Publication:
AUTONOMOUS VEHICLE FIELD DATA GENERATION FOR MACHINE LEARNED (ML) COLLISION PREDICTING ORACLES (CPOS)

Loading...
Thumbnail Image
Authors
Bridget, London IV
Subjects
autonomous vehicles
cross entropy
Advisors
Drusinsky, Doron
Date of Issue
2024-03
Date
Publisher
Monterey, CA; Naval Postgraduate School
Language
Abstract
Machine Learning algorithms have played a transformative role in the development of autonomous vehicles. As the race for fully autonomous vehicles accelerates, the pivotal role of machine learning algorithms in decision-making, object detection, and predictive modeling becomes increasingly apparent. However, the training of these algorithms poses significant challenges, especially with regards to scale, variability and quality of data required. This research begins with an exploration of how rich, diverse, and controlled environments provided by simulators can yield a large dataset that is crucial for the initial training of machine learning models. Simulation data offers the possibility to quickly generate and test multiple scenarios, including rare events, which are challenging to obtain from real-world driving. Moreover, the study illuminates how simulation provides a controlled and repeatable testing environment that mitigates the risks associated with real-world testing. Subsequently, the focus transitions to real-world data’s role in validating and testing the models. Despite the advantages of simulation data, it is essential to confront models with the complexity and unpredictability of real-world scenarios for thorough validation. This ensures the autonomous system’s safety and robustness under a wide range of conditions.
Type
Thesis
Description
Series/Report No
Department
Computer Science (CS)
Organization
Identifiers
NPS Report Number
Sponsors
Funder
Format
Citation
Distribution Statement
Distribution Statement A. Approved for public release: Distribution is unlimited.
Rights
This publication is a work of the U.S. Government as defined in Title 17, United States Code, Section 101. Copyright protection is not available for this work in the United States.
Collections