Modeling of confidence

Confidence is defined by: "Every observation in the automotive domain is an estimated observation coming from the aggregation of several observations from different sensor or devices taken at the same time or at consecutive times very close to each other. Therefore, a confidence is always associated to each mesurement to indicate that it has not 100% accuracy."

Associated entities are: hasConfidenceUnitOfMeasure hasConfidenceValue hasConfidence, isConfidenceOf

and subclasses:

• :SpeedConfidence (value is int: an absolute accuracy (?!) in cm/s) ;
• :AccelerationConfidence (value in float, absolute accuracy with a predefined level (we don't know what level))
• :PositionConfidenceEllipse (represents horizontal position accuracy in a shape of ellipse with a predefined confidence level.

Although it may be useful to describe the confidence level of an observation (or in general: a procedure execution), and this generalizes outside the automotive domain, I believe the current modelling is erroneous and overly complicated.

The confidence value could simply be modelled as a datatype property attached to the observation, with a decimal value between zero and one. For example:

saref:hasResultConfidenceLevel a owl:DatatypeProperty ;
  rdfs:domain [ a owl:Class ; owl:unionOf ( saref:Observation saref:Actuation ) ] ;
  rdfs:range xsd:decimal ;
  rdfs:label "has result confidence level"@en ;
  rdfs:comment "links an observation or actuation to its result's confidence level, expressed as a value between zero and one"@en .