As the ENACT project progresses, one key task is nearing completion: the development of advanced methods to optimise how environmental sensors are deployed in real-world settings.
This effort addresses a simple question: where should sensors be placed to capture meaningful environmental data as efficiently as possible? The answer, however, requires a sophisticated combination of mathematical modelling, geospatial intelligence, and user-oriented design.
From abstract graphs to real environments
At the core of this work lies a graph-based approach to modelling environments. In this framework, locations are represented as nodes, while the connections between them capture distances, accessibility, or communication constraints.
This structure allows researchers to explore a wide range of placement strategies. Algorithms can systematically evaluate possible configurations and group nearby sensing points into clusters that reduce network congestion.
Simulation models are used to test how different deployment strategies perform under realistic conditions, including varying communication ranges, battery constraints, and potential system failures. The goal is to ensure that environmental data collection remains reliable even in less-than-ideal scenarios.
Two complementary approaches to sensor placement
Building on this foundation, the ENACT team led by South-East Technological University (SETU) has developed two complementary algorithms for sensor placement.
The first is a grid-based approach, which provides a clear and robust baseline. It divides the target area into regular segments and assigns sensor locations accordingly. This method is computationally efficient and easy to interpret, making it a useful reference point.
The second approach uses particle swarm optimisation, a more advanced technique inspired by collective behaviours observed in nature. In this model, multiple candidate solutions evolve iteratively, “learning” from each other to converge towards more optimal configurations. This allows for more flexible and adaptive placement strategies, particularly in complex environments.
Both approaches have been significantly enhanced through integration with OpenStreetMap data to ensure distances are computed along realistic paths, avoiding natural and urban obstacles.
Bridging optimisation and real-world constraints
While SETU’s work focuses on large-scale optimisation strategies, VICOMTECH and RFSAT Limited developed complementary tools that address the sensor placement challenge from different, practice-oriented perspectives.
The system developed by VICOMTECH starts from a common real-world constraint: domain experts often identify multiple points of interest, such as hospitals, schools, or high-risk urban areas, but do not have enough sensors to cover each location individually. The goal, therefore, is to maximise coverage with limited resources.
To achieve this, the system follows a multi-stage workflow. It begins with a clustering-based approximation, then refines sensor locations using graph-based optimisation over real street networks. By leveraging shortest-path algorithms and local search techniques, it minimises the distance between each point of interest and its nearest sensor, while ensuring that all proposed locations are feasible in practice.
The system does not produce a single “optimal” solution. Instead, it generates several alternative configurations that balance performance with practical constraints. This gives decision-makers the flexibility to select the most appropriate deployment based on regulatory, physical, or operational considerations.
In parallel, RFSAT Limited has developed a web-based application that tackles the problem from a broader urban and environmental modelling perspective. Their approach explicitly incorporates terrain, building structures, and road networks as key factors influencing both pollutant dispersion and sensor effectiveness.
The tool evaluates where sensors should be placed by combining several dimensions: ground elevation, which affects how pollutants move; urban morphology, including building footprints and heights that can block or channel airflow; and road network characteristics, as major sources of emissions. At the same time, it ensures that spatial coverage remains sufficiently uniform, avoiding blind spots in monitoring.
Two complementary modes are supported: users can either define a fixed number of sensors or allow the system to determine the optimal number automatically. The application is designed for accessibility and flexibility, running directly in a web browser and supporting both predefined pilot cities and fully custom regions.
From algorithms to a functional demonstrator
These advances are not confined to theory or isolated components. The developed algorithms have been integrated into a dockerised environment with a web-based interface, resulting in a functional demonstrator, expected by June 2026.
This platform will allow users to explore different sensor placement strategies. They will be able to visualise configurations, compare alternatives, and better understand how algorithmic choices translate into real-world deployments.
At this stage, the core functionality is largely in place. The remaining work focuses on fine-tuning algorithm parameters and further integrating complementary systems developed by other partners.
Supporting the broader ENACT vision
Optimising sensor placement is not an isolated technical exercise. It plays a central role in enabling ENACT’s broader ambition: to generate high-quality, spatially meaningful environmental data that can feed into exposomic risk models.
Accurate and reliable sensor networks are essential for understanding how environmental exposures affect health. By improving how these networks are designed and deployed, ENACT strengthens the foundations of its data-driven approach to health risk prediction.