AR Discovery Scenarios

AR Discovery can be separated into 5 distinct problem areas. Context identification refers to the functions for determinating the context of an actor or setting. The context is the the core input for filtering relevant information.
 * Context identification
 * Information Filtering
 * Information Packaging
 * Information Delivery
 * Information Presentation

Information filtering refers to functions for selecting information that is relevant within a given context.

Information packaging refers to the formats of AR information provisioning. In this category falls smart data prefetching in order to offload large data streams for mobile use-cases. Part of the packaging problem is the identify when to load/send which parts of data in order to shape the user experience.

Information delivery refers to the protocols and data channels for providing information to the end user devices.

Information presentation is related to the user experience of getting access to relevant information.

Context Identification
Context identification refers to functions for determinating the context of an actor or a setting. In a first discussion for defining AR discovery scenarios we identified three dimensions. The location dimension combines three characteristics for framing an actor's perspective. Positions are anchored on the geo grid and can be determined via GPS or sender triangulation.
 * 1) The location dimension,
 * 2) the context dimension, and
 * 3) the "referrer" dimension.
 * Position
 * Orientiation
 * Relation

Orientiation include aspects such as "direction" or "speed". Orientation can be determinated by gyroscopes, accerelometers, or barometers.

Relational information refers to the presence and relations of actors and POIs on the grid. These relations are typically described as "next to", "in front of", "nearby", "in view", "around the corner" etc. For example, such relations can be identified via Bluetooth, wireless sender detection, or proximity sensors. The context dimension refers to meta-data or linked data that describes the context of an actor or of a POI. The two extreme positions of filtering relevant information are POI-centered discovery and user-centered discovery. Both positions may include different context dimensions for defining and controlling the discovery process (such as described by Zimmermann, Lorenz & Oppermann, 2007 ).

Finally the "referrer" dimension refers to external references that guide the discovery process but are external to the users' devices. Such referrers provide pointers to or even the actual information to agents. This way actors can "find" new information by utilizing the sensorium of the envrionment. Such referres can be dumb markers that provide "static" references. Typical dumb markers are bar-codes, QR codes, and RFID.

To the other end of the spectrum or external referers are smart sensors that can dynamically change depending on environmental (e.g., the room temperature) and other contextual factors (e.g., the state of a machine). Smart sensors can build on NFC, bluetooth or WiFi communication for information distribution.

The combination of these dimensions and the integration of the perspectives of these dimensions allows to describe characteristic scenarios for the different forms of AR discovery with respect to the required technologies and interactions. A possible grid is illustrated on the right.

The Kurio project may serve as an example for illustrating the interplay of the different dimensions. The Kurio project uses tangibles for providing personalised museum tours. The system uses markers and beacons to identify the relative position of the visitors to the museum exhibits. Each marker provides a reference for selecting appropriate information for present location (POI meta-data) and the selected tour (user context).

Information filtering
Information filtering describes the process of relecting relevant information of a given context. This process can be static, based on pre-defined data(-types) or based on dynamic interferrence.

Static filtering includes pre-selected content (e.g. catalogues) and information channels (e.g., ATOM-feeds). Dump markers can trigger static information filtering by providing filter references similar to BitTorrent "magnet-links". Linked data concepts fall into this category, too.

Filtering based on pre-defined data types uses existing meta-data attributes that are explicitly associated to the information. This meta-data can be used for identifying the relevance of given information in a given context. For example, if an actor is looking for a restaurant at 2AM in the morning, a POI database can be queried based, on the POI-type (= "restaurant"), its location ( e.g., < 2km radius from the actors position), and the restaurant's opening hours.

Dynamic interferrence looks for similarities of unstructured information within a given context.

Information packaging
Information packaging addresses the data-structures and -formats of AR-related information. Such formats bundle contextualised information to be delivered effiently to the clients. These packages might be a set of HTML pages or audio-files that are just connected by the context.

Depending on the application scenario requirements a packaging format can include the actual information (like a ZIP-file, e-book, or SCORM-package) or a set of references (like an ATOM-feed).

Information delivery
Information delivery addresses the delivery of relevant information to a client device. This basically involves the network protocols.

Typical smart phones have several data channels ranging from cellular data via WiFi and Bluetooth to NFC. Depending on the application scenario relevant information can be offloaded to different channels. For example, for AR in a museum using cellular data might not be feasible for tourists due to roaming charges. Alternative to WiFi offloading, preinstalled NFC or Bluetooth beacons (that are also smart referrers) could provide low-threshold channels for exchanging small chunks of contextual data.