The development of a multi-objective decision making and learning model for the use in unmanned systems is the focus of this project. Starting with traditional game theory and psychological learning theories developed in the past, a new model for machine learning is developed. This model incorporates a no-regret decision making model with a Bayesian learning process which has the ability to adapt to errors found in preconceived costs associated with each objective. This learning ability is what sets this model apart from many others. By creating a model based on previously developed human learning models, hundreds of years of experience in these fields can be applied to the recently developing field of machine learning. This also allows for operators to more comfortably adapt to the machine's learning process in order to better understand how to take advantage of its features. One of the main purposes of this system is to incorporate multiple objectives into a decision making process. This feature can better allow its users to clearly define objectives and prioritize these objectives allowing the system to calculate the best approach for completing the mission. For instance, if an operator is given objectives such as obstacle avoidance, safety, and limiting resource usage, the operator would traditionally be required to decide how to meet all of these objectives. The use of a multi-objective decision making process such as the one designed in this project, allows the operator to input the objectives and their priorities and receive an output of the calculated optimal compromise.
| Identifer | oai:union.ndltd.org:ucf.edu/oai:stars.library.ucf.edu:etd-4589 |
| Date | 01 January 2008 |
| Creators | Howard, Matthew |
| Publisher | STARS |
| Source Sets | University of Central Florida |
| Language | English |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Electronic Theses and Dissertations |
Page generated in 0.0122 seconds