Markov Logic Networks for context integration and situation assessment in maritime domain

The detection of anomalies has a critical role in situational assessment. In this paper, we break down the concept of anomaly in the maritime domain into different levels and relate them to the JDL fusion model. We also show how uncertain context knowledge can be encoded through Markov Logic Networks (MLNs) which offer a convenient framework leveraging both the expressive power of first order logic and the probabilistic uncertainty management of Markov networks. Every formula in the knowledge base is assigned a weight indicating its confidence. Different types of knowledge with associated uncertainty can therefore be fused together within MLNs and on-line inference can be performed as input data is processed by the system, and the formulas are grounded in the knowledge base. Promising examples are demonstrated on a sample set of rules for maritime event and anomaly detection.