Visual Servoing Of Nonholonomic Wheeled Mobile Robots

Visual sensors imitate the visual sense of human beings and are allowed to conduct the non-contact measurement of the environment.Due to the advantages of low cost,a wealth of information provided,and high reliability,vision cameras are often used as sensors of wheeled mobile robot(WMR)systems to precisely localize the robots and detect the targets.Their application domain involves the domestic service,the intelligent transportation systems,the aerospace,etc.The vision-based visual servoing of WMRs has also become a research hotspot.Visual servoing of WMR is a technique which uses feedback information extracted from vision sensors to control the WMR to reach a specified posture or track a desired trajectory.WMRs belong to a class of strong nonlinear underactuated nonholonomic systems,and the various uncertainties introduced by vision sensors make them uncertain nonholonomic systems,and their controller design is very challenging.Moreover,limited by the mechanical structure of a WMR,the robot frame may not be consistent with its camera frame,and the extrinsic parameters need to be calibrated,which will also bring inconvenience to the practical applications.Recent years have witnessed increased developments in dealing with these problems.However,the existing methods still have some defects.Firstly,the most of existing visual servoing methods are designed by the adaptive control.Although they can guarantee the asymptotic convergence for the closed-loop errors,the control gains are sensitive to the initial states of the systems,which are difficult to be adjusted in practice.Secondly,in most of the existing methods,the time-varying control laws are adopted,and their convergence speed is very low.Thirdly,sometimes it may be difficult to strictly distinguish between the stabilization and trajectory tasks types of tasks,then a non-unified controller will bring inconvenience to practical applications.Finally,most of the existing controllers are designed at kinematic-level,which require that the designed velocity could be achieved immediately.The controllers designed at kinematic-level may be ineffective due to the existence of mass and inertia of the WMRs.To address the aforementioned problems,this dissertation delves into visual servoing of nonholonomic WMRs,the main results and contributions are summarized as follows:1.Invariant-manifold-based visual stabilization of a class of WMRs.An adaptiveinvariant-manifold-based two-stage control law is designed to solve the visual stabilization problem for a class of WMRs in the presence of unknown extrinsic parameters.The exponentially bounded convergence of the closed-loop errors can be guaranteed and the controller gains are insensitive to the initial states and the image depth.By resorting to the two-stage process,the numerical singularity problems of the invariant manifold will not occur.2.A non-adaptive switching approach for visual stabilization of a class of WMRs.For the same control problem as in 1,an invariant-manifold-based non-adaptive three-stage control law is presented,and the exponentially bounded convergence of the closed-loop errors can also be achieved.For the three-stage controller,its software implementation is more simple since there is no need to estimate the unknown extrinsic parameters and the image depth,while keeping perfect properties of the control law in 1.Moreover,the overshoot of the systems can be reduced as much as possible when the three-stage control law is applied.3.Simultaneous stabilization and tracking visual servoing of WMRs.A simultaneous stabilization and trajectory tracking controller is designed to avoid frequently switching between the stabilization and trajectory tracking tasks.Compared with the existing methods,the assumptions on the reference velocities are more intuitive and reasonable.4.Visual stabilization of a class of WMRs at the torque-level.A global time-varying controller using backstepping method is designed with adaptive control techniques to realizing the visual stabilization problems of WMRs in the presence of uncalibrated extrinsic parameters.5.Visual regulation of WMRs at the acceleration-level.An extended I&I is proposed to deal with the problem of realizing acceleration-level pseudo-dynamic visual stabilzation for a class of WMRs.To validate the proposed control laws,they are tested in simulation systems and/or experimental systems,and these results show that the proposed approaches can achieve the visual stabilization and trajectory tracking tasks.Finally,the main work and the drawbacks of the thesis,as well as the future works are summarized.