.2. The design method of visual servo stabilized controllers is presented for eye-to-hand robot visual servo system. Moreover, based on the properties of control Lyapunov functions, the results on global stabilization of the control law are obtained. This method proposed here does not require to estimating inverse image Jacobin’s matrix. Finally, simulation demonstrates the effectiveness of the method.3. The actuators and workspace of visual servo system is limited, in other words, visual servo control is in essence a restricted (or constraint) control problem. In order to satisfy the constraint conditions of the visual servo system, the Fibonacci method is introduced to online compute the optimal value of the adjustable parameters, which yields an optimized control of the constrained visual servo system. Finally, simulation demonstrates the effectiveness of the method proposed here.4. It sets up four degrees of freedom robot visual servo system experiment platform, including the vision subsystem, the software environment and the robot system. The method we proposed is better than inverse Jacobian matrix control method through comparing.