The technological developments in the field of microeletronics and sensors devices, as well as in the telecommunications technologies, have led the way to interesting applications scenarios for the development of new generation monitoring systems, the so-called Distributed Networks of Sensors. It essentially deals with ad-hoc monitoring networks, conceived for managing even a very high number of sensors, distributed in a very broad but hardly invasive manner (even throughout very extended territories), able to deliver data continuously and in real-time, entirely wireless, with low investment and managing costs. The main characteristics of the system carried-out by Pirelli – Telecom Italia are:

* Modularity and scalability
* flexibility and operability 
* reliability
* real-time data collection and wireless mode
 * advanced network functionality (plug&play of sensors, auto reconfigurability, etc...)

This technological platform of general purpose type, can be “specialized” with appropriate sensors, adapting it to various application needs. Thanks to the modular structure the system is furthermore easily updated in anticipation of future upgrades, both at a hardware level as well as of protocols and standards used.
The elevated amount of data made available in real-time by these systems, appropriately managed and processed, can enable innovative operating and managing tools (and related services) based on highly valued information. (resulting, for example, from the integration with predictive models and/or decision support).

Application for the Monitoring of Traffic

The state of the art of the hardware for the monitoring of traffic foresees the use of devices such as video cameras, radar, inductive loops, piezoelectric sensors, etc…
These systems generally present high costs and difficulties in set-up, maintenance and management, which in fact prevent a broad circulation consequently limiting even the potential  set-backs in terms of quantity and quality of the info connected to them.
The Pirelli/Telecom Italia Systems, combining specially developed innovative sensors to the technology of Distributed Networks, offer the possibility of using systems for the monitoring of traffic able to give back in real-time high quality measures with low costs. Thus enabling  to broaden and make more effective the data collection of traffic with positive set-backs in terms of the reliability  and efficiency of the models for the forecasting and management of mobility.

Here below the characteristics of the Pirelli Sensors/Telecom Italia system are described in further detail. The proposed system is articulated on three levels:

Level 1 (Sensors/ Actuators)

At the lowest level there are the “wireless sensor nodes”, installed under the road surface (5 cm depth), which allow the detection of the traffic parameters of interest, such as speed, calculation and classification. Technical details:

Level 2 (Unit/Gateway)

The data of the sensors is transmmited in a wireless and real-time mode to a roadside control unit, which covers various functions: it coordinates and manages the sensors, acting as an access point (with the possibility of a further processing of the same data) allowing long range transmission, always in real-time. It further offers the possibility of integrating practically any type of sensors used for various purposes (parameters of climate, safety, pollution, etc…) Technical Details:

* GPRS or Ethernet Connections
* Reconfigurable by Remote or via local server
* Short and long range Wireless
* Advanced radio capabilities (Network coordinator LAN)
* Back-up batteries with  the possibility of solar panel power supply or network, of  220 V

Functional Details:
* Short range/long range Gateway
* Access Point/Sensors Hub
* Management of network synchronicity for real-time applications
* Management of  Plug&Play sensors

Level 3 (Network Manager)

Software for managing the distributed network. It handles the network of units and sensors and coordinates the gathering, archiving, and validation of data in the relative database. It offers the possibility of access to external clients (via Web), implementing the levels of safety and of management of the services offered to accredited Clients. Finally it also implements the administration software which allows the fully remote management for regular and extra maintenance (for example software upgrade). This software can be integrated in already existing management structures (ex: operative units) rather than being installed independently on a PC.

The level of integration used allows the implementation of a high number of vertical applications utilizing the “general purpose” platform, and exclusively specializing it at a sensor system level or dedicated software. This gives the possibility of having a sole integrated system of control and multifunctional management, with low investment and management costs. 

Artificial Intelliegence based on Neural Networks

In the context of cooperation with TELECOM Italia SpA, D.E.C.A. has carried out the IT system of Data Management and Data Analysis described below:
The DecaVelta project is based on innovative technologies of information processing that result from Artificial Intelligence studies. Artificial Neural Networks (ANN) in particular are widely used, the software is able to process in a non-linear way large amounts of information producing  models of adaption stimulus-response that accomplish predictive models of complex phenomena. Such models are applied to foresee in an efficient manner the onset of plant diseases, reducing and rationalizing the use of treatments.
The lowest environmental impact that results from it, besides the considerable lesser impact on the ecosystem, determines a greater economic return for companies given the lower expenditure, and a better quality  and health of the product , resulting in an actual certification chain of highest profile.
ANNs were created to simulate the neural structure of the human brain, imitating the known mechanisms of learning, predominantly on an experimental basis. The most frequent use of ANNs is the supervised  learning. On the basis of the processing of a data set compared to a specific output, it’s possible to correct the parameters of the model to find relations between the data and thus generate the correct output value, by means of parallel computational operations. In this way they are able to deduct quite efficient rules to solve predictive-type problems. ANNs are non-linear mathematic models by definition,  but contrary to techniques such as non-linear regression they don’t carry out any type of hypothesis on the form of data;  instead they are indipendently able to identify interactions between variables which need to be explicitly specified in models of statistic regression. ANNs are thus appropriate for analyzing an objective/event variable in the context of a strong non-linearity of incoming data (or of inputs). 

Use of ANNs in DecaVelta

The data deriving from the Agriveltha system is pre-processed by Expert System and Statistics Technologies, which allow to implement malfunctioning signals and failures of sensors and/or of RTUs, and are subsequently processed with Artificial Intelligence techniques, especially ANN and Genetic Algorithms.

After two years of experimentation our ANNs have demonstrated that they are able to correctly indicate the infections which actually have taken place, that they never give false negatives (absence of danger warning with infection underway), and, most of all, of showing better performances than those expressed by mathematic models commonly used in viticulture.
All of the studies carried out until now in the context of this project have been the object of various pubblications at international conventions in the agronomic field.

Main pubblications on the Project

A Neural Network-Based Forecasting Model for Plasmopara viticola infection also allows to manage changes introduced by the new climate scenario: a two years study (R. Bugliosi, G. Spera, A. La Torre, L. Campoli) Congrès sur le Climat et la Viticulture - Zaragoza (ES) - 10-14 April 2007

Advanced use of Artificial Intelligence Techniques on Viticulture for the reduction of the use of Plant Protection Products. Achievement of a Forecasting Model for Plasmopara Viticola Infection (R. Bugliosi, G. Spera, A. La Torre, L. Campoli) XXIX World Congress of Wine and Vine, Logroño, Spain, June 2006

Artificial Intelligence approach with the use of Artificial Neural Networks for the creation of a forecasting model of Plasmopara vitivola infection (R. Bugliosi, G. Spera, A. La Torre, L. Campoli, M. Scaglione) 58th International Symposium on Crop Protection, Gent, Belgium, May 2006