The main objective of SemsorGrid4Env is to specify, design, implement, evaluate and deploy a service-oriented architecture and middleware which allows application developers to build open large-scale semantic-based sensor network applications for environmental management. Such architecture and middleware will enable the rapid development of thin applications (e.g., mashups) that require real-world real-time data coming from heterogeneous sensor networks, making it possible to use sensors for other environmental management purposes than those that they were originally expected to have (hence reducing sensor network deployment costs) and to combine their real-time data with historical data from other data sources, opening possibilities of improving current decision-making procedures in a variety of situations (emergencies, monitoring, etc.).
The following clips provide a rapid visual introduction to the key partners and activates that have gone into the SemsorGrid4Env project:
by David DeRoure
(University of Southampton & University of Oxford): SemSorGrid4Env recognises existing sensor network deployments and brings them together with diverse data sets and applies them in the real world. Its task is to take data through from the sensor to the understanding and decisions of real-world expert users. In seeking to cope with this deluge of incoming data, SemSorGrid4Env takes a novel approach to demonstrating that it is possible to integrate established techniques and data (sensor and other). To this end, a service-oriented architecture is employed, comprising back-office storage and data management with front office user interaction with a web-based interface. This makes it easy for users to use the system and for developers to develop new systems. Linked Data make sensor network data accessible to and reusable by anyone - and the SemSorGrid4Env approach can cope even when this reuse is unanticipated. In a world with many data sources, finding the right data is critical, and the semantic registry approach adopted by SemSorGrid4Env (led by the Athens partner) allows this by handling descriptions of sensors and networks. The project runs "in the wild", meaning in the real world of network deployment and the plethora of existing systems and data sources - thus SemSorGrid4Env fits in with existing standards and practices and avoids being isolated. It is part of a global community through which to establish and share best practice.
Registry - The Athens Contribution
by Manolis Koubarakis
(National and Kapodistrian University of Athens): The Athens contribution covered two themes. The first was analysis techniques for low-level data from sensors - for example, to detect sensor malfunction or produce forecast values for the next hour. This work, undertaken with University of Manchester, implemented data analysis in the query execution stack. The University of Athens took the network execution engine provided by its partner and enhanced its functionality to offer forecasting and outlier detection mechanisms. This work was undertaken using test sensors which possessed significant computing and analytical power. The second theme at Athens was registry-based and developed a semantic web data model for capturing information about sensors, such as their location and properties, together with a database management system for managing and querying such information efficiently. Given this data model and database management system, it was possible to build the registry services needed by the SemSorGrid4Env architecture. There were two implementations of the semantic registry, one centralised and one distributed. The project (University of Madrid) also developed an ontology supporting service descriptions such as a web service with data sets covering specific regions, each with a spatial extent defined by a geometry. Thus every service provider could use these ontologies to provide web service descriptions and store them in the registry where they could be discovered by prospective clients. This technology can also play other roles such as refinement operations on hotspot products produced by the National Observatory of Athens. They use remote sensing techniques to support emergency response management. Every 5 minutes an image is created to depict disaster events such as forest fire or flooding across the whole Greek territory. SemSorGrid4Env applications are now used to improve the outcomes of this data processing. One of the strengths of SemSorGrid4Env is this use of real-world examples. The approach is also very scalable and can handle large data volumes which make it useful in practice.
Architecting the System
by Alvaro Fernandes
(University of Manchester): Manchester brought two main strands to the work of SemSorGrid4Env. Firstly, the architecture of the project combined in a novel and interesting way the two worlds of service-oriented architecture and more agile work on the web. Secondly, previous Manchester work on sensor network pre-processing was advanced to reduce the development and maintenance cost and deployment time of sensor networks. Rather than programming the network, the users' interest in data is expressed in the form of a query as if accessing a database. The sensor network then becomes in essence a distributed database made up of many individual nodes, and the query mechanism offers a well-established and well-understood mechanism for integrating resources. The architectural work thus focussed on connecting sources to applications using mechanisms that lie between these two layers (registry and data integration technology). The architecture has a service-oriented structure in the bottom layer and a user-facing resource-oriented application layer at the top. At the bottom are the concrete resources (sensor networks, databases etc.), and their descriptions are stored in the registry structure developed for SemSorGrid4Env by University of Athens. Applications can then discover accessible data sources from the registry. However, the applications might not understand the sources because they describe their data in one form and the application expects a description in another form. To tackle this, a semantic integrator was developed to transform and combine data from multiple sources using an ontology that users could understand. The top layer of the architecture carries application-oriented user-facing services such as the high-level API developed at University of Southampton. Thus SemSorGrid4Env combines very high-level concerns with very low-level artifacts, prime amongst which are battery-powered sensor nodes with on-board processing power. These are electronics boards including sensors and radio communications. The demonstration system for SemSorGrid4Env used one node as a base station and others to supply data. A query to the nodes computes average sensor values in near real time. Once the network is running, a drop in one sensor value translates instantly into a drop in the displayed average. A great advantage of this approach is that it reduces the time and cost of network deployment. A network can be implemented remotely from the user's computer in a very agile manner. SemSorGrid4Env has thus significantly advanced the process of connecting low-level technologies all the way up - with levels of abstraction that allow users to interact with information comfortably in a coherent and principled way.
HLAPI - The Southampton Contribution
by Kirk Martinez
(University of Southampton): In SemSorGrid4Env, linked data and web data combine to allow people and computers to find environmental information. For example, people locate weather data by web browsing using keywords. A computer finds this very difficult, and thus needs specialised techniques to know what is in a data repository and extract the right information. People sift and select information very effectively, but a computer needs specific software to achieve the same selective access to information. SemSorGrid4Env takes each data source and wraps it in a description that computers can understand. For this to happen, people worldwide should ideally use a standard language and description to describe things in an agreed way. This has traditionally been approached by using metadata, but even so in practice the required global standardisation is unachievable. So the semantic web approach employs common groups of descriptions which in effect link together different ways of describing an object - and this approach is based on ontologies. These are fixed and inter-linked groups of descriptors that can be applied across an information domain. For SemSorGrid4Env , specific existing ontologies have been chosen and used to mark up data sets semantically so that computers can find them and reason about them automatically far faster than a human web search. One emergent technology that can be managed in this way is wireless sensor networks, which are cheap but semi-intelligent data sources that can communicate between themselves and make decision about data. SemSorGrid4Env has experimented with integrating such networks into the information infrastructure. Mashups are then used to permit one data source to feed many different uses and applications.
Supplying the Data - Sensor Network Management
by Samantha Roe
(EMU - Marine Development, Research and Planning): EMU's key contribution to SemSorGrid4Env is in collecting data in the marine environment - with all the physical and logistical challenges that entails. Great attention was paid to the physical side of sensor deployment as well as the networking side. The principal aim of SemSorGrid4Env was to create an IT infrastructure that facilitates rapid integration of different types of data. A project demonstrator was developed at University of Southampton that represents these data types in one application - modelled data, real-time sensor data and historical data sets. Data are transmitted live from offshore wave buoys every second to a shore station, and every 30 minutes average wave activity at that site is calculated. These data are then forwarded to the Channel Coastal Observatory (UK government-funded repository) and displayed on a public web site. EMU worked with CCO to harness these rapid 1-second updates from wave buoys and add them to the web site as well as capturing them as a data stream for the SemSorGrid4Env demonstrator. In addition, EMU customised a data collection campaign for SemSorGrid4Env, deploying sea-bed sensors to measure waves, tides and currents, and these data were used as model calibration points. EMU currently monitors incoming data streams from 51 UK coastal locations, and constantly quality-controls these real-time data. On the quay wall, wave-overtopping sensors were also deployed for SemSorGrid4Env, which communicate to each other and back to a base station. These data could feed an early-warning system. Consultation with user groups early in the SemSorGrid4Env project made it clear that using real-time sensor data to increase the accuracy of modelled and predicted information is of great interest to the user community.
ABP - Using the data
by William Heaps
(ABP Southampton - Associated British Ports): ABP is the largest port operator in the UK, with 21 ports around the country. Its aim is to bring ships into port safely and efficiently, which involves many practical skills including the traditional seamanship skills of the ship pilots. Increasingly, however, there is a need for information gathered from many sources that must be assimilated to plan the voyage. ABP provides several services, notable amongst which is the Vessel Traffic System (VTS), which is similar to an Air Traffic Control function. The VTS provides information to approaching vessels and to their pilots. The data needed by pilots are wide-ranging and increasing - for example, weather, ship positions and tides. As well as being the Harbour Authority for safe navigation, ABP has wider responsibilities for harbour incidents such as pollution or even collisions. When such events happen, the need for data suddenly increases rapidly - often involving data that are not used on a day-to-day basis. During incidents, data will be needed from other data providers such as the Channel Coastal Observatory (meteorological and tidal data), the Environment Agency (river flows) and in the case of pollution, the Maritime Coastguard Agency (tide and current models). All this information will be brought together and processed in the VTS to support incident management and decision-making. ABP Southampton has found that the amount of data to be collected and rapidly assimilated is increasing, and the issue now is not finding the data but processing it conveniently and rapidly enough to support decision-making. It was this problem that brought ABP Southampton to SemSorGrid4Env, which offered a potential solution to these challenges. The work focussed on pilotage as a driver for the SemSorGrid4Env project demonstrator. ABP Southampton was particularly attracted to SemSorGrid4Env because it seemed to offer a research-based answer to questions that were already being asked.
Spreading the Word - SemSorGrid4Env at Santander
by Sarah Watt
(EMU - Marine Development, Research and Planning) and Oscar Corcho
(Universidad Politécnica de Madrid): After several years of work on SemSorGrid4Env, the project was taken to a user-focussed international oceanographic conference and exhibition in Santander, Spain. The data demonstrator, ontologies and technologies developed by the project were presented to a range of potential users. The project demonstrator includes a flood inundation model to provide flood forecasts and combine these with other data layers including flood defences. Thus far, the application is similar to a Geographic Information System (GIS) tool, but the benefit of SemSorGrid4Env is to overlay real-time data on top of the modelled and database data. These include real-time ship positions, road traffic, time-based population distribution, and sensor-derived waves and tides. Combining all of this into a single application is the real power and benefit of SemSorGrid4Env for the user. Now that more data sources are being opened up and made more publically available, much greater added value can be achieved than before. Taking the SemSorGrid4Env project to Santander was a very useful basis for discussions with many potential users to alert them to the power of model data integration with real-time sensor data. These discussions also produced feedback on the project and the demonstrator, and identified possibilities for future funding and development of the work so as to take the SemSorGrid4Env concepts on beyond the end of the project.
Reflections on SemSorGrid4Env - The Aims of the Project
by Oscar Corcho
(Universidad Politécnica de Madrid). The data produced by public administrations and companies are useless unless they are integrated meaningfully. SemSorGrid4Env aims to combine sensor and database outputs into a single information space. The project involves universities from Spain, Greece and the UK, together with companies who supply and use the resulting data. The key aim is to enable developers to create tools that allow users to find diverse data and draw them together. In order to achieve effective integration, SemSorGrid4Env has developed a high-level API that can be understood easily by developers, and allows them to access data sources. Ontologies are then used to achieve the data integration. These ontologies need to be understood by both developers and users, and the SemSorGrid4Env knowledge architecture operates at both a general and a subject-specific level. The project has successfully integrated technologies that can be used by any third party developer interested in creating semantic sensor applications, and has implemented this in a flood and coastal demonstrator application. Some of the results can be framed in the context of larger government open data initiatives. Lessons learned about putting sensor data streams onto the web can now be extended to other countries. The project has been visionary in promoting the idea of the semantic sensor web, and has addressed many of the issues that arise when integrating different technologies and data sources. As a result, semantic web sensor applications can now be produced more rapidly and more easily.