Valuation Network for Ongoing Assessment of Threat to an Underwater Vehicle

Abstract

The paper develops a valuation based system for reasoning under uncertainty in the context of threat assessment onboard an underwater vehicle. The focus is on threat posed by the nearby contacts, while the vessel is navigating busy waters with warships, merchant ships and fishing vessels. A graphical model of a valuation network is developed, representing the (uncertain) contextual prior knowledge and received observations over the course of time. Two types of valuations are considered in this context: (1) probability mass functions, assuming that all probabilistic values are known precisely; (2) credal sets (sets of probabilities), when probabilistic values are specified only as the confidence intervals. The performance of two valuation networks is presented, using a typical scenario-log, involving a varying number of different types of contacts over time.

Keywords

• Machine reasoning
• graphical models
• valuation networks
• maritime threat assessment