Sensing in SSN (and SOSA)

This wiki page is about the concept of Sensing in the old/original SSN and how to deal with it in the reworked version of the SSN as well as SOSA. It addresses Action 302. The concept of Sensing and its relation to Process and Observation has been a source of confusion in the old SSN with users not being clear about their relation.

Following these two potential interpretations, in the old SSN Sensing was necessary either because Observation in SSN was not an event but a Situation/Context. The 'act of sensing' as we called it in the original SSN is what is now called Observation in SOSA/new-SSN, namely the action that results in the estimation, or calculation, of the value of a phenomenon; see also Simon's clarification: https://lists.w3.org/Archives/Public/public-sdw-wg/2016Aug/0144.html . In SOSA and the new SSN, the act of creating samples is called sampling, the act of changing a state in response to some trigger is called actuation, and the act of transforming a stimulus into a result is called an observation.

As far as the other interpretation is concerned, procedures in SOSA and SSN are like recipes and not actual executions. One potential source of confusion about Sensing in the old SSN was what exactly is meant by the 'process of sensing', if it is indeed the description of how the sensor operates, and where the actual event of it operating (the observation) is modeled. One can see where the confusion arises from sentences such as 'We assume that objects are sensors while they perform sensing' followed by 'Features of Interest are entities in the real world that are the target of sensing' (note that this does not refer to types but instances of FOI), and 'Consequently, sensors can be thought of as implementations of sensing methods where different methods can be used to derive information about the same type of observed property [...] Simplifying, one can think of sensing as recipes for observing.'

Sometimes we tried to escape this problem by refering to 'sensing method', 'sensing object', 'sensing infrastructure', 'sensing capability', and 'sensing tasks', clearly some of them relate to events while others relate to the process/procedure view.

Proposal: For the sake of simplicity and to avoid confusion on the side of users, we could deprecate 'Sensing' and define it as a subclass of Procedure (in SOSA/new SSN). If we want to keep the concept (and can provide implementation evidence), we need to clarify that it is more about the principal functioning of the sensor than about the actual execution of events that lead to (observation) results.

While this is out of scope here, similar simplifications can be applied to the relation between Sensor, Device, and SensingDevice in old-SSN.

Simon Cox (talk) 20:16, 11 April 2017 (UTC) added

The PROV-O alignment by Compton, Corsar and Taylor [SSN-PROV] is helpful in disentangling, which I also used in [OM-Lite]. For each of Sensing, Sampling and Actuation we have a an 'Act of X' (a kind of prov:Activity) 'Procedure for X' (recipe for and 'Device that implements procedure for X' (a kind of prov:Agent).

I also introduced separate X-Procedure and Sampling-device and Sampling-event in the O&M alignment http://w3c.github.io/sdw/ssn/#OM_Alignment_utility

[OM-Lite] Ontology for observations and sampling features, with alignments to existing models. Simon Cox. IOS Press/Semantic Web. 6 December 2016. URL: http://content.iospress.com/articles/semantic-web/sw214

[SSN-PROV] undefined. Michael Compton; David Corsar; Kerry Taylor. CEUR: 7th International Conference on Semantic Sensor Networks. 2014. URL: http://ceur-ws.org/Vol-1401/paper-05.pdf