Uncertainty propagation methods for collision avoidance in autonomous navigation (GANESHA)
Abstract
This project aims to analyze the behavior of error propagation methods applied to autonomous maritime navigation. Different uncertainty propagation methods exist in the literature. Each of them has advantages and disadvantages. The methods that require lower computational cost are almost exclusively applicable to linear dynamic systems, while the methods developed for non-linear dynamics require a higher computational effort. In the autonomous navigation problem, several objects need to be tracked at the same time, so it is advisable to keep the computational cost low. At the same time, the method should be adaptable to complex non-linear dynamics. Inthis project, the aim is to adapt and develop uncertainty propagation methods for the autonomous navigation problem. In particular,tracking and filtering methods will be developed to estimate collision probabilities in realistic scenarios. In parallel, it is proposed to analyze different artificial neural networks for obstacle localization and their behavior under low visibility conditions. Many autonomous ship surveillance systems are based on optical imaging of the environment. In conditions of reduced visibility, classification algorithms fail to detect objects or misidentify them. In this project, the behavior of neural networks and other classification methods will be studied under low visibility conditions and with images with a low signal-to-noise ratio.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 09, 2024
- Source ID
- N629092412095
Entities
People
- Javier López Santiago
Organizations
- Office of Naval Research
- United States Navy
- Universidad Carlos III de Madrid