Section: New Results
Spatial Computing approach and RFIDs
Participants : Michel Banâtre, Paul Couderc [contact] , Yann Glouche, Arnab Sinha.
In the line of our previous research in pervasive computing, we are working on spatial computing approaches in the context of RFID. Spatial computing consists in data structures and computing processes directly supported by physical objects. RFID is an attractive technology for supporting spatial computing, enabling any object to interact in a smart environment. Traditionnal RFID solutions use a logical model, where the RFID tags are simple identifiers referring to data in a remote information system. In our approach, we use the memory of the tags to build self-contained data structures and self-describing objects. While featuring interesting properties, such as autonomous operation and high scalability, this approach also raises difficult challenges: the memory capacity of the tags is very limited, requiring compact and efficient data structures.
Our research in the context of domestic waste management is broadly investigating the use of RFID at item level to provide early waste sorting, to avoid incompatible mix of waste and to prevent hazards [3] , [4] . Several innovative aspected are studied in this project. First, the design of an autonomous computing architecture for the waste items and smart containers, enabling early processing in the waste management: for example waste bags can be accepted or rejected accordingly to their content and its conformance with the recipient container. Hazard prevention and human operator safety can also be improved with the knowledge of the nature of the waste.
Autonomy is important as it would be possible to depend on a remote information system for each waste insertion, due to obvious scalability, energy and network costs. An ontology based system has been proposed to determine the possible interactions of tagged products based on their properties and the external conditions [6] . This ontological model is simple enough to be supported entirely by a low power embedded computer at the container level, but can still support the waste application requirements. An unconventional aspect in this architecture is that semantic properties are directly written in the RFID tags, instead of semantic-less identifiers typically used in most RFID applications.
A second innovative aspect of the research is to consider the set of containers in a city as a particular case of sensor network, and developing energy efficient protocol to enable information reporting to a supervising infrastructure.
In the context of this research, some limitations of existing RFID technology become challenging: unlike standard RFID application scenarios, pervasive computing often involves uncontrolled environment for RFID, where tags and reader have to operate in much more difficult situations that those usually encountered or expected for classical RFID systems. In a near future, we seek to work with a team who has a strong expertise in antenna design and radio signal behaviour.