Environmental Sensing

Wireless sensor networks offer scientists the opportunity to measure and record information about the natural environment at a spatial and temporal scale that has never previously been possible. Despite this, wireless sensor networks have not yet been widely embraced as a scientific tool for environmental monitoring. In this project we are addressing impediments to the wide-spread adoption of wireless sensor networks, including network reliability and lifespan, ease of deployment, total cost of operation, and data integrity. We are also developing tools that will allow scientists to more effectively query and analyse the information available from sensor networks.

Wireless Sensor Networks are gather information from the environment by attaching sensing devices to small, low-power computer systems with digital radio communications. These sensor network nodes self-organise into a network to deliver the data collected, and perhaps processed, by the sensor nodes to a base station, where it can be made available to scientists and other users through the Internet.

Environmental Sensing is a project in the CSIRO Sensors and Sensor Networks Transformational Capability Platform (SSN TCP) intended to:

• increase the range of environments in which Wireless Sensor Networks can be deployed;
• improve techniques for data quality management and the design of sensor networks to optimise data quality/cost tradeoffs; and
• reduce the software engineering costs associated with deploying sensor networks.
• improve the usability of the information made available from sensor networks, through new methods for analysing and presenting information, and for querying and interacting with the network.
  

This work is in pursuit of the Sensors and Sensor Networks TCP goal to "create sensor network technologies that will transform our understanding and management of the environment and resources that underpin Australia's way of life by providing a transformational data-driven approach to scientific discovery."

The project draws on the skills and capabilities of the CSIRO ICT Centre's Autonomous Systems Laboratory, Information Engineering Laboratory, Networking Technologies Laboratory and Tasmanian ICT Centre, and CSIRO Mathematical and Information Sciences.

Wireless Sensor Networks in more environments

The Environmental Sensing project is investigating ways to extend the range of environments in which Wireless Sensor Networks can operate, and increase the lifespan of such networks. Wireless Sensor Networks need to be able to be set up in situations where there is no power and communications infrastructure. Typically, sensor nodes are equipped with a small solar panel and rechargeable batteries to make them independently powered. They carry their own communications infrastructure, to at least deliver data to the base station, where it may be delivered into the Internet via existing land lines or through the mobile phone network.

However, sometimes there isn't enough solar energy available to power a sensor node continuously. This can affect its ability both to collect data, and to forward data from other nodes to the base station In this project we are addressing these challenges by:

• adapting the rate at which data is collected and sent to the amount of energy the node is able to gather from its environment;
• adopting sensor hierarchies so that low-power sensors can trigger the activation of higher-power sensors only when those sensors are likely to capture useful information;
• compressing or extracting features from high-volume data in the node to save power by taking advantage of the relatively low cost of processing on the node compared to radio communication [Read more];
  
• using autonomous vehicles to "visit" sensor nodes and make calibration measurements, so that less expensive and lower power sensors can be used in the nodes;
  
• using connection quality sensitive routing of messages to improve the reliability of the network and reduce the total transmission cost of the messages (for example, by avoiding re-transmission); and
• modelling sensor node energy consumption, predicting the amount of energy that can be harvested from the environment by the node, and adapting the node behaviour so that it continues to function and does not exhaust the energy it has available.
  

in Wireless Sensor networks.

Data quality management and sensor network design

Data quality management is an important part of any system intended to capture scientific information.  The project is developing statistical and operation research techniques to improve the estimation of missing data, quality assurance and quality control of data, real-time detection of anomalous events, and to help design location and sensor complement of the nodes to improve the trade-off between system cost and data quality.

Reducing the software engineering cost of sensor deployment

Sensor networks are difficult to program as they typically have highly constrained resources (especially memory), few debugging facilities and are generally programmed in low-level languages. The program that runs within the sensor nodes is often built specifically for the purpose of that network, and the interface to the data collected is often purpose-built, too.

In this project we aim to reduce the software engineering cost of sensor network deployments by:

• using standards-compliant access to sensor data using Open Geospatial Consortium Sensor Web Enablement standards;
• providing rich computer-interpretable descriptions of the data collected from the sensor networks using the standards developed in the W3C Semantic Sensor Network Incubator Group [Read more];
• contributing to the W3C Semantic Sensor Network Incubator Group;
• using queries over the semantic description of the network and its capabilities as a language for generating the programs to run in sensor networks;
• providing mechanisms to help maintain and manage sensor networks remotely, including:
• light-weight remote procedure calls to control sensor node activities and report on status;
• using queries to collect sensor status data; and
• reprogramming sensor networks over the network
• using standard sensor network operating system software to reduce the cost of maintaining the low-level functions of the network.

Partners and deployments

The techniques being developed in the Environmental Sensing project are being demonstrated in four sensor network deployments. In turn, the challenges presented by these deployment scenarios, and our experiences with the long-term operation of large-scale sensor networks in these environments inform our research directions.

Deployment of a sensor network at Mt Springbrook in South-East Queensland is a collaborative activity between CSIRO and the Queensland Department of Environment and Resources who are interested in monitoring microclimate in a rainforest regeneration area, to both measure the impact of microclimate on regeneration, and to observe the effect on microclimate of the regeneration. The rainforest environment presents particular challenges for a sensor network deployment due to low levels of sunlight (and hence limited power available to the sensor nodes) and conditions of high attenuation for radio communications. Thus the Springbrook deployment is a key test-bed for the technologies being developed in the project for energy management, hierarchical sensing and data compression. [Read more]

A deployment of audio sensors is being use to monitor endangered frog populations in the Somerset Dam catchment near Brisbane. With the support of Seqwater, CSIRO researchers are working to develop networks of acoustic sensors that can recognise, record, and ultimately classify frog vocalisations. This deployment demonstrates the use of hierarchical sensing, remote procedure calls for control of the nodes, and will eventually also incorporate event-based activation of the nodes - whereby audio recording is triggered when the nodes detect rainfall. [Read more]

In another collaboration with Seqwater we have deployed a network of over 200 sensor nodes for monitoring water quality indicators in Lake Wivenhoe, environmental conditions in the surrounding catchment area, and the movement of cattle. This network employs techniques for improved networking using link quality measurements and hierarchical sensing to reduce the use of power-hungry GPS sensors on cow collars. It also involves the use of an autonomous boat to check the physical status of sensor nodes on moored buoys and to make sensor calibration measurements. [Read more]

In collaboration with CSIRO Plant Industry's Phenonet project in we are developing a sensor network to measure the impact of microclimate and soil moisture on the development of test plantings that compare the productivity of grain varieties. This deployment will make use of semantic queries to generate the programs that run in the sensor network to collect and forward the data.

A common standards-based data front end is being developed for use in all of these deployments, as well as other sensor network deployments in CSIRO.