Online Path Planning for Autonomous Underwater Vehicles with Large Inertia in Uncertain Environments
Abstract
This project aims to develop a robust and efficient algorithm for online planning of large underwater vehicles that incorporates uncertainties and variability in the environment. Many naval systems (e.g., submarines, ships) are designed to sustain the oceanic environments while carrying necessary equipment and payloads which require large and heavy vehicles. These vehicles possess a large amount of inertia as they travel through the underwater environment making the problem of path planning challenging due to their restricted motion behavior. Also, there are many uncertainties in the underwater environment (e.g., spatio-temporally varying ocean currents, hidden unknown obstacles) that must be incorporated into the path planning problem to ensure safety of the vehicle. While an ample amount of techniques have addressed the path planning problem in literature, very few algorithms focus on planning for large underwater vehicles with inertia operating in uncertain environments and in presence of ocean currents. In this regard, this project aims to develop a robust and efficient algorithm for online planning of large underwater vehicles that incorporates uncertainties and variability in the environment. To achieve this, the research will solve a sequence of correlated tasks. First, the problem of path planning for large vehicles will be examined in known oceanic environments where the obstacle locations are a priori known to the planner. The space will be partitioned into a tiling structure, divided into reachable and unreachable areas, and then a configuration space will be established by encoding augmented vehicle kinematics information (e.g., course angle and velocity) as states for each tile. A potential-field based method is then proposed to significantly reduce the configuration space by deleting states that are not feasible or safe for the AUV. This will reduce the computational complexity of the path planning algorithm. Further, a multi-resolution based incremental real-time planning algorithm will be developed that efficiently guides the vehicle to its destination. Next, we extend this approach by incorporating spatio-temporally varying current maps as additional uncertainty. The ocean current maps could be generated using measurement systems (e.g., satellites, costal radars) or models (e.g., Regional ocean Modeling System (ROMS)). Since the momentum of the vehicle is drastically effected by these currents, the research will integrate the current velocity with the vehicle dynamics into the tiling states. Then, the potential-field based method will be used to reduce the computational complexity and delete states that are hazardous to the vehicle. The multi-resolution planning algorithm will then be extended to include the effects of ocean currents. Finally, the assumption of known environment will be relaxed. As the vehicle moves, obstacles will be detected using an on-board sensing system. This requires the vehicle to re-plan its path in situ to ensure its safety and the efficiency. However, re-computing the full configuration space dynamically is computationally complex. Therefore, only selected localized regions of the configuration space in front of the vehicle will be updated and searched for planning. The multiresolution planning algorithm will be extremely useful for real-time re-planning. The proposed approaches will be validated on a high-fidelity simulation platform (e.g., Player/Stage, Gazebo, UWSim/ROS) and system level simulations will be conducted to validate the efficiency, reliability, and real-time execution of the proposed methods.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 23, 2016
- Source ID
- N000141613032
Entities
People
- Shalabh Gupta
Organizations
- Office of Naval Research
- United States Navy
- University of Connecticut