WINSOC PROJECT SUMMARY
Wireless sensor networks are currently receiving huge attention as a basic tool to detect emergency events or monitor physical parameters of interest, such as radiation, pollution, temperatures, pressures, and so on.
The key idea of WINSOC is the development of a totally innovative design methodology, mimicking biological systems, where the high accuracy and reliability of the whole sensor network is achieved through a proper interaction among nearby, low cost, sensors. This local interaction gives rise to distributed detection or estimation schemes, more accurate than that of each single sensor and capable of achieving globally optimal decisions, without the need to send all the collected data to a fusion center. The whole network is hierarchical and composed of two layers: a lower level, composed of the low cost sensors, responsible for gathering information from the environment and producing locally reliable decisions, and an upper level, composed of more sophisticated nodes, whose goal is to convey the information to the control centers.
The key issue is the interaction among nearby, low cost, sensors in a way that increases the overall network reliability, decreases the probability of congestion around the sink nodes, provides scalability and tolerance against breakdown or stand-by of some sensors, and eliminate the necessity for battery recharge. Building on this idea, the consortium has put together expertise from large companies, academies, research centres, end-users and SME's, to create a strong synergism between the academic world, industries and end-users. The goal is, on one side, to develop a general purpose innovative wireless sensor network having distributed processing capabilities and, on the other side, to test applications on environmental risk management where heterogeneous networks, composed of nodes having various degree of complexity and capabilities, are made to work under realistic scenarios. More specifically, the project will address applications to small landslide detection, gas leakage detection and large scale temperature field monitoring.
Scrutinizing the state of the art of the paradigms typically employed in sensor networks, it is possible to recognise a common critical factor: the current paradigms greatly reflect (although scaled and adapted) a well known and consolidated methodological approach borrowed from TLC networks, which however has been developed to cope with totally different requirements and constraints, with respect to a sensor network. The most typical solutions try to adapt classical telecommunication protocols, except for a much greater emphasis on energy-efficient design (see, e.g., ZigBee). However, they still require rather sophisticated network protocols and management overheads in applications where the bit rate required by the sensor network is relatively small and what is really necessary is only to bring the event of interest from the source to the right control node. Typically, the congestion around the sink nodes is only alleviated, but not avoided and the network is not scalable.
In WINSOC, we envisage the development of a very innovative concept of sensor network that represents a significant departure from current proposals. The network is organized in two hierarchical levels. At the low level, there are very simple nodes that gather relevant information and interact with each other to achieve a consensus about the locally observed phenomenon. The interaction occurs through a very simple mechanism that does not require complicated modulation, MAC, or routing strategies. This interaction among the sensors is the key feature, as it improves the reliability of the local decisions and, at the same time, it yields fault tolerance and scalability. The decisions taken locally are then communicated to the upper level nodes that take care of forwarding them to the appropriate control centers.
Building on this fundamental structure, WINSOC has the following primary objectives:
- Develop and test innovative algorithms implementing the self-organization capabilities of the low level sensors and devise the most appropriate radio interface responsible for the interaction among nearby sensors; this technology has a rather broad scope and it is especially useful for environmental monitoring;
- Develop and test three system level simulators addressing the following applications in environmental monitoring:
i) detection or prediction of landslides (according to US government reports, landslides have killed more than 500 people from 1998 to 2001);
ii) detection of gas leakage, to prevent hazard situations or simply avoid unnecessary wastes of energetic resources;
iii) monitoring of temperature fields, as a way to detect fires or, even better, to predict the risk of a potential fire in a given area. The three simulators will incorporate the emulation of the physical environment under test, for the applications mentioned before, the reaction of the network to hazardous events, the performance of the network in terms of reaction time, probability of detection, estimation accuracy, localization, fault tolerance;
- Develop a reduced scale experiment for testing the proposed approach in the case of temperature monitoring and obtain the deployment experience from landslide detection experiments with in-situ monitoring.