Air pollution causes 4.2 million premature deaths annually (source: World Health Organization). 91% of the world’s population lives with poor air quality and 90% of children are exposed to air pollutants daily. This number will rise as cities expand, increasingly impacting developing countries. Air Pollution Sensors as Wireless Sensor Network nodes (AIR-WSNs) are readily available and monitor air quality globally. Modern sensor technology is capable of providing information on air pollution based on: ozone, particulate matter, carbon monoxide, sulfur oxide, or nitrous oxide. The current technology can provide an accurate snapshot of its immediate environment, but depending on sensor placement this information may not be valuable. Current placement techniques of AIR-WSNs are based on statistical understanding of the target area. Our world is dynamic, resulting in air quality migration over time, so this traditional sensor placement method may lead to obsolete data readings negatively impacting decision making.
Dana is an adaptive pollutant sensing network, integrating existing sensor technologies with machine learning to optimize sensor placement for improved accuracy. As Dana learns about its serviced environment, sensor redeployment will be suggested to concentrate sensors to regions where air quality variation is higher, redistributing sensors as required in response to environmental migration.
Dana’s operating principles are:
1. Initial deployment of a large number of sensors to form a fine measurement grid.
2. Environmental learning through data collection and analysis.
3. Redeployment of sensors based on analysis results.
4. Repeat the process in response to environmental changes.
Dana is a future-proof solution for air quality monitoring to adapt to environmental changes on a long-term scale. The network becomes customized and the sensors are placed at locations that best represent the given environment. The aim is to present accessible data that is representative of surroundings using the right number of sensors, in the right places. Our world is constantly changing, and with machine learning, Dana keeps networks relevant by adapting the sensor network to reflect changes in environmental shifts.
WSNs have been established for over a decade, and the associated components (e.g. sensor, communication device, micro-controller, and enclosure) are readily available. The value of Dana arises from the machining learning algorithm. Our team members hold advanced degrees across a wide range of engineering disciplines and have been involved with various technology ventures, thus have the understanding to develop and integrate the software and hardware.
An outline of the algorithm is:
1. Sets of three sensors measure air quality variation in a subset using triangulation.
2. Air quality variation between subsets is measured.
3. Areas of insignificant variation are identified.
4. Suggestions made to relocate subsets such that:
4.1. Subsets in the same area with no variation is replaced with one subset.
4.2. Subsets that have been made obsolete are moved to areas with high variation to increase resolution.
5. Repeat steps until system stabilizes.
We anticipate that after one month sufficient data will be collected to advise the first subset relocation. Subsequent relocation will be notified by the software and depends on the size of the area serviced as well as any transformative environmental changes.
The installation, software, and maintenance costs are comparable to traditional WSNs. A minor cost comes from the occasional sensor relocation but is offset by greatly improved accuracy with allocated sensors.
The deployment cost of the sensors does not differ to any current methods (hardware is the same) which can vary from $500 to $5000 depending on the device. To create a turnkey software solution we need a predicted one year of technical development. This is an estimate of $500k initial capital investment. A copy of the software can be replicated and installed anywhere. Operational costs are the installation of sensor nodes and their subsequent maintenance. This is minimized with the use of a fine AIR-WSNs, and removing or relocating the sensor nodes as instructed by the software. This allows the recycling of sensors to either expand the client’s network or to minimize costs. Software maintenance is done on the air (cloud-based) with minimal expense.
Dana can create impact in the following ways:
1. Lifestyle information for the public. For example:
1.1. Level of exposure to pollutants on their daily commute, and to make an informed decision on route planning.
1.2. Identify areas of good air quality for recreational planning.
2. Feedback for governing bodies for city planning. For example:
2.1. How new infrastructure has affected air quality.
2.2. Identifying areas of high pollution where plants or parks can be established to aid.
2.3. Infrastructure planning to manage air pollution in affected areas.