Accurate prediction of radio wave propagation is essential for the design and optimization of future wireless networks, particularly in complex environments such as buildings, factories and urban areas. However, conventional approaches often require extensive and costly measurement campaigns to characterize each new environment. Artificial intelligence (AI) offers promising alternatives, but AI models trained on simulated data do not necessarily perform well in real-world conditions.

The objective of this PhD is to develop AI-based methods capable of accurately predicting site-specific radio propagation using a minimum number of real measurements. The research will investigate how knowledge acquired from simulated environments can be transferred to previously unseen real environments. Three complementary directions will be explored: identifying the most suitable representations of radio propagation for efficient transfer learning; developing intelligent strategies to select the most informative measurement locations; and exploiting additional environmental information, such as images, maps or 3D geometry, to improve prediction accuracy.

The expected results include new measurement-efficient AI methods, a systematic evaluation of their prediction accuracy versus the number of required measurements, and validation on realistic wireless propagation scenarios. Ultimately, the thesis aims to significantly reduce the effort and cost of deploying accurate site-specific radio propagation models.