Research On Constrained Visual Servoing For Robotic Manipulators Based On Predictive Control

Visual servoing control technology has been widely used in field of the robotics,which makes the robotic systems more flexible and faster.The visual servoing methods use the image features which are regarded as feedback signals to design the control law,and the control signals are obtained to control the motion of the robotic manipulator,then the robotic manipulator can be drived to reach to the desired pose by using the visual information.The advantage of the feedback control is that it is robust to errors which exist in the system.Many traditional visual servoing control methods do not take into account system constraints.Usually,visual servoing systems are in the face of the visibility constraints and actuator saturation.It is well known that one of the advantages of the predictive control is its ability to explicitly deal with constraints.Therefore,it is significant to study the model predictive control(MPC)for constrained image-based visual servoing(IBVS).Then,this paper further research the model predictive control methods for constrained image-based visual servoing.In this paper,the controlled plant is the monocular eye-in-hand and eye-to-hand camera configuration.Firstly,a kinematic IBVS prediction model based on the depth-independent Jacobian matrix is established.Then,the constrained visual servoing predictive control method with terminal matrix is designed,which can be used to cope with the system's input and output constraints.The terminal constraints and the terminal cost functions are added to the constrained optimization problem of the MPC-based IBVS method,and the stability of the IBVS system is analyzed.Most of the existing MPC-based IBVS methods do not consider the uncalibrated problem in the IBVS system.However,the camera calibration is a tedious and error-prone process.Moreover,for the monocular vision system,it is also difficult to obtain the depth information.Therefore,in this paper,the adaptive model predictive control method for IBVS is proposed,which can be used to simultaneously deal with system constraints,unknown camera intrinsic and extrinsic parameters,and unknown depth information.Under the depth-independent Jacobian matrix framework,a novel parameter estimation method is proposed.The model predictive control method is combined with the parameter estimation algorithm.At each sampling time step,the parameter estimation algorithm uses the latest system's inputs and outputs to update the model parameters,while the MPC controller adopts the latest updated model parameters to calculate the optimal control inputs.The simulation results demonstrate that when the MPC method is used to control the system with unknown parameters,the parameter estimation method is an effective way to cope with the unknown model parameters.The comparison simulation results verify the validity of the proposed control method in constrained and uncalibrated environments.The control precision of the visual servoing system is not only influenced by the model uncertainty,but also influenced by the external disturbances.Therefore,in order to improve the disturbance rejection ability of the constrained visual servoing system,the adaptive disturbance observer based adaptive model predictive control method for IBVS is proposed,which can be used to simultaneously cope with system constraints,the model uncertainty and disturbances.The proposed method is composed of two parts which are the feedforward compensation part based on the adaptive disturbance observer and the feedback regulation part based on the adaptive model predictive control.Moreover,the adaptive disturbance observer is proposed.The traditional disturbance observer is designed by using the nominal plant model which is fixed.Different from the traditional disturbance observer,the proposed adaptive disturbance observer is designed by using the estimated plant model.The depth-independent image Jacobian matrix is employed to construct the estimation of the plant model,which can be used to facilitate the updating of the estimated plant model.The iterative identification algorithm is incorporated in the adaptive model predictive controller,which can be used to provide the model parameters to both the adaptive disturbance observer and the adaptive model predictive controller,and to minimize the model uncertainty and the influence of the disturbance observer which affects the plant dynamics.The simulation results demonstrate that the satisfactory control performance can be achieved under the control of the proposed method.Most of the visual servoing schemes either do not consider the constraints or consider the constraints at the kinematic level.Moreover,many MPC-based IBVS methods take into account the constraints,but they do not simultaneously consider the nonlinear robot dynamics and the model uncertainty.Hence,the sliding mode observer based adaptive model predictive dynamic control method for IBVS is proposed,which can simultaneously deal with system constraints,the model uncertainty and the nonlinear robot dynamics under the condition that the joint velocities are unavailable.The proposed control method can be applied to both the eye-in-hand and eye-to-hand camera configurations.