Research On Visual Servo Strategy Of Wheeled Mobile Robots

The vision based mobile robot system is obtained by combining wheeled mobile robots and vision sensors.This system can take the advantages of both mobile robots and vision sensors,and it has been applied widely in forefront and high-end areas such as aerospace engineering,intelligent transportation,and home services.The visual servoing of wheeled mobile robots mainly refers to utilizing image information,which is collected by vision sensors,to make the mobile robots track a given trajectory or regulate to a desired pose.This technique can enhance the perception and adaptation ability for external environments of the agent,and it effectively improves the intelligence level and working ability of the agent.With the rapid development of various technologies involving computer vision,pattern recognition,and automatic control,the topic of visual servoing for mobile robots has been concerned by numerous researchers both at home and abroad,which makes it become one of the most hottest topic in robotics and automation field.However,the main problem of the vision system is the lack of depth information of the scene.On the other hand,the mobile robot is a typical underactuated device having nonholonomic motion constraint.Thus,the vision based mobile robot system turns to an uncertain nonholonomic system,and it brings great challenges for the high performance control design of the system.Till now,though fruitful results have been derived for visual servoing of mobile robots,many drawbacks are associated with existing methods from the practical application perspective.Firstly,the intrinsic parameters of the camera is the basis for achieving visual serving tasks.However,existing calibration methods give unsatisfactory precision results and demand complex operation.Then,the current visual servo regulation methods have low efficiency,making the mobile robot produce redundant paths.Next,because of the short baseline degeneration problem,the essential matrix is seldom applied in visual servoing,implying that it is difficult to take the advantages of the essential matrix based visual servoing method.Moreover,current visual servoing schemes can only implement trajectory tracking or regulation task separately.Finally,the existing methods,which only work well with accurate camera intrinsic parameters,can not be directly employed for uncalibrated visual servoing method.Motivated by the desire to tackle the above-mentioned problems and drawbacks,this dissertation carries out insightful studies for visual servoing of mobile robots,and provides several effective solutions specifically.This work can improve the accuracy,flexibility,and intelligence of the vision-based mobile robot system,providing technical accumulations for its practical application.In general,the main contributions of the dissertation are summarized as follows.1)A unified rotation-based self-calibration method for cameras.For the selfcalibration problem of camera intrinsic parameters,a unified calibration scheme that can be applied to both center catadioptric cameras and pin-hole cameras is proposed in this paper.Firstly,the two types of cameras are described by a unified spherical projection model.Then it is proved that when the camera frame makes pure rotation,the distance between spherical projection points for any two static feature points remains unchanged,based on which,a set of constraint equations with respect to camera intrinsic parameters is designed.An optimization function is constructed and finally solved by numerical algorithms.Simulation and experimental results are collected,showing that the proposed strategy presents such merits as good robustness regarding image noise and undesired small translation,easy implementation,and sufficiently high calibration precision.2)Selected-strategy-based visual servo regulation of wheeled mobile robots.In order to improve the practicability of the system,an efficient visual regulation scheme is provided.Firstly,based on the pose estimation by vision information,relative position and pointing angle errors are utilized for designing the visual servo regulation controller by use of the Lyapunov techniques.Secondly,for resolving the problem that this controller will generate large rotation angles when the robot at some initial positions,the negative regulation controller is obtained by changing the errors' form,then the selection strategy is designed for the above two controllers.At the same time,the active vision is combined to keeping the target in the camera's field of view.Simulation and experiment results show that the visual servo regulation controller can accomplish the regulation tasks efficiently by using the same controller parameters from different initial poses,and the selection strategy keeps the pan-unit's rotation angle in a reasonable range.3)Essential-matrix-based visual servoing of mobile robots without short baseline degeneration.To deal with the short baseline degeneration problem occurring in essential matrix based strategies,a switching control scheme is presented to achieve efficient regulation for a nonholonomic mobile robot by using several essential matrices.Firstly,three essential matrices of the system are introduced,which are then estimated up to a determinable constant scale factor with respect to their true value without any additional constraint for the feature points.Then,to deal with the short baseline degeneration problem of essential matrices and the nonholonomic motion constraint,a three-stage switching control law is designed by selecting proper essential matrices,despite the lack of metric information of the scene.The rapid convergent property of system errors is proven with mathematical analysis.The performance of the proposed strategy is validated by both simulation and experimental results.4)Model-free simultaneous tracking and regulation visual servoing of wheeled mobile robots.To avoid frequently switching between different visual servoing tasks,a unified tracking and regulation strategy is proposed.Specifically,a novel composite error vector,including both image signals and rotational angles,is defined and then calculated by comparing the feature points of the unmodeled target object from the current image,the reference image and the desired sequence of images.Here the 2-1/2-D visual servoing scheme is employed to keep targets in the camera field of view.After that,an error transformation is designed to construct the chained-form error system,for which a time-varying continuous controller is designed to accomplish both tracking and regulation tasks.As shown by Lyapunov analysis,the proposed unified controller achieves asymptotical stability even in the presence of uncertainties regarding object model and depth information.Both simulation and experimental results are collected to investigate the feasibility of the proposed approach.5)Visual servo regulation of wheeled mobile robots with an uncalibrated onboard camera.When using an uncalibrated camera for visual servoing tasks directly,this method can still regulate the mobile robot well.Specifically,a novel fundamental matrix-based algorithm is firstly proposed to rotate the robot to point toward the desired position,with the camera intrinsic parameters estimated simultaneously by employing the fundamental matrix and a projection homography matrix.Subsequently,by utilizing the obtained camera intrinsic parameters,a straight-line motion controller is developed to drive the robot to the desired position,with the orientation of the robot always facing the target position.Another pure rotation controller is finally adopted to correct the orientation error.The exponentially convergent properties of the visual servo errors are proven with mathematical analysis.The performance of the proposed uncalibrated visual servo regulation method is further validated by both simulation and experimental results.