The Sampling-Assisted Pathloss Radio Map Prediction Competition

Cagkan Yapar, Stefanos Sotirios Bakirtzis, Andra Lutu, Ian Wassell, Jie Zhang, Giuseppe Caire

Keywords: Radio map, pathloss, deep learning, challenge, dataset

Abstract: To encourage further research and facilitate fair comparisons of deep learning–based pathloss estimation methods in indoor environments, particularly in the less-explored case of having access to sparse ground truth pathloss samples in tandem with physical propagation environment information, we organized the MLSP 2025 Sampling-Assisted Pathloss Radio Map Prediction Data Competition. This overview paper describes the sampling-assisted indoor pathloss prediction problem, the datasets used, the competition tasks, and the evaluation methodology. Lastly, it provides an overview of the submitted methods and the results of the challenge.

Submission Number: 9