Internet des objets
Objects that are able to communicate
© Inria / Photo Kaksonen
Nathalie Mitton is the leader of the Pops project team at the Inria Lille – Nord Europe Research Centre. She obtained her habilitation to advise doctoral theses in a rapidly expanding field: the internet of objects. RFID tags and sensor networks – what is the current focus of researchers’ efforts?
What is the internet of objects and how does the subject of your research fit in with it?
Nathalie Mitton : The internet of objects is an extremely vast field, which deals with the ability to make objects communicate via a network. Objects can be connected directly to the network or be equipped with sensors (cameras, microphones, radar trackers, thermal sensors, etc.) or RFID tags, which transmit their data via wireless links. My field of research relates more specifically to RFID and sensor networks. RFID tags are made up of a chip and an antenna, which contain, for example, information about the manufacture of a product and which allow it to be tracked. These tags don’t have any batteries. You need a reader to power them and to be able to read the data they contain.
Unlike RFID tags, sensors have a battery and are able to transmit data at any time. Sensor networks are made up of a group of sensors that communicate with each other. They are used in particular for environmental monitoring, for example to monitor volcanic activity , using seismic sensors, or to identify outbreaks of forest fires using temperature sensors. They can also be used to detect changes in structures or to monitor hospital patients’ vital signs.
What are the main scientific challenges associated with RFID?
Nathalie Mitton : Creating these networks requires a great deal of research into the hardware . The sensors need to be designed to be small, inexpensive, use very little energy and have a limited impact on the environment. The batteries also need to be small, as well as biodegradable and able to last as long as possible. We need to design tiny antennae, which are able to transmit a clear signal that is relatively undisrupted by obstacles. Other challenges in terms of hardware are emerging for specific uses. For example, biologists would like to use sensors to be able to track wild animals, such as penguins. In this case, the sensors would also need to be able to withstand the cold, water and salt. There are the same concerns for RFID tags, with additional specific challenges associated with their use. They must, for example, be able to withstand washing, for RFID tags attached to clothing, and sterilisation, for those placed on surgical instruments.
In addition, there are algorithmic challenges, which affect us more directly and which relate to intended applications. A particular challenge relating to the use of RFID tags is to successfully read the greatest possible number of tags in the least possible time, in order to be able to rapidly identify a set of pallets moving along a conveyor, for example. Another objective is to limit conflicts between readers, in order to improve the reading rate, as, when reader signals overlap, they fail to recognise the tag signal. There are also a number of security issues : it is important to ensure that RFID tags on garments or in the bags of passers-by cannot be read by just anyone. These chips are currently deactivated at the till. However, manufacturers that have monitored the object’s entire life using the chip would like to be able to receive this information as part of their after-sales service. Integrating a password would be one option, but would not be sufficient to protect privacy. We are seeking an alternative solution.
… and for sensor networks?
Nathalie Mitton : One challenge for sensor networks is to find the best way of ensuring that these sensors communicate effectively . Due to their short range, they cannot directly send data gathered to their base station, which could be some distance away. They have to make use of relays - in this case, other sensors located between them and their base station.
The aim is to devise algorithms that would allow these sensors, with extremely limited computing, memory and power capacities, to identify the most appropriate recipient sensor for their data , in order for it to arrive safely. I am concentrating specifically on this aspect by studying self-organisation algorithms. These algorithms enable sensors, when dropped by a plane onto a volcano, to identify other neighbouring sensors and recognise which links to maintain with these other sensors, by means of very simple calculations. They also make it possible for sensors to recognise when to switch to standby in order to save power and which data to send, at what intervals (as infrequently as possible) in order to extend their battery life.
What is the next stage for these networks?
Nathalie Mitton : At present, new components, known as actuators, are being incorporated into sensor networks. While the sensor gathers information about the environment, the actuator is able to have an effect on the same environment . For example, when the sensor detects an increase in temperature caused by fire, the actuator triggers the operation of fire hoses. These actuators may be capable of moving about, like little robots. By combining sensors and actuators in a fleet of robots, it is possible to monitor a specific event.
Another stage involves testing our algorithms on a large number of sensors, in order to test their reliability. Thanks to experimental platforms, it is possible to load the code to be tested automatically onto all the sensors or actuators at the same time — rather than one-by-one — via a web interface, and to use tools to analyse the test and gather a large volume of data, on the consumption of nodes, for example. We are currently working on the Senslab platform, created in 2009 and combining 1,024 sensors on four sites (Lille, Grenoble, Rennes and Strasbourg). Senslab will soon form part of the FIT excellence-in-equipment project, which will incorporate sensors and actuators, among other components. This new equipment will allow us to test all our new algorithms.
Without any visibility, how is it possible to select the appropriate neighbouring sensor able to convey data to the necessary destination, and ignore all other sensors, while at the same time ensuring that this will not result in another sensor that is more remote from the base no longer being to reach its recipient (i.e. the same base station)? This is the problem faced by researchers working on sensor networks. Within these networks, data reaches the base station by “short hopping” from one sensor to another. These are known as multi-hop radio connections. To resolve this problem, they use local algorithms, which are based on information relating to what takes place between neighbours: I can talk to A and B and I know that A and B can talk to each other. I can therefore forget about A because I know that I can talk to it via B. It goes without saying that the problem becomes substantially more complicated when the sensor is displaced by water, for example, or is attached to an animal that is moving about!