Research On The Method For Autonomous Target Capturing Of Free-Floating Space Robot

Based on motion prediction and nonlinear optimization, a method for computing feasible robot motion for capturing a target satellite with known geometry and inertial parameter autonomously is presented. This method can be divided into 2 steps. First, the target trajectory, meaning the translational motion of its center of mass and the rotational motion about it, is observed and then predicted. Then, the robot capturing motion is planned, using the motion planning approach of zero-disturbance to base’s attitude in point-to-point manipulation. After the first step, based on the prediction, a suitable time for capturing the target can be selected. So, through the motion planning approach in point-to-point manipulation, the motion planning of the robot capturing the target can be solved, since we know the pose of the handle on the target at the time of the space robot capturing the target. The main contributions of the thesis are as follows:The algorithm for long-term general three-dimensional motion prediction of a free-floating rigid body in space is studied. Long-term prediction is useful for the case in which observation is not possible for a long time(e.g. occlusions, limited communication to ground) or in which a task which requires long-term planning is of interest(such as autonomous grasping). The algorithm is divided into three steps: observation of the target’s motion using binocular stereo camera on the space robot, the identification of a set of parameters from the observation, and prediction with the identified parameters. The identification of the parameters can be solved by optimization method, whose objective function is the difference between the observation and the pose expressed by this set of parameters analytically. Part of this set of parameters can be estimated by the observation. This estimation can help to reduce the computation cost of this nonlinear optimization problem. A simulation system for general three-dimensional motion prediction of free-floating space object is established to test this motion prediction algorithm. The statistical analysis of a great deal of numerical simulation shows the accuracy and the reliability of this method.A method which combines motion prediction of target and motion planning of robot is also presented. This method is for capturing a target satellite autonomously by a space robot. First, based on the theory of multi rigid body dynamics, the differential equations of the base’s attitude motion and of the end-effector’s motion is established. In these equations, the manipulator joint motion is the entries. Then, the manipulator joint motion is discretized with some appropriate parameters. At last, through the motion planning approach of zero-disturbance to base’s attitude in pointto-point manipulation, the optimal parameters can be solved. In the motion planning task, firstly, the motion prediction results of the target satellite are used to determine whether the robot arm will collide with the target when it moves along the planned trajectory, these planned trajectories will be eliminated; then, catching the target satellite is the primary objective, at the same time, minimizing the disturbance to the base’s attitude is also considered. At last, Adams and a program is used to simulate the space robot’s capturing the target using this method, the results shows that this method is effective.An effective method for autonomous target capturing of free-floating space robot is proposed. This study is closed to the application of space robot, and the presented method for motion prediction of free-floating object and motion planning approach have a significant potential application for the development of our space robot.