Image-based Robot Visual Servoing Control

Currently, the research on image-based robotic visual servoing control has become a hotspot in robot domain. This thesis made an in-depth study in the key problems involved in image-based robotic visual servoing system such as tracking and locating algorithm, Jacobin matrix based online identification and the construction and implementation of H∞robust controller, which can be specified as follows:(1)A Shifted Rayleigh based particle filter is proposed for tracking and locating. In order to solve the problems involved in the particle filter based tracking algorithm, such as particle shortage, high computational complexity, long operation cycle and inability to meat the real-time requirement of robotic visual servo system, the bearings-only target tracking-Shifted Rayleigh based filter is introduced into particle filter. Firstly, Shifted Rayleigh based filter is utilized to obtain the trajectory of the moving target, then the target is accurately located by the particle filter, which uses the color feature of the target. As a result, both the deficiency that Shifted Rayleigh Filter only considers the motion information and particle shortage, high computational complexity involved in the particle filter are overcome.(2)In particle filter based online identification of Jacobin matrix, the proposal distribution is usually the suboptimal predictive density from the system equation without regarding the latest observation, thus heavily depends on the system model. Further, if there exists uncertainty in the system model or observation noise, the proposal distribution can hardly reflect the true distribution. In order to solve this problem, a particle filter based on H∞is proposed in the thesis. In the proposed filter, the uncertainty of Jacobin matrix and external disturbance is fully considered. More specifically, the non-linear factors of the system (i.e. the unobservable factors) and dynamic features generated by the system parameters and structure is treated as the colored noise, then the Hoo robust filter capable of dealing with the colored noise is introduced into the particle filter, termed as the H∞based particle filter. The particles sampled from this filter are much close to the true position of the object through utilizing the rough estimation from the Jacobin matrix and the effective proposal distribution. As a result, the sampling efficiency of particle filter is enhanced and the computational complexity of recursive estimation is also reduced.(3)Aiming at the uncertainty of robotic visual servoing system and external disturbance, a state-space equation based H∞robust control method is proposed, in which a state space model of robotic visual servoing system is established and a H∞robust controller is designed.The simulation experiments to test the positioning ability of the visual servoing system of a2-DOF robot is implemented and the experimental results demonstrate the validity of the proposed approach.