| On October 19, 2016, China’s tiangong-2 and shenzhou-11 were successfully docked.Astronauts and space manipulators carried out the first international man-machine coordinated on-orbit maintenance scientific experiment. The man-machine coordinated on orbit maintenance has become one of the three key test tasks of tiangong-2. In the future, China’s space station construction, on-orbit maintenance and scientific experiments will also need a space manipulator to assist astronauts. The research on the path planning of the manipulator in the space environment is of great significance for space robot and astronaut to cooperate to complete the on-orbit mission.Nowadays, domestic and foreign research on space manipulator is still in the initial stage,and there are still many key problems that need to be resolved. When a space manipulator cooperates with astronauts to complete on-orbit tasks in a dynamic and uncertain space station environment, the collaborative manipulator will not only be affected by inherent constraints such as joint limit and self-collision, but also be affected by tasks constraints such as maintaining a specific posture. At the same time, it needs to maintain the anthropomorphic configuration to enhance the comfort of human-computer interaction. This puts forward higher requirements for the already complex high-dimensional motion planning of collaborative manipulator, which mainly involves two key problems: Path Planning under task constraints and motion anthropomorphization.In this research topic the 7-DOF collaborative manipulator is taken as the research object,the anthropomorphic obstacle avoidance motion planning of collaborative manipulator under task constraints and the obstacle avoidance motion planning based on the method framework in a dynamic environment are taken as the main research lines. Gradually carry out the research and exploration of the anthropomorphic obstacle avoidance motion planning of the collaborative manipulator under the dynamic environment. The specific research content is as follows:For the part of anthropomorphic obstacle avoidance motion planning of collaborative manipulators under task constraints, the path planning problem of task constraints and the motion configuration anthropomorphic problem separately according to the existing methods,which cannot realize the motion anthropomorphization of collaborative manipulators while meeting the constraints. A learning-based anthropomorphic motion planning method for a collaborative manipulator under task constraints is proposed. This method uses random sampling planning algorithm to plan the end path of collaborative manipulator which meets the inherent constraints and task constraints in the task space where the manipulator meets the constraints directly; establishes Gaussian process regression model including the upper and lower arm length to learn and map the human arm motion to the collaborative manipulator, and obtains the anthropomorphic arm configuration; adopts inverse kinematics based on arm angle parameters for the collaborative manipulator; adopts inverse kinematics based on arm angle parameters to solve the complete null space of invalid sampling points, and selects Sub-anthropomorphic arm configuration. This method (1) can simultaneously solve the task constraint motion planning and motion anthropomorphization problems of the collaborative manipulator; (2) establish the Gaussian process regression model of collaborative manipulator, and realize the random sampling planning algorithm to directly sample in the task space which meets the constraints, so as to avoid the constraint satisfaction method, and improve the efficiency of algorithm planning; (3) take the upper and lower arm length as the input of the model and adds it to the training of the model to avoid the data re-collection and training caused by the change of the collaborative manipulator model. (4) describe the self-motion manifold completely to ensure the probability completeness of the planning algorithm.For the part of obstacle avoidance motion planning in dynamic environment, based on the above part of the planning method, this paper analyzes the advantages and disadvantages of global path planning algorithm and local path planning algorithm in the manipulator tip constraint space. Aiming at the problem that the global path planning algorithm planning algorithm cannot solve the obstacle avoidance path planning problem of the manipulator in a dynamic environment, and the local planning algorithm cannot get the optimal solution and easily fall into the local-minimum, a hybrid algorithm combining the RRT* algorithm and the artificial potential field method is proposed. The algorithm (1) ignores the dynamic obstacles in the global path planning stage, uses RRT * algorithm to obtain an asymptotically optimal planning path, and takes the global path as the path guidance of the local planner; (2) The improved artificial potential field method is used in the local planning stage. The improved potential field method is improved by adding velocity, acceleration and global path guidance potential field on the basis of the traditional position potential field, so that the manipulator can make full use of the global planning to avoid obstacles in real time and walk along the asymptotically optimal path in dynamic environment. |
