Mathematician Nazanin Abedini focused her research on the process of data assimilation: the intelligent combination of mathematical models and real-world measurements to produce the most accurate possible picture of reality. In many sectors, from weather forecasting to technical systems, predictions are based on models that are never perfect in practice. At the same time, sensor data is often susceptible to noise, incomplete or irregularly available. Abedini compares the situation to a LEGO set with a partially missing manual: one cannot rely solely on the instructions, but must also look at the available pieces and other examples to arrive at the correct construction.

The principal finding of the research is that predictions can be surprisingly robust, even when data is flawed. Systems prove capable of tracking reality accurately, even when data contains noise or partial gaps. Combining multiple data sources can further improve accuracy, provided this is done in the right way. A key factor is the frequency with which switching between data sources occurs: this largely determines how stable and reliable the predictions are. A further notable finding is that it appears possible to estimate in advance how accurate a prediction will be.

The findings carry broad societal relevance. In many modern applications, decisions must be made on the basis of incomplete information, such as autonomous vehicles continuously analysing their surroundings, energy networks balancing supply and demand, wireless sensor networks in smart cities and systems for climate and environmental monitoring. The way sensors are deployed, and how frequently switching between different information sources occurs, has a direct impact on the quality of predictions. This is particularly important in systems where data is not continuously available, for example due to limitations in battery capacity or bandwidth.

For engineers, the results provide concrete guidelines. Switching between data sources too infrequently can reduce accuracy, while a well-chosen switching frequency leads to more stable and higher-quality predictions. In practical applications such as environmental monitoring with battery-powered sensors, the findings can be used to determine how often sensors need to be active in order to save energy while still delivering reliable information. The insights are directly applicable in algorithms for robotics, smart infrastructure and climate models. Nazanin Abedini will defend her doctoral thesis on 7 May at Vrije Universiteit Amsterdam.