Maritime intent estimation and the detection of unknown obstacles

Abstract

The benefits of using Unmanned Undersea Vehicles (UUVs) in maritime operations are numerous. However, before these benefits can be realized, UUV capabilities must be expanded. This thesis focuses on improving certain aspects of the Maritime Reconnaissance and Undersea Search and Survey capabilities of a UUV. An algorithm is first presented which provides the UUV with the ability to estimate the intent of the contacts it is observing (intent estimation). This was accomplished by developing a probabilistic model of the contact's possible intents and then using those models to estimate the contact's actual intent. The results from that algorithm are used to analyze the contact's observed path to determine a probabilistic belief of the potential location of obstacles in the environment (obstacle detection) that the contact is avoiding. These values are recorded in an obstacle inference map, which is capable of incorporating the results from the analysis of any number of observed paths from multiple contacts. The laws of probability were used to develop the algorithms in this thesis-with an emphasis on Bayes' rule. Various scenarios are presented to demonstrate the capabilities and limitations of the intent estimation and obstacle detection algorithms.