An Application of Graph of Convex Sets Trajectory Optimization to the Marine Robotics Domain
Name
largaespada-rlargaes-meng-eecs-2023-thesis.pdf
Description
Thesis PDF
Size
25.57 MB
Format
Adobe PDF
Checksum (MD5)
a324fe6d8d4f85e4cf62857dc4dc8e97
Author(s)
Largaespada, Raul Alexander
Advisor(s)
Leonard, John J.
Bennett, Andrew
Date Issued
September 2023
Publisher
Massachusetts Institute of Technology
Abstract
Autonomous unmanned surface vehicles (USVs) and unmanned underwater vehicles (UUVs) are becoming ubiquitous in applications exploring marine environments, and the design of path planning algorithms for these vehicles remains an open area of research. For marine environments, to save on energy a path between two points should be optimized to minimize distance traveled while remaining smooth to reduce changes in speed and account for the dynamic limits of the vehicle.
The Graphs of Convex Sets (GCS) trajectory optimization motion planner from the MIT Robot Locomotion Group is a recently developed planner which has been demonstrated to return smooth and optimal paths navigating around complex environments filled with obstacles, but this planner has not been applied to marine environments. The early successes of the GCS planner and the smoothness of the trajectories returned suggest that GCS could be effectively applied to USV and UUV path planning.
This project implemented the GCS planner as part of the MOOS-IvP so software suite for autonomous marine robotics. The robustness of the trajectories returned from GCS was evaluated via Monte Carlo trials on a simulated USV traversing a field of randomized known and unknown obstacles. The performance of GCS was compared against alternate planners implementing the D* Lite algorithm or relying only on existing MOOS-IvP obstacle avoidance capabilities, running the the same simulation environment.
In testing, the GCS planner was not as successful as the D* Lite planner in navigating dense obstacle fields, but returned smoother and shorter paths than D* Lite which were easier for the vehicle to follow. Testing also suggested future modifications to the GCS planner which could be added to further increase its robustness when applied to USVs operating in dense obstacle fields.
All code developed for this project may be found at: https://github.com/rlargaespada/moos-ivp-monte-carlo.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link
