Fast Motion Planning of UAVs
Kaixi Cathay Wang · DigitalCommons - CalPoly (California State Polytechnic University) · 2017
Fast motion planning (FMP) of autonomous vehicles has been advanced rapidly for robotics research, particularly for trajectory planning of spacecraft. The FMP team at JPL and Caltech has developed an algorithm for autonomous vehicles in environments with many fixed obstacles. The spherical expansion and sequential convex programming (SE-SCP) algorithm is computationally efficient and guarantees any-time local optimality for a given function on top of being faster than other sampling-based motion planning methods. Spherical Expansion (SE) is randomized sampling to explore the workspace of the autonomous vehicle and it finds an initial cost-minimized path. Sequential convex programming (SCP) then optimizes this path and computes a locally optimal trajectory. Current development and simulation of the SE-SCP algorithm is still being tested with MATLAB software as well as the collaborative robotics software called the Robot Operating System (ROS). ROS has advantages over MATLAB since it is a flexible framework for writing robotics software and includes a collection of tools, libraries, and conventions specifically for robotics improvement. By developing a SE-SCP simulation in ROS, a ROS package can be created and uploaded online, which provides opportunities for the public to easily utilize the software and apply the SE-SCP algorithm for motion planning to their own autonomous vehicles.