Hide/Show Apps

Feedback motion planning of unmanned underwater vehicles via random sequential composition

Ege, Emr
In this thesis, we propose a new motion planning method to robustly and computationally efficiently solve (probabilistic) coverage, path planning, and navigation problems for unmanned underwater vehicles (UUVs). Our approach is based on synthesizing two existing methodologies: sequential decomposition of dynamic behaviors and rapidly exploring random trees. The main motivation for this integrated solution is a robust feed-back based and computationally feasible motion planning and navigation algorithm that takes advantage of these two planning approaches. To illustrate the main approach and show the feasibility of the method, we first performed 2D simulations in MATLAB. We then implemented our method using a realistic fully dynamic 3D UUV simulation environment based on a platform built on the Robot Operating System (ROS)/Gazebo interface to test the overall performance and applicability for real applications. We also tested the robustness of the method under extreme environmental uncertainty (water current that is half the maximum speed of the UUV). 2D and realistic 3D simulation results indicate that our method can produce robust and computationally feasible solutions for a broad class of UUVs and Unmanned Surface Vehciles (USVs).