The understanding of complex environmental phenomena, such as deforestation and epidemics, requires observation at multiple scales. This scale dependency is not handled well by today's rather technical sensor definitions. For instance, to understand the impact of deforestation on the local fauna, it is necessary to track the path of individuals as well as the path of populations within a biotope. Movement patterns of individuals reveal information about change in territory and foraging, while the changed behavior of one population impacts the behavior of others. At the scale of the population, a sensor network should produce a single trajectory based on the tracks of the individual animals. Current definitions of sensors, sensor systems, and sensor networks are too technical to capture these abstractions of observations. For example, the definition of geosensor networks as “distributed ad-hoc wireless networks of sensor-enabled miniature computing platforms that monitors phenomena in geographic space” (Nittel et al., 2004) does not admit animals as sensors and cannot relate the phenomena to those observed at other scales. These definitions also exclude human sensors which are the key to volunteered geographic information. We propose definitions of sensors as information sources at multiple scales, relating physical stimuli to symbol systems. An algebraic formalization shows the aggregations, compositions, and generalizations. It also serves as a basis for defining consistent application programming interfaces to sense the environment at multiple scales of observations and with different devices.

Full paper to be submitted March 31 to http://www.comlab.ox.ac.uk/geosensornetworks/

The Context Toolkit A toolkit for context-aware applicationsAnind K. Dey (dey@cs.berkeley.edu) –> http://www.cs.cmu.edu/~anind/context.html

Modeling sensors as well as sensor systems and sensor networks as a single sensor allows more consistent interface specifications. The Abstract Sensor Interface specifies the basic functionality for interacting with several types, systems and networks of sensors. The developer gets an API (Application Programming Interface) for any sensing system and does not need to know whether the accessed platform is in fact a single sensor or some composition of sensors. Implementations of the Application Programming Interface for a specific sensor type will serve as a simple way of encapsulating the different abstraction levels of the sensor network stack. The API is based on algebraic specifications for the composition of sensors. The work will elaborate design patterns for the composition of sensors and will incorporate these in the abstract interface specifications.

note: interface to our algebra! USE THE SUNSPOTS!

Outlet: Discussion paper for W3C incubator group on geosensors and for OGC

Sensor A sensor is an implemented map from physical stimuli into a symbol system.

Example: Step counter for joggers, blood pressure sensor

Geosensor A geosensor is an implemented map from a physical stimulus into a system of georeferenced symbols.

Example: air temperature sensor, acid rain level sensor, human counting aedes ssp. eggs in a egg trap

Sensor System A sensor system is either an aggregation or a composition of sensors, which can behave like a sensor.

Example: system for monitoring blood pressure and heart frequency

Geosensor System A geosensor system is a sensor system containing at least one geosensor. A geosensor system like a single geosensor.

Example: weather station

Sensor Network A sensor network is a spatially distributed and connected network of sensors.

Example: Company wants to know how many of their vehicles are moving at any point in time.

Geosensor Network A geosensor network is a sensor network whose nodes are geosensors.

Example: Wireless Sensor Network for monitoring air pollution at selected locations

Sensor Web “A Sensor Web refers to web accessible sensor networks and archived sensor data that can be discovered and accessed using standard protocols and application program interfaces (APIs).” (Botts et al. 2007)

Example: Weather Forecasting Sensor Web

Observation an act of observing a property or phenomenon, with the goal of producing an estimate of the value of the property. A specialized event whose result is a data value.

Measurement an observation whose result is a measure

The understanding of complex environmental phenomena, such as deforestation and epidemics, requires observation at multiple scales. This scale dependency is not handled well by today's rather technical sensor definitions. For instance, to understand the impact of deforestation on the local fauna, it is necessary to track the path of individuals as well as the path of populations within a biotope. Movement patterns of individuals reveal information about change in territory and foraging, while the changed behavior of one population impacts the behavior of others. At the scale of the animal population, a sensor network should produce a single trajectory based on the tracks of the individual animals. Current definitions of sensors, sensor systems, and sensor networks are too technical to capture these abstractions of observations. For this reason, a framework which allows to pick sensors, tools, statistical methods, standards and models will be developed to support decisions and understanding in dynamic geospatial applications at multiple scales. The basis for the framework will be an algebraic specification for sensors at different scales. Based on this algebraic specification, different meaningful abstraction levels of sensor data will be identified. To enable the retrieval and composition in or between the different abstraction levels, an ontology of processes and data sources will be elaborated. Additionally, statistical methods supporting the conversion between the different abstraction levels and the assessment of uncertainty will be developed.

Practical outcome: cookbook.

Application of the algebraic view on sensors to the dengue fever (Aedes spp. eggs) use case. This is at the same time our mobility measure: ifgi –> inpe.

In recent years, there have been several outbreaks of dengue fever epedemics in Brazil. To monitor the mosquito agents of the dengue fever, a network of mosquito eggs buckets has been deployed. The algebraic view on sensors depicts a formalism of sensors on different spatial scale levels. In this master thesis, it will be investigated whether this formalism could be applied to the dengue fever use case. The master student will have the possibility to join the team at the Brazilian National Institute for Space Research (INPE) in Brazil to work on his thesis.

Development of suitable tools for different sensor abstraction levels.

This Master thesis aims at the development of tools for the different levels of abstractions which can be applied in sensor network scenarios. These levels include the Sensor, Sensor System, Sensor Network, Sensor Web and Application level. The tools are supposed to facilitate the deployment of sensor networks, ease the management and administration or improve the usage of sensors and their gathered data. The work includes the analysis of an existing sensor network scenario and the utilization of the developed tools in the scenario.

The thesis involves the opportunity to participate in an exchange program with INPE.

At which abstraction level do we need to introduce objects? can we avoid them?

The hypothesis of this work is that objects (like forests or water bodies) are not needed for most change models and can be substituted by observations, surfaces, substances, and media. Avoiding objects as long as possible in the “food chain of sensors” has the advantage of reducing the semantic heterogeneity of models, as object classes vary across information communities. The MSc student will study at least two change questions (e.g., deforestation, glacier melt), work out the chain of sensor abstractions leading to answers, and show at what level object notions need to be introduced, if any.

Ontology (SWRL) of processes and data sources for retrieval and composition in the framework (see proposal)

Selecting sensors manually to observe a complex process is a difficult task. Which sensors should be used? Which temporal resolution should be chosen? Where to place the sensors? Instead deciding manually, one could define an ontology of processes. Such ontology would describe processes in terms of how their are observed. Picking a process from this ontology is like selecting a cookbook for a specific recipe. Besides providing the necessary information on which sensors to chose and how to configure them (the composition step), the ontology would also support retrieval of sensors (the retrieval step) by aligning them to the processes observable by them.

Statistical methods for the conversion between the different abstraction levels of the framework (see proposal).

Different types of data need different statistical methods to be analyzed. For instance, raw sensor data can be analyzed using a given set of techniques, whereas to analyze aggregated, we have to consider this previous data processing. Therefore, once we have identified the meaningful abstraction levels for the sensor data, we propose to identify the proper statistical methods to analyze these data. These methods will be part of a general framework which allows to pick sensors, tools, statistical methods, standards, and models to support decisions and understanding in geospatial applications at multiple scales.

Goal: simplify the sensor web architectures Link to document

Abstract

The understanding of complex environmental phenomena, such as deforestation and epidemics, requires observation at multiple scales. This scale dependency is not handled well by today’s rather technical sensor definitions. For instance, to understand the impact of deforestation on the local fauna, it is necessary to track the path of individuals as well as the path of populations within a biotope. Movement patterns of individuals reveal information about change in territory and foraging, while the changed behavior of one population impacts the behavior of others. At the scale of the population, a sensor network should produce a single trajectory based on the tracks of the individual animals. Current definitions of sensors, sensor systems, and sensor networks are too technical to capture these abstractions of observations. For example, the definition of geosensor networks as “distributed ad-hoc wireless networks of sensor-enabled miniature computing platforms that monitors phenomena in geographic space” (Nittel et al., 2004) does not admit animals as sensors and cannot relate the phenomena to those observed at other scales. These definitions also exclude human sensors which are the key to volunteered geographic information. We propose definitions of sensors as information sources at multiple scales, relating physical stimuli to symbol systems. An algebraic formalization shows the aggregations, compositions, and generalizations. It also serves as a basis for defining consistent application programming interfaces to sense the environment at multiple scales of observations and with different devices.

Introduction

Related Work

Current sensor models and definitions are designed from a technical perspective. In the engineering community, sensors are defined as devices that produce analog signals based on the observed phenomenon. These signals are converted to digital signals by analog-to-digital converters (ADCs). Sensor networks comprise a large number of sensor nodes “that are densely deployed either inside the phenomenon or very close to it” (Akyildiz et al., 2002). From the viewpoint of the Open Geospatial Consortium (OGC)(FOODNOTE: The Open Geospatial Consortium ) a sensor is “a transducer which converts a physical phenomenon into a digital data representation” (Havens et al 2007). Nittel&Stefanidis (2005) introduced the term Geosensor Network by defining a Geosensor Network as a distributed ad-hoc wireless network of sensor-enabled miniature computing platforms that monitors phenomena in geographic space.

The applications of sensor networks range from detection of natural disasters (earthquakes, seaquakes, volcanic eruptions) and non-natural disasters (biological contamination, oil spilling, fires), to habitat monitoring and also include traffic organization and smart environments. The most known projects are the Great Duck Island Project for habitat monitoring (Mainwaring et. al, 2002) and the ZebraNet Project for wildlife monitoring (JUANG et al., 2002)

There are several topics of current research on sensor networks. A lot of work has been done on reducing the in-network communication cost to reduce energy consumption (Reis et al, 2009; Silberstein et al., 2006; ?? MAYBE ADDITIONAL REFERENCES). Other research focusses on developing distributed algorithms to enable the in-network detection of changes or events regarding the observed phenomenon (Worboys&Duckham 2006). Additionally, several work has been done to improve the localization of the sensor nodes (Reichenbach et al. 2008).

Recently, Goodchild (2007) proposed to extend geosensor networks to include humans either as sensor platforms or as sensors themselves. These human sensor networks could serve as the basis for the Volunteered Geographic Information (VGI) enabled by Web 2.0 technologies. An example in this context is the birdpost application (www.birdpost.com) which enables its users to report bird sightings or to search for birds sightings by location or characteristics.

To enable the webbased exchange of geosensor data and the integration of sensor data into spatial data infrastructures, OGC's Sensor Web Enablement (SWE) initiative provides a framework of standards for the realization of the Sensor Web. Following Botts et al. (2007) a Sensor Web refers to web accessible sensor networks and archived sensor data that can be discovered and accessed using standard protocols and application program interfaces (APIs).

Research on the Semantic Sensor Web (Sheth et al., 2008) investigates the role of semantic annotation, ontologies, and reasoning to improve discovery on the Sensor Web. It combines OGC's vision of a web of sensors with the reasoning capabilities of the semantic Web. Besides discovery, a semantic layer would improve interoperability between sensor networks and help to make sensors situation aware. An ontological analysis of the OGC standards on observation and measurements has been done by Probst (2006). (TODO ask anu for input) However, the integration of semantics into sensor networks and sensor applications is still a challeging research task and a thoroughly defined model for sensors from an information perspective is currently missing.

Use Case

Main Use Case: Wildlife tracking to monitor the impact of deforestation on the local fauna

To illustrate and clarify the definitions and algebraic operators which will be developed below we introduce an exemplary scenario of a sensor network use case. We consider a wildlife tracking system which has been deployed as a real world scenario several times in the past (Sheppard et al., 2006; Ferguson and Elkie, 2004; Deutsch et al., 2003 ; Walker et al., 1992). Tracking sensors are attached to the individuals of an animal population to record their paths through the local habitat. Information which can be gained from such sensor network deployments are fundamental to understand the foraging behaviour of free-ranging animals. Besides general biological and environmental research, the information allows policy makers to manage land resources in accordance with endangered species. Without such information, hypotheses may be falsely accepted and result in poor land-use decisions (Hulbert 2001). We assume that the example use case is similarly designed as the existing ZebraNet project (Zhang et al., 2004). Each animal of the monitored population carries a small sensor platform consisting of a global positionig system (GPS) and a wireless transceiver to establish communication between the platforms. The animals are considered as the nodes of the sensor network and propagate gathered data from platform to platform to finally forward it to a mobile base station which accompanies the population. Besides positioning sensors also other sensor types may be attached to single animals. These additional sensors could measure variables like heart frequency or blood pressure. But also environmental data, e.g. temperature or luminosity, could be gathered sensors attached to the animals.

ZebraNet Link: [1] http://www.princeton.edu/%7Emrm/zebranet.html

Example: Step counter for joggers, blood pressure sensor

Example: air temperature sensor, acid rain level sensor, human counting Aedes spp. eggs in egg traps (??? Monteiro et al. or just the link http://www.dpi.inpe.br/saudavel/documentos/ArtigoSAUDAVELAgo2004.pdf ???)

Example: system for monitoring blood pressure and heart frequency

Example: Weather Forecasting Sensor Web

Definitions (find a better name; combine with axiomatization)

In this section…

To include technical as well as animal sensors, we state that a sensor is a map from the domain of physical stimuli into a symbol system. Examples are technical blood pressure sensors or the human somatosensory system. Geosensors are a specific kind of sensors where all symbols are georeferenced.

Symbol

Georeferenced_Symbol init:: time x space x symbol –> georeferenced_symbol

Sensor observe:: Physical_stimuli –> symbol

Geosensor observe:: stimuli –> georeferenced_symbol

class ( ) SYMBOL is there a Haskell GROUP class we can use here?

class (SYMBOL geosymbol, STR str) ⇒ GEOSYMBOL geosymbol

– A sensor is a map from physical stimuli into a symbol system – A sensor is modelled as a class of functions (not objects) – Example: Step counter for joggers, blood pressure sensor, but also all sensing systems

class (STIMULUS stimulus, SYMBOL symbol) ⇒ SENSOR stimuli → symbol where

  observe :: [stimulus] -> symbol

– A geosensor is an implemented map from a physical stimulus into a system of georeferenced symbols. – Example: air temperature sensor, acid rain level sensor, human counting dengue eggs in bucket

class (SENSOR geosensor, GEOSYMBOL symbol) ⇒ GEOSENSOR geosensor where

  observe :: [stimulus] -> symbol

– space :: geosensor → space – time :: geosensor –> time – where = getSpace – when = getTime

class (TIME time, SPACE space) ⇒ STR time space where

  getTime :: str -> time
  getSpace :: str -> space
  
  

class (SENSOR sensor) ⇒ SENSORSYSTEM sensorsystem sensor where

   addSensor :: sensor -> sensorsystem -> sensorsystem
   removeSensor :: sensorsystem -> sensor -> sensorsystem
   getSensors :: sensorsystem -> [sensor]
   

class (SENSORSYSTEM homog_sensorsystem) ⇒ HOMOG_SENSORSYSTEM homog_sensorsystem where – implementations of addSensor need to guarantee same sensor (or can this be done here?)

instance SENSORSYSTEM

class () ⇒ GEOSENSORSYSTEM

class () ⇒ SENSORNETWORK

class () ⇒ GEOSENSORNETWORK

Geosensor A geosensor is an implemented map from a physical stimulus into a system of georeferenced symbols.

Example: Tracking sensor of animal

Sensor System A sensor system is either an aggregation or a composition of sensors, which can behave like a sensor.

Geosensor System A geosensor system is a sensor system containing at least one geosensor. A geosensor system like a single geosensor.

Sensor Network A sensor network is a spatially distributed and connected network of sensors.

Geosensor Network A geosensor network is a sensor network whose nodes are geosensors.

Sensor Web “A Sensor Web refers to web accessible sensor networks and archived sensor data that can be discovered and accessed using standard protocols and application program interfaces (APIs).” (Botts et al. 2007)

Observation an act of observing a property or phenomenon, with the goal of producing an estimate of the value of the property. A specialized event whose result is a data value.

Measurement an observation whose result is a measure

Algebra

Conclusions and Further Work

References

Akyildiz, I. F., Su, W., Sankarasubramaniam, Y., and Cayirci, E.: Wireless Sensor Networks: a Survey. Computer Networks, vol. 38, 393–422 (2002).

Deutsch, C. J., Reid, J. P., Bonde, R. K., Easton, D. E., Kochman, H. I., and O'shea, T. J.: Seasonal Movements, Migratory Behavior, and Site Fidelity of West Indian Manatees along the Atlantic Coast of the United States. Wildlife Monographs. 151, 1-77 (2003).

Ferguson, S. H. and Elkie, P. C.: Seasonal movement patterns of woodland caribou (Rangifer tarandus caribou). J. Zool., Lond.,262, 125-134 (2004).

Havens et al (2007)

Hulbert, I.A. and J. French (2001): The accuracy of GPS for wildlife telemetry and habitat mapping

Juang, P., Oki, H., Wang, Y., Martonosi, M., Peh, L., and Rubenstein, D., “Energy-Efficient Computing for Wildlife Tracking: Design Tradeoffs and Early Experiences with ZebraNet.,” in International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS-X), 10., 2002, pp. 96 – 107.

Mainwaring, A., Polastre, J., Szewczyk, R., Culler, D., and Anderson, J., “Wireless Sensor Networks for Habitat Monitoring,” in ACM International Workshop on Wireless Sensor Networks and Applications, 2002, pp. 88 – 97.

Monteiro, A. M. V. et al., “SAUDAVEL: Bridging the Gap between Research and Services in Public Health Operational Programs by Multi-Institutional Networking Development and Use of Spatial Information Technology Innovative Tools”, Document Prepared for the first General Assessment of the MCT PD&I-TI Program (http://www.dpi.inpe.br/saudavel/documentos/ArtigoSAUDAVELAgo2004.pdf)

Nittel, S. and Stefanidis, A.: GeoSensor Networks and Virtual GeoReality. In: GeoSensors Networks, S. Nittel, Stefanidis, A., Ed.: CRC Press, 2005, pp. 296.

Reis, I. A., Camara, G., Assunção, R. M., and Monteiro, A. M. V., “Suppressing temporal data in sensor networks using a scheme robust to aberrant readings,” International Journal of Distributed Sensor Networks, to be published, 2009.

Amit Sheth, Cory Henson, and Satya Sahoo, “Semantic Sensor Web,” IEEE Internet Computing, July/August 2008, p.78-83.

Sheppard, J. K., Preen, A. R., Marsh, H., Lawler, I. R., Whiting, S. D., and Jones, R. E., “Movement heterogeneity of dugongs, Dugong dugon (Müller),over large spatial scales,” Journal of Experimental Marine Biology and Ecology, vol. 334, pp. 64 – 83, 2006.

Silberstein, A., Braynard, R., and Yang, J., “Constraint Chaining: On Energy-Efficient Continuous Monitoring in Sensor Networks,” in SIGMOD 2006, 2006, pp.

Walker, M. M., Kirschvink, J. L., Ahmed, G., and Dizon, A. E., “Evidence that Fin Whales Respond to the Geomagnetic Field during Migration,” J. exp. Biol., vol. 171, pp. 67– 78, 1992.

WORBOYS, M.F., DUCKHAM, M., 2006. “Monitoring qualitative spatiotemporal change for geosensor networks”. International Journal of Geographical Information Science, v20 n10, 1087-1108.

Zhang, P., C.M. Sadler, S. A. Lyon, and M. Martonosi (2004): Hardware Design Experiences in ZebraNet