Motivation

Air pollution is the greatest health risk based on environmental factors. According to the latest statistical projections of the World Health Organisation (WHO), one in eight deaths worldwide is caused by polluted air. In the EU, according to a recently published report by the European Commission, about 400,000 people lose their lives every year due to secondary diseases caused by air pollution - ten times as many as from traffic accidents.

On the other hand, the amount of information available about one’s environment is virtually multiplied by wearable sensors, one’s own senses practically multiplied. This makes it all the more important to derive correct and reliable recommendations for action from the data that can be collected. After all, more data does not necessarily mean more applicable and useful information. The ActOnAir concept combines sensor data obtained from an individual environment with a Big Data analysis and thus measured values and prognoses with medical knowledge.

The specific objectives are:

• Provide low-cost, mobile sensors for comprehensive measurement of air pollutants.
• Development of an integration platform for connecting a wide variety of sensors and for the secure transmission of sensor data to cloud-based analysis modules.
• Provision of customised data mining and forecasting methods to determine individual exposure patterns and provide medically valid recommendations for action in real time.
• A user-friendly and expandable overall solution for individuals consisting of the above-mentioned components for recording their exposure to pollutants and personal symptoms of exposure, calculating individual exposure patterns and providing timely recommendations for action.

Activities

At the start of the project, an overall architecture was developed (see Figure 1), which was also transferred into an initial data flow concept. The architecture shows the 4 main components, of which the i3mainz takes over the area of the integration platform for connecting the most diverse sensors. For this purpose, a system was developed that generates and processes data flows. These data flows consist of input sources, process steps and data stores, which can be flexibly combined into data flows. In the further course of the project, the individual services that are necessary for the processing were developed and tested in several scenarios.

Results

As a result of the project, a prototype was provided which can read in and process different sensor data. The collected data is then analysed and personalised risks are calculated. The integration platform developed by i3mainz uses microservices for processing, which are orchestrated by Spring Cloud Data Flow (see Figure 2). This is done by defining data flows that exchange messages via so-called messaging services and then process them further. The type of services can be divided into “Sources”, “Processors”, “Sinks” and “Task”. The first three are used for data streams and the latter for “batch jobs”. The persistence and standardisation of sensor data is done via OGC standards, such as the Sensor Observation Service (SOS). In the project, such a service was set up using the 52°north framework. Other components developed include:

• API to integrate sensor data from the mobile application: This component allows the storage of sensor data from the mobile smartphone application, for example.
• Data provision to the data mining component: This component acts as a “sink” and allows data to be stored in the data mining component via a database interface. Metadata and sensor information can be considered separately.
• Storage of sensor information in an SOS: This component allows the storage of sensor data via the transactional profile of the Sensor Observation Service. This service has been designed to allow multiple modes of sensor creation. For example, the sensor can be initially registered from fixed defined data when the service is started, or by dynamic generation based on the processed messages.
• Adapters for environmental data services: Adapters for different data and sensor concepts have been developed here. These enable the addition of different weather and environmental factors, such as temperature, particulate matter or ozone. The respective services can be controlled via configuration parameters.
• Data processing: For the internal processing of the data, essential processes have been implemented. This mainly includes the development of services that adapt the data to the necessary structure for further processing. However, a prototype service has also been implemented which can determine an isochrone polygon which could then be used in further sensor data selection.
• Data Fusion Component: In order to link and process the different data into a data set that can be used for data mining, components were developed that are combined into a kind of batch job. These components allow the merging of several measured values via different methods, which can be decisively configured for the individual measurement phenomena.