Academic
Publications
Adaptable data models for scalable Ambient Intelligence scenarios

Adaptable data models for scalable Ambient Intelligence scenarios,10.1109/ICOIN.2011.5723138,Alessandra De Paola,Giuseppe Lo Re,Fabrizio Milazzo,Marco

Adaptable data models for scalable Ambient Intelligence scenarios   (Citations: 1)
BibTex | RIS | RefWorks Download
In most real-life scenarios for Ambient Intelligence, the need arises for scalable simulations that provide reliable sensory data to be used in the preliminary design and test phases. This works present an approach to modeling data generated by a hybrid simulator for wireless sensor networks, where virtual nodes coexist with real ones. We apply our method to real data available from a public repository and show that we can compute reliable models for the quantities measured at a given reference site, and that such models are portable to different environments, so as to obtain a complete, scalable and reliable testing environment. Index Terms—Ambient Intelligence, Hybrid Simulation, Wire- less Sensor Networks, Environmental Data Modeling. Ambient Intelligence is a branch of AI that focuses on adapting the environmental conditions to maximize the user's comfort, and aims to do so transparently by applying meth- ods and ideas borrowed from such fields as pervasive and ubiquitous computing. The underlying assumption is the avail- ability of tools for extensive and timely monitoring of the environment under observation, as well as the construction of predictive models that reliably reproduce the behavior of the physical phenomena of interest. A sensing and communication infrastructure that is increasingly gaining popularity in this context is the Wireless Sensor Network (WSN) technology (1), thanks to its versatility and to the possibility of carrying on limited computations on board of the nodes. A common approach to assessing the validity of AmI sys- tems is to develop a full functional prototype of the intelligent application, and to actually deploy it into the environment. This is for instance the solution adopted for iDorm (2), a prototype for a student dormitory that allows the simulation of different everyday life activities. AmI applications usually require the creation of predictive models from sensed data; for instance, in the Neural Network House (3) a neural network system was used to forecast future environment state and users' occupancy. The intrinsic drawback of AmI tests based on actual deploy- ments is that it may prove costly in complex environments, such as entire buildings; moreover, it does not allow to test application scalability, nor to evaluate the application behavior across different configurations. An alternative approach consists in simulating the whole control loop, from sensing the physical phenomena of interest, to performing artificial reasoning, and finally to modifying the environmental conditions. The Intelligent Home (4), for instance, is a simulated testbed intended as a support for the development of multi-agent systems. However, while the application logic is in general easily reproducible, it is dif- ficult to capture the runtime overall behavior of the whole system. Moreover, early detection of design errors, and fine tuning of critical factors, such as the position and number of sensor nodes in the various areas of the test site, may avoid
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...The validity of such consideration has been experimentally confirmed by checking that the monitored quantities did present intrinsic redundancy, as shown by the autocorrelation plots reported in our previous work [16]; namely, both temperature and humidity present autocorrelation higher than 0.9, while for light is is higher than 0.75...

    Alessandra De Paolaet al. Predictive models for energy saving in Wireless Sensor Networks

Sort by: