Machine olfaction is a challenging research area that try to build comprehensive solid state systems able to mimic the surprising capabilities (sensibility, versatility, reliability, portability, adaptability, etc.) of the animal (mostly mammalians) olfaction systems. This challenge involves the work of researchers coming from multiple disciplines from Chemistry to Electronics and Pattern recognition.  E-noses, on of the most promising architecture, are actually build by an array of solid state chemical sensors whose responses are sampled and tipycally processed by pattern recognition algorithms in order to qualify (i.e. classify) or quantify  a particular gas mixture (i.e. detrminig components concentrations). Actually, solid state sensors suffer from low stability and low specificity problem, although by means of PR algorithms low specificity can sometimes be turn in an advantage, the stability issue  still retain a great negative performance influence. Application of e-noses now ranges from food industry (e.g wine, oil and cheese origin identification, antifraud, food freshness determination etc.) to environmental analysis. In our labs two different e-nose architectures are developed starting from device design, development and characterization to pattern recognition techniques implementation. The first developed platform was  designed for outdoor environmental monitoring and civil protection applications. Geochemical monitoring of fumaroles emissions in the Solfatara volcane (Campi Flegrei) is, in fact, its primary test field. The second platform (see below) has been designed for overcoming current e-nose platform limitations in indoor scenarios. Most e-nose architecture has been designed for fixed operation with power line availability. Our idea is to build a wireless e-nose platform (w-nose) called TinyNose able to act as a single sensitive node in a wireless network scenario in which multiple nodes can cohoperate in order to reconsruct an "olfactive" image of the sensed environment. The w-nose should actually have a very small dimensional and energy consuming footprint while being self powered by relying on batteries or  a reliable energy harvesting technology. At the moment the w-nose project is in an advanced devolpment phase being equipped with 4 non conductive polymeric sensors connected to a Crossbow commercial mote platform (Telos rev.B). Custom software architecture has been developed to allow sensors driving, data acquisition, and transmission in a mesh shaped topology network. A component interface has been foreseen for allowing on-board pattern recognition algorithms implementation (Neural Networks and SVMs).
In our lab, PR tecniques are also used in an off-line fashion for exploratory data analysis purposes and for the quantification of single components concentration in complex mixtures and harsh environments. One of the most attractive scenario is the use of portable devices for city atmospheric pollution monitoring. Portable devices can, in facts, be used for overcoming limitations of traditional  analyzers  that are characterized by both high dimensional and cost profile that does not allow for a sufficient density of a monitoring grid or for the use in cities historical centres.  Using Support Vector Machines in regression schemes it has been proven possible to exploit solid state chemiresistors responses and air traffic pollutants particular  multivariate distribution to quantify Benzene concentration over one year interval with very interesting performances.  
             Applicative Scenarios:
                    a) Complex mixtures single components concentration quantification: Surface contamination detection for safety and security
                    b) Pollution Monitoring, replacing state of the art air analyzers
                    c) Instrument Fault Accomodation Schemes for traditional air traffic monitoring analyzers
                    d) Geochemical monitoring of fumaroles emissions in Campi Flegrei caldera.

    The pervasive computing revolution, together with multi level sensor fusion advancements, is changing the environment sensing world fostering new scenarios in which distributed sensing is the key for building a more comprehensive image of the sensed environment. Air quality monitoring is a scenario that will greatly benefit from the use of multiple, distributed intelligent sensing units because of the fluidodynamic influences on local concentrations of gas species. In such scenarios, I am investigating the use of a group of multiple, self powered, electronic noses that cooperate for extracting an olfactive “image” of the air quality of a complex area such as an office building or a research centre, thus extending the e-nose concept to a novel wireless nose (w-nose). In particular, such devices could be profitably used for the detection and quantification of VOC released by furniture, cleaning products, solvents etc.  Actually, VOC represents one of the major threats for indoor pollution in houses, offices and other manned working environments both for short and long (worst) term exposure effects.Wireless e-noses to be used in such scenario should have, as common in  wireless sensing,  a low dimensional and power consumption profile in order to allow a seamless integration with the sensed environment and acceptable operative life duration while maintaining sufficient detection and quantification performances. In this sense, room temperature operating polymeric sensors  offer a viable solution in this specific application domain. Furthermore, the computing and communication capabilities of each sensing node should allow the cooperation with the other units in order to let data reach the “data sink” in which the olfactive image would be reconstructed via sensor fusion algorithms.Performing local sensor fusion activities will allow the implementation of local reaction as well as power saving strategies. In facts, data trasmission is the primary power draining source for these architectures, modulating the data sensing and transmission rate with estimated pollutants concentration or simply transmitting data only when a significative event has been detected could allow for extra operative life length not only for the single node but also for each node involved in the transport of information between the original node and the data sink ultimately extending the entire network life.Today e-noses architectures have a very limited suitability for distributed olfactive measurement since they have been typically designed for single point of measure fixed application with power line availability, high temperature operating sensors and often operate analysing the headspace of sample vials. Furthermore their connectivity is typically based on wired interfaces. On the other hand, attention in merging wireless sensor networks and chemical sensors fields is now growing, despite the very different disciplines involved in such projects, as witnessed by a limited number of recent works.ENEA UTTP-MDB is currently working to the development of an actual w-nose called TinyNose whose first functional release was shown at AISEM 2007 Conference. TinyNose, the wireless ENEA Nose, has now reached a prototype phase whose design is primarily aimed on stage decoupling for tuning purposes. His architecture is based on a commercial core mote featuring data acquisition and computing power via a TI MSP4300 microcontroller. Communication capabilities are achieved through a CC240 Zigbee compliant radio. The commercial mote (TelosB by Crossbow) is coupled to a polymeric sensor array  via a signal conditioning stage while hosting a complete set of software components capable to coordinate the acquisition-processing-transmission duty cycle. Signal conditioning stage is based on resistance to voltage conversion obtained through operational amplifier in classic configuration. A power conversion circuit provide power stabilization and adaptation of the 12V battery input. In this prototypal stage, each sensor is mounted on a separate card in which amplification and coupling configuration can be calibrated  via potentiometers and/or choosing from an array of fixed resistance allowing for new sensor development. Sensors base resistance can vary from 100Ω to  500K Ω. As regards as software components, our group developed a comprehensive architecture based on the component programming paradigm and the TinyOS operating system features. This architecture foresee sensor drivers components that take the responsibility to empower data acquisition via the microcontroller ADC and local data processing components that can empower local implementation of sensor fusion algorithms. On the data sink side, a packet forwarder developed in java allow for data inception and forwarding from the wireless sensor network to a GUI that is able to visualize data coming from the different motes recording them on secondary storage units for further analysis, i.e. second level sensor fusion implementations. The software GUI offer also the capability to estimate packet error performances of single links as well as controlling the inner duty cycle of each mote by selecting the data sampling rate. In the previous development stage, TinyNose revealed capable to correctly classify different sources of indoor pollutants, we are now devising to use the present stage prototype in Terpenes detection/concentration estimation problems.

References
1. D.J. Cook, S.K. Das, Smart Environments: Technologies, Protocols, and Applications, John Wiley (New York, 2004).
2. Indoor Air Quality – EPA Information Sheet,  http://www.epa.gov/iaq/voc.html
3. R.Shepherd et al.Monitoring chemical plumes in an environmental sensing chamber with a wireless chemical sensor network, Sensors and Actuators B, 121 (2007) , 142-149
4. C.A. Grimes et al., A sentinel sensor network for hydrogen sensing, Sensors 3 (2003) 69–82.
5. S. De Vito et al. Enabling Distributed VOC Sensing Applications: Toward Tinynose, A Polymeric Wireless E-Nose, Proceedings of XII AISEM Conference, Naples Italy, 2007
6. J. Polastre et Al., Telos: Enabling Ultra-Low Power Wireless Research Proceedings of IPSN/SPOTS, April 25-27, 2005
7. A. De Girolamo Del Mauro et al., “Towards an all polymeric electronic nose: Device fabrication and characterization, electronic control, data analysis”. Proceedings of Transducers&Eurosensors XXI, Lione France, 2007.

    A limited number of thesis, concerning computer science related topics, are permanently available. Available thesis can be  classified in two  main classes:

		Wireless Chemical Sensor Network in Indoor scenarios
		Thesis works in this area will require the analysis/simulation/design/development of  SW Architectures and  protocols (formation, routing, in network query processing etc.) for wireless sensor networks operating in indoor scenarios. Estimated duration : 4 to 8 Months. Basic networking topics knowledge is  suggested as a prerequisite.
		Sensor fusion (Electronic noses)
		Sensor fusion thesis will be based on analysis/development of different approaches (Fuzzy logic, Neural networks, Evolutionary computing, Combination of multiple classifier, etc.) to information fusion typically  applied to Enea solid state gas sensors (Inorganic Gases, Volatile Organic Compounds). This sensors shows poor selectivity performance and hence sensor fusion applied to matrix of slightly different sensors should help to gain far better performance. Estimated duration : 4 to 8 Months. Previously gained basic skills on Pattern Recognition or Knowledge engineering are suggested as a prerequisite.
		Wireless Sensor Networks for Energy Efficiency
		Energy efficiency is of paramount importance for reaching a true sustainable development. A federation of tiny multisensor device could measure appliances real time consumptionin apartments and small offices making data available anywhere, anytime from  PCs, Android based phones and IPhone apps allowing for the increase of energy consumptions awareness . integration with social networking architectures, will empower the adoption of more sustainable lifestyles. Furthermore, this paradigm can be applied to datacenters and smart buildings for obtaining a continuous monitoring of energy consumptions allowing the implementation of HVACs fault prevention and detection as well as accomodation.
		GIS Empowered Distributed Sensing and Decision Support Systems in Utilities
		Efficient water management systems require an in deep knowledge of several parameters across a geographically distributed distribution network. A distributed chemical and ohysical variables monitoring system, through the power of GIS models, could provide an effective way to support water distribution managers in their decision making process.