| As a miniaturized marine robot,the unmanned surface vehicle(USV)can replace humans to complete repetitive and dangerous operation tasks,and are playing an increasingly important role in both civil and military fields.However,the global positioning system(GPS)is usually deployed to measure position states for motion control,which will limit applications of the control algorithm.In this context,the USV motion control problem in GPS-denied environments has become one of hot topics focused by scholars.In this dissertation,the monocular camera is employed as a perception and measurement unit,so as to improve the environment perception and motion autonomy by endowing the USV with visual functions.Focusing on the motion control problem caused by the monocular vision,the monocular visual servo stabilization control,tracking control and swarm control of the USV are investigated to propose corresponding visual servo control methods.Main research works are as follows:Aiming at problems in the monocular visual servo stabilization control,the extreme learning and finite-time extended state observer(FESO)based stabilization control methods are respectively proposed for the USV.To retrieve pose errors conveniently,the homography decomposition technique is used to establish the monocular visual servo stabilization control system of the USV.Firstly,for the problem of unknown image depth and model dynamics,the single-hidden layer feedforward neural network with double-channel learning mechanism is employed to approximate lumped unknown terms,and input and output layers of the network are directly connected via hyperbolic tangent functions,so as to improve approximation rate and accuracy.As a result,the extreme learning-based stabilization control method is proposed for the USV,such that stabilization errors can tend to zero with arbitrary accuracy.Then,the velocity measurement noises are further considered,thereby the FESO is designed to simultaneously reconstruct USV velocities and total disturbances for avoiding control input chattering induced by measurement noises,and homogeneity-satisfied nonlinear feedback to directly accommodate unknown image depth for decreasing computational complexity.As a consequence,the FESO-based stabilization control method is proposed for the USV,rendering stabilization errors practical finite-time stable.Finally,the effectiveness of proposed methods in the USV stabilization control is validated by simulation and comparison results.Focusing on problems in the monocular visual servo tracking control,the barrier functionbased target tracking control method and the position observer-based trajectory tracking control method are proposed for the pan-tilt camera and USV,respectively.To maintain camera visibility,the active vision system is employed for tracking the visual target.Firstly,for the problem of unknown image depth,singularity of Jacobian matrix and restricted view field in the active visual servo target tracking control,the one-dimensional target tracking error is designed to avoid control singularity by the pseudo-inverse technology,and the barrier function with the finite escape property to make target tracking errors satisfy view field constraints.Together with the parameter adaptive method to estimate image depth-related parameter online,the barrier function-based adaptive pseudo-inverse control method is proposed for the pan-tilt camera to make target tracking errors asymptotically stable,and thereby continuously providing visual information for the monocular visual servo trajectory tracking control of the USV.Then,to further deal with unknown image depth and model dynamics in the trajectory tracking control,the finite-time position observer is devised by briskly calibrating the camera extrinsic parameter online,thereby achieving the finite-time monocular vision localization of the USV.Combining with the FESO and non-singular terminal sliding mode control,the position observer-based trajectory tracking control method is proposed for the USV,making trajectory tracking errors finite-time stable.Finally,the effectiveness of proposed methods in the target and trajectory tracking control is verified by simulation and comparison results.For problems in the monocular visual servo swarm control,the prescribed performance function-based formation control method and the barrier Lyapunov function-based selforganizing swarm control method are respectively proposed for the USV.To improve the motion autonomy,formation and self-organizing swarm control tasks are respectively designed according to the line of sight distance and angle.Firstly,to handle the problem of visibility maintenance and safe collision avoidance in the formation control,exponentially decaying performance functions are exploited to prescribe feasible evolution regions for formation errors,thereby transforming foregoing problem into the prescribed performance control issue.Together with backstepping control,neural networks and parameter adaptive method,the prescribed performance function-based formation control method is proposed for the USV,such that formation errors can satisfy prescribed performance and are uniformly ultimately bounded in spite of complex unknown dynamics.Then,to further solve the multiple objective control problem in the self-organizing swarm control,both control and barrier Lyapunov functions are deployed to describe global and local control tasks,guaranteeing real-time operational performance of the control algorithm by automatically weighing priority of each objective.Combining with sliding-mode control,neural networks and parameter adaptive method,the barrier Lyapunov function-based self-organizing swarm control method is proposed for the USV,so as to make the swarm control system asymptotically stable in the presence of complex unknown dynamics,and thereby achieving visibility maintenance,swarm aggregation,swarm separation and velocity consensus.Finally,the effectiveness of proposed methods in the formation and self-organizing swarm control is validated by simulation and comparison results. |
PDF Full Text Request