5 Motivation

The assumption of many existing oneM2M applications is that they interact with other oneM2M applications through known resource structures. They either create the resources themselves or are configured to use specific resources. Information is typically stored in containers, sometimes as base64-encoded content instances, with the implicit assumption that applications have a-priori knowledge of the syntax and semantics of this information.

Depending on a-priori knowledge of the structures and data works well for small-scale vertical deployments of IoT devices. When the deployment evolves to include new devices, the existing applications change to reflect the new additions. However, in larger systems of IoT devices where the IoT devices may be a part of a legacy deployment or more than a single vertical solution, changes to all existing applications may become impractical. To enable growth and diversity of IoT devices in large heterogenous settings, applications need to be able to discover the structure and meaning of data from devices and how to use the services of the devices. In oneM2M Release 1 support for discovery of resources based on specific attribute values and the use of labels was defined. The agreement of a fixed set of labels (using a-priori knowledge) can be a viable solution for small deployments.

Figure 1: Semantic understanding of device and data in IoT deployments

For medium or large deployments of heterogeneous IoT devices a more expressive approach for describing and discovering IoT devices is provided by oneM2M. Each type of device in a heterogeneous deployment can model services and data in the oneM2M Service Layer using different structures and syntax of data. For example, temperature sensors may report measurements using different units such as Celsius, Fahrenheit and Kelvin. Additionally, those IoT devices may measure different aspects, such as indoor temperature, outdoor temperature, refrigerator temperature, etc., and the representation of the measurement may differ as well.

Figure 2: Meaningfulness of data from IoT devices

With semantic annotations in oneM2M, all the different aspects of IoT devices can be described using RDF triples, which is a standard semantic format. The vocabulary used for a semantic description can be defined according to an ontology such as SAREF. With semantic discovery, applications can describe precisely what information they need or can deal with. This is powered by specifying a semantic filter using the SPARQL query language. The SPARQL filter is matched against the respective semantic annotations of each resource within the discovery scope. This feature in oneM2M helps applications to properly handle the data from the IoT devices.

Besides differences in the data from an IoT device in oneM2M the information model of devices can be modelled in a variety of ways. As with most IoT platforms, oneM2M supports custom information models that are defined for a specific use case and work well in small scale or single vertical scenarios. Another method that oneM2M defines to model devices is based on the semantic description of a device that is mapped to a resource structure (see TS-0030). A third approach to modelling devices in oneM2M is the use of Smart Device Templates (see TS-0023).

With all these options available to model a device the ability to have a-priori knowledge of a device model becomes less likely as IoT deployments scale beyond small vertical use cases. The oneM2M Base Ontology addresses this and enables developers of these different models to make them interoperable if the appropriate semantic annotations are made and semantic filtering is used to discover the appropriate API for a model. The focus of the remainder of this developer guide is to demonstrate this process.