In the agricultural sector, sensor data is highly fragmented and its use is often limited to irrigation control service providers. This restricted view of the information greatly hinders the development of advanced data analytics solutions and the application of technologies for automation and decision support.
Cloudskin proposes a "from ground to cloud" solution that establishes standards and mechanisms for capturing and sharing sensor information in a scalable and secure manner. This includes demonstrating the feasibility of developing advanced geospatial analytics services integrated with this data, including optimizing processing costs in distributed cloud services.
In this environment, we face two main challenges:
On the one hand, the need to overcome reluctance to share data, ensure its security and control over its use, identify its meaning and scale (such as units of measurement), facilitate its search and sharing, and require a platform independent of the origin and frequency of data availability, regardless of whether it comes from sensors or edge information services, or from the private or public cloud.
On the other hand, it is necessary to establish the feasibility of advanced and integrated information analysis, which may involve the use of serverless distributed resources. Furthermore, it is necessary to control costs and maximize processing performance to ensure the development of advanced information services.
Cloudskin addresses these challenges through two experiments:
The first involves developing an agricultural and environmental dataspace in the cloud. This agricultural dataspace ensures data usage and sharing through upload options, dataset usage contract management, customized management using data dictionaries to standardize information semantics, and unified search and dataset management services for projects that feed back into the data space.
The second challenge is addressed by developing a geospatial analytics service integrated into the data space. This service performs advanced data analysis and management, leveraging artificial intelligence models for the predictive estimation of computing resources needed to run these services, which are based on pipelines related to agricultural and geospatial data analysis.
As a result of the project, we will have an agricultural and environmental dataspace that allows for the secure search and sharing of sensor information. We will therefore be able to incorporate all available sensor data from a geographic area, such as the Region of Murcia, after the necessary acceptance of its terms of use.
Furthermore, a cloud service like AWS will integrate this information with data from satellites, analyzing humidity and irrigation information in a unified way with atmospheric data. At last, an automated analysis service, trained with models like XGBoost/Optuna, will determine the optimal configuration of processing resources needed to perform advanced analyses with the lowest resource cost and highest performance.
First, the proposed cloud data space, ideally at a European level, improves the secure use and exploitation of sensor data, especially on small and medium-sized farms. This solution abstracts the data from its technical origin, establishing a communication channel between field sensors and intermediate servers, transforming disconnected data into accessible information.
Furthermore, demonstrating the use of the technology in serverless, distributed environments allows us to showcase the ability to leverage this information and automate decision-making processes. Specifically, our indicators show a 79.9% optimization in terms of speed and a 30% reduction in resource costs, compared to resource consumption without the support of a predictive system using artificial intelligence. These results demonstrate how this project lays the foundation for the development of advanced global information services, paving the way for improved production processes and decision-making in the agricultural and nature conservation sectors.
