A development in the line of the composite objects (see section ) are self-describing objects. While previous works enabled integrity checking over a set of physical objects, these mechanisms were limited in two aspects: expressiveness and autonomy. More precisely, objects support the detection of special conditions (such as a missing element), but not the characterization of these conditions (such as describing the problem, identifying the missing element). Moreover, this compromises the autonomous feature of coupled objects, which would depend on external systems for analysing these special conditions. Self-describing objects are an attempt to overcome these limitations, and to broaden the application perspectives of autonomous RFID systems.

The principle is to implement distributed data structure over a set of RFID tags, enabling a complex object (made of various parts) or a set of objects belonging to a given logical group to "self-describe" itself and the relation between the various physical elements. Some applications examples includes waste management, assembling and repair assistance, prevention of hazards in situations where various products / materials are combined etc. The key property of self-describing objects is, like for coupled objects, that the vital data are self-hosted by the physical element themselves (typically in RFID chips), not an external infrastructure like most RFID systems. This property provides the same advantages as in coupled objects, namely high scalability, easy deployment (no interoperability dependence/interference), and limited risk for privacy.

However, given the extreme storage limitation of RFID chips, designing such systems is difficult:

In the context of RFID system, the resiliency property of such data structures enables new information architecture and autonomous (offline) operation, which is very important for some RFID applications. We previously applied the self-describing objects approach to the waste management domain , which has shown to be a specially challenging situation for RFID. This challenge is found more generally in pervasive computing scenarios involving RFID reading in uncontrolled environments (see section ).

Pervasive support for RFIDs.

We propose to apply our approach to improve the robustness of RFID inventories / batch checking: when many objects are read at once by an RFID reader, miss read are common and raise reliability and operational issues for applications. An innovative solution to this problem is to take advantage of the multiplicity of tags by leveraging them as a distributed memory shared by a logical group. In this way, it is possible to support error detection as well as information recovery. We proposed a flexible protocol to support robust EPC retrieval in adverse reading conditions. The proposed protocol uses erasure correcting techniques to enable error-free recovery of misread EPCs . It is further customizable with respect to the rate of misread tags and application requirements. This work was the object of an Inria patent Patent filed in April 2015 - Inria 179. Fine-tuning the protocol parameters is still the object of on going experimentation in the context of the Pervasive_RFID project (see section ).

At the software level, RFID inventory reliability issue is usually addressed by anti-collisions mechanisms and redundancy mechanisms. Anti-collisions protocols limit the risk of data corruption when multiples tags have to reply to an inventory request. Redundancy is often implemented in RFID readers by aggregating the results of multiple inventory requests over a time frame, to give the tags multiple opportunities to reply. While useful, these strategies cannot ensure that a given inventory is valid or not (in other words, one or more tags may be missing without being noticed). In situations where we have to read large collection of objects of various types, the performance is difficult to predict but may still be adequate for a given application. For example, some application can tolerate missing some tags, provided that miss read probability could be characterized. In some cases, read reliability could be improved using mechanical approaches, such as introducing movements in objects or antenna to introduce radio diversity during read. Finally, distributed data structure can be used over a set of tags to be used to mitigate the impact of misread (by using data redundancy) and to help the reading protocol by integrating hints about the tag set collection being read.

We studied extensively by experimentation the behaviour of existing RFID solutions in the context of uncontrolled environment (meaning, random placement of tags on objects mixing various materials) in order to characterize their real-world performance regarding the parameters of such as tags numbers, density, frequencies, reader antenna design, dynamicity of objects (movements), etc. From these experimentations, we would like to identify the conditions that are favorable to acceptable performance, and the way where there are hopes of improvement with specific design for these difficult environments. These results should also allow improving the performance: high level integrity checks can guide low level operations by determining whether inventories are complete or not. This cross layer strategy enables faster and more efficient inventory protocols.

Tacoma group is focussed on the conception and implementation of innovative services for the Smart Home/Building. The range of considered services is broad: from "optimizing the energy consumption" to "helping users to find their way in a building". One of our goals is to build a pervasive platform with constrained performance and cost , without disrupting existing spaces. Within this idea, we explored in 2015 the services provided by different modes of interaction in a physical space between neighboring objects, and also between an object and a nearby user.

More precisely, we conducted some experiments with LEDs. Instrumented via a short distance radio interface, a lighting device becomes an unobtrusive connected object that is easy to integrate to a mesh network. A relevant aspect of this platform is the consideration of potential conflicts in data access offered by the connected objects. One of the first scenarios we considered is to operate an LED-based light path to guide the evacuation of a building in the case of a fire alarm. When our objective is to multiply the uses of LED devices ("go beyond lighting", see section ), the question is then the priority of access to resources offered by the platform distributed in the environment. Specifically, we addressed the following issues (similar to some of the issues presented in section ):

To provide services for smart Homes, automation based on pre-set scenarios is ineffective: human behavior is hardly predictable and application should be able to adapt their behavior at runtime depending on the context. We focused on recognizing user's activities to adapt applications behaviours. Our aim is to compute small pieces of context we called context attributes. Those context attributes are diverse, for example a presence in a room, the number of people in a room etc.

Building efficient and accurate context information using inexpensive and non-invasive sensors was and is still a great challenge . We proved, through the use of dedicated algorithms and a layered architecture that it is achievable when the targeted Home is known - due to the specific and non automated calibration process we used. Among all the available theories, we used the Belief Function Theory (BFT) as it allows to express uncertainty and imprecision.

Context is computed by a chain a tasks as illustrated in figure :

Alleviating the complexity of the platform configuration and maintenance is a prerequisite for the adoption of Smart-Home environments by consumers. Currently the BFT theories requires a huge calibration process. We focussed our efforts on the semi-automated building of belief functions, required by the theory, that have to be provided by each sensor.

Automated configuration of sensors.

The belief model is provided to the platform by us and a component is in charge of transforming a sensor value in a belief function. The fine tuning of a belief function can be a tedious task. It must be done by a specialist who understands the belief function theory and knows the behavior of the sensors. The model is often built iteratively by experimenting. This may take several hours or days. Moreover, this method is directly connected to the output of each sensor. Biased and noisy measures can cause major modifications on the resulting beliefs.

Ideally, the calibration of the model should be as automatic as possible (few interaction with the user during calibration). The person setting up the sensors should not have to understand the belief function theory. We proposed to generate our belief model from a training set of sensor data. We mainly focused on k-nearest neighbors (KNN) algorithm . We used a training data set to compute the presence belief model. We acquired a set of data with someone present in the experimentation room and a second data set with nobody in the room, which gives us a labelled data set. This principle is illustrated in figure .

Pervasive environment brings into play complex interactions between a large number of heterogeneous entities: computing units executing third-party applications delivering multiple services to users, with various (sometimes conflicting) requirements, based on the information provided by dynamically (un)available smart object or sensors. The development of pervasive application is consequently hard and must be supported by architectures and frameworks that propose solution to manage the heterogeneity, to organize the interaction of distributed entities, to support the dynamic discovery of the entities, to ensure the privacy of collected data and inferred context, to organize and structure information sharing, and to enforce access control over data and entities.

To alleviate the development of such application (see section ), we worked on a distributed pervasive environment made of several processing nodes (or gateway) managing interacting Smart Spaces (i.e a room, a corridor etc.). A Smart Space contains one or more nodes that coordinate to provide services to users. A node is a low cost computing unit with constrained performances. Each node is responsible for the management of entities and services available in their close proximity: they dynamically discover available devices and source of information, computes contextual information and offer services to nearby users. Nodes are organized hierarchically: in each space one supervisor node is responsible for coordination between nodes (e.g. managing conflicting requirements and enforcing global policies) and communication with neighboring Smart Spaces. The whole environment (a Smart Building or a Smart Home) is controlled by a master node that distributes policy between Smart Spaces to provide global services (e.g. global energy management). Data are stored and processed by nodes as close as possible to the users in order to enforce data privacy (see section ).

Each node supports the execution of several services and application. To help the development of these services, applications are built on top of a framework (Matriona) running on each gateways.

Matriona proposes a unified representation of the concepts manipulated by applications in order to hide the heterogeneity of technologies to the application. We rely on the concept of resource. A Matriona's resource is quite similar to a REST resource but is not identical: HTTP protocol is only used for remote call; the structure of a resource is constrained; a resource can be dynamically discovered; it can provide notification (PUSH operation); it is uniquely identified in the environment; it is any object used or shared by applications, by bridges, or by the system itself: a device, a room, a platform, a user or a contextual information; it is implemented as a standard object and has a type specified by its interfaces (annotated Java interface). Types describe the data and the operations on resources. CRUD (Create, Read, Update, Delete) and PUSH operations are used to represent the semantic of operations. Specifying the semantic of operations allows to operate some generic processing (e.g. gathering all the information provided by a resource) on resources without knowing their types.

Matriona enables dynamic discovery between nodes and inter-platform communication between the applications and resources. The platforms automatically discovers other platforms using a discovery mechanism. Each platform enables the discovery and the use of its resources to other platforms' applications. Remote resources are using exactly the same API as their internal counterparts: using remote resources is completely transparent to the application. The data is serialized / deserialized automatically by the platform during the call. Calling remote resources induces a performance cost equivalent to the use of a traditional REST resource.

Matriona allows to dynamically add new properties and new behaviors to a resource using decorators. Resources are built using multiple layers in the same way as "Russian dolls". Each layer is responsible for implementing a specific behavior such as retrieving, conversion operation or adding new properties (e.g. localization). For instance, the implementation of a thermometer resource will consist of 1) a core layer providing standard information (id, date of creation, the groups to which it belongs); 2) a protocol layer able to communicate with the real devices; 3) a conversion layer (Kelvin to Celsius). The application interacts with the top layer that exposes all or part of information and treatments offered by lower layers. While some layers are static (part of the resource declaration) and cannot be removed, most can be added afterward by the applications. It creates a virtual resource composed of the original resource and the new layers. This virtual resource can be used and discovered as any other resources.

Matriona allows to organize the information. Resource may reference other resources, for example the localization of a "thermometer resource" refers to the resource representing the room in which it is located. The value of the property is the id of the referenced resource. This allows applications to easily find resources and their interactions. It is also possible to create composite resources using aggregation mechanisms provided by the framework. The virtual resource can be used directly by applications as any other resources.

Matriona provides a basic language queries. Applications use resources directly or send queries to the platform registry. The query language allows to apply filters and to aggregate data on available resources. A request is represented by a specific URL. For instance, the mean temperature of thermometers in the whole meeting rooms of a building can be obtained using the URL /*/*/$ location/meeting_room/temperature!mean. The query language can be extended by providing new decorators and new filters.

Matriona provides access management: each resources belongs to one or many groups. The groups are defined when the resource is created or during its lifecycle by the owner of the resource. Groups gather applications that share the same permissions on access resources. Groups are managed by a "group owner" that can limit members permissions. Permissions describe the ability of an application to read, write, update, delete, manage or lock a resource. Resource locking avoids conflicting requests to be performed by different applications. Locks are given to applications for a fixed period of time. A resource can always be unlocked by the platform itself or by "critical" applications (e.g. emergency fire alarm, see section ).

Matriona allows applications to extend resource properties and to share these meta-information with others. Each application can add new information to a given resource. Tagged information are available only for the application and its group. Meta-information are stored by the platform and associated to the resource until the latter is destroyed. This mechanism allows application to easily share information on the resource they used. For instance, this can be used to retrieve previously used resources or to rate the quality of service provided by a given resource. For the application, meta-information are part of the resource. It is then possible for an application to only use resources that have been approved by other applications of their group. This mechanism can also be used by application to add some task-relevant information (e.g. a medical application can tag resources that have been used by a patient).

This research activity is supported by Fondation Rennes 1 through the chair "Smart Home and Innovation", since January 2014. This activity is centered on the concept of metamorphic housing (see section ). During the first year, we had identified the goals of the research project, also taking into account the trends of future housing industry, provided by the enterprises and public authorities that support the chair. We also had identified a case study, the on-demand room, to be displayed as the main application of the research results in scientific communications and vulgarization. It consists in a space that is physically shared by a small group of apartments, but is assigned for the sole use of one or few particular ones at the time. The room is designed so as to make occupants feel they did not leave their apartment at all. They seamlessly move from their dwelling to the on-demand room and conversely, without noticing the difference, as the room adapts to their preferences. During 2015, second year of the chair, we organized our work following two main axes: (i) solving the research problems, illustrated in the rest of this section; (ii) demonstrating the results using mixed reality as combination of virtual reality and off-the-shelf domotic devices, described in section .

The research problems underlying the on-demand room are numerous: we illustrated them in the research report "A Case Study of Metamorphic Housing: The on-Demand Room” . We started by addressing the problems associated with the goal of "plugging” the room into different apartments. This requires to dynamically change the rights to control and customize the room's equipment, including lights, appliances, heating, ventilation and air conditioning systems (HVAC), etc. This must be done in a transparent fashion, so that off-the-shelf devices and appliances can be used.

To solve these problems, we started a collaboration with the DIVERSE team http://diverse.irisa.fr/. The goal is to use the Kevoree http://kevoree.org/ software framework to dynamically reconfigure the networks and domotic system of the room and of the apartments. When the on-demand room is owned by an apartment, their computer networks are interconnected; appliances, sensors and controllers in the room and the apartment can communicate with each others; devices reflect user preferences. Kevoree will enable these reconfigurations by running on key appliances and dynamically adapting and customizing their behavior to the owner of the on-demand room.

As part of the collaboration, some research goals have already been identified. The underlying challenges will be addressed and the results will be integrated in a comprehensive mixed reality demonstrator. This will represent the final iteration of the ongoing demonstration process, illustrated in the platform section (for more details, see ).