A Bayesian Network for Combat Identification
Abstract
This paper reports the results of an investigation at TNO Physics and Electronics Laboratory for the Royal Dutch Navy on how a Bayesian network can be used to introduce more transparency in decision aid as to the level of confidence of the information that is used. To assess the feasibility of a Bayesian approach to new decision support system concepts, we have chosen to use the identification process of air contacts aboard navy frigates as a case study. The identification process is a time intensive and mind consuming process that is critical to anticipate an air attack. Wrong decisions may have fatal consequences and for example: identifying a neutral aircraft as an hostile aircraft may cost the lives of many people and may cause undesired political instability. Vice versa, mistakenly taking a hostile aircraft for a friendly one will give opponents the tactical advantages of a surprise attack. Typically a list of identification criteria is used by air defense personnel to discriminate hostile aircraft from neutral and friendly aircraft. These predefined criteria may change from mission to mission but will always follow a strict scheme of if-then-else clauses. Although identification involves reasoning with uncertain information, current procedures do not make this uncertainty explicit. Embedding Bayesian inference techniques in decision support systems would enable us to reason with uncertainty in a scientifically sound and consistent manner. Bayesian networks can express the likelihood of a hypothesis such as the identity of air contact being hostile as an explicit value, even when information about a contact is uncertain and incomplete. Making this uncertainty explicit, enables navy personnel to know how much confidence it should have the probability of hypotheses that are based on it.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2004
- Accession Number
- ADA428330
Entities
People
- Henk Jansen
- Ing S. Pier Van Gosliga