Bi-RRT* Based Trajectory Optimization and Obstacle Avoidance for a Serial Manipulator

Abstract

Nowadays, due to the rapid development and application of robotics and artificial intelligence technology, the collision-free operation of robotic manipulators has attracted a great deal of interest in the industrial manufacturing field. In complex manufacturing situations catering to one specific task, one critical issue is to develop an obstacle-free trajectory for a manipulator. In this research, we adopt one novel trajectory planning method based on the Bi-RRT algorithm to achieve a smooth and fast planning process for a serial manipulator. We first build a distance estimation and collision detection model utilizing the point cloud data. Different from the previous research of utilizing the mesh information of dynamic objects, only the position information will be used, of which is simple to develop. Moreover, GPU and parallel programming will also be applied to accelerate the procession of distance calculation. Second, a collision-free global path will be generated by using a random sampling method, namely RRT-connect algorithm, which will largely improve the convergence speed of finding a feasible path from start to finish while also avoiding obstacles. Third, a smooth trajectory will be generated via quintic polynomial interpolation. This method guarantees that the trajectory's acceleration will remain constant during the whole planning process. Finally, we conduct a numerical test on a UR5 manipulator, with an efficient solution being provided to the problem of collision-free trajectory generating for a manipulator.