Font Size: a A A

Research On Coordinated Tracking Control For Multi-Robot System Based On Binocular Visual

Posted on:2016-02-01Degree:DoctorType:Dissertation
Country:ChinaCandidate:N ShaoFull Text:PDF
GTID:1108330479450985Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
With the expanding of application range of robot technology, and the increasing of task complexity, multi-robot systems has become a very active research subject in the academia, and it has important practical value in many fields. Compared with single robot, multi-robot systems has its special control requirements: Firistly, each follower needs to plan its own path by using the status informations of other robots, so as to realize the consistency of exercise behavior and avoid collision with each other; Secondly, multi-robot systems is a complex nonlinear system, and consider the influences of system modeling imprecision, parameter perturbation, external disturbance, which put forward higher control requirement on the design of system controller; Furthermore, each follower’s own position information and movement velocity are often applied in the controller, but each follower is equipped with two sensors are bound to increase the cost of control system, and sensor is easily affected by the influence of environmental factors.We take the coordinated tracking control of multi-robot systems as the research topic, and adopt binocular vision system as the sensing device to provide status information for the followers, and then conduct in-depth research on the distributed containment control of multi-robot systems with dynamic and static leaders respectively, as well as the dynamic following control problem of multi-robot systems. The main contents of this paper are as follows:A binocular vision model is proposed based on depth information sensitivity for the establishment problem of the image Jacobian matrix, which make the image Jacobian matrix can be calculated by using the image plane coordinate values of target points, and then avoid the estimate of depth information; combine the binocular vision model and the directed graph in graph theory together, and according to the task requirements of containing control and formation control, then complete the design of guide system based on the image informations of neighbor robots, and then plan out virtual target points as the actual tracking objects for followers to guide their movements.Aming at the distributed containment control problem for multi-robot systems with static leaders, considering the characteristic that the final velocity of each follower is zero, 1) dissipative Hamilton model of followers is built, and nonlinear disturbance observers are designed to observe the total uncertainties in the system model dynamically, then the outputs of designed observers are introduced into the designed dissipative Hamilton controllers for compensation, which enhance the robust stability and tracking control precision of system. 2) The full state observers are designed, and the observations of joint angles of followers are introduced into the designed feedback dissipative Hamilton controllers for compensation, which realizes position-sensorless control and reduces the equipment cost of control system, and the followers realize effectively tracking control for the virtual target points.Aming at the distributed containment control problem for multi-robot systems with dynamic leaders, 1) considering there exists uncertain parameters in the system models of followers, the designed sliding mode adaptive controllers can drive the followers to track the virtual target points, which realize adaptive estimations for the uncertain parameters. 2) Considering there exists both uncertain parameters and external disturbances in the system models of followers, radial basis function neural networks are adopted to approximate the total uncertainties, and the approximations and the robust inhibition items of approximation errors are introduced into the designed sliding mode controllers for compensation, which improve the control precision of system effectively.Aming at the dynamic following control problem for multi-robot systems, 1) in order to improve the global convergence speed of system, tracking controllers are designed based on two-power reaching law and nonsingular terminal sliding mode, and radial basis function neural networks are used to approximate the external disturbance uncertainties of the system. 2) To improve the robust stability and control precision of the system in the overall process, regard the complex items and disturbance items as total external disturbances, for which state observers are built to conduct dynamic observation, and the observations and the adaptive estimations of upper bound of observation errors are introduced into the designed global sliding mode backstepping controllers for compensation, which weakens the chattering, and make the followers realize tracking control for particular formation.
Keywords/Search Tags:multi-robot systems, binocular vision servo control, distributed containment control, dynamic following, Hamilton theory, observer, sliding mode adaptive control, radial basis function neural network
PDF Full Text Request
Related items