Physics-Guided Machine Learning for Path Loss Prediction: A Review
Abstract
Accurate path loss prediction is essential for the planning and deployment of wireless sensor networks, Internet of Things (IoT) systems, and sixth-generation (6G) networks, particularly in the emerging frequency range 3 (FR3, 7.125–24.25 GHz) band, which has been identified as strategic spectrum for next-generation systems. This article presents a systematic literature review (SLR) that explicitly distinguishes physics-guided machine learning (PGML) from pure machine learning (Pure ML) and hybrid approaches for path loss prediction, with emphasis on the FR3 band. Following the PRISMA 2020 guidelines, 70 studies published between 2018 and 2026 were selected from IEEE Xplore, Scopus, and Web of Science and analyzed across 12 dimensions, including frequency tier, methodology, validation strategy, and prediction performance. The resulting three-category taxonomy comprising Pure ML (66%), Hybrid (26%), and PGML (8%) shows that physics-guided methods, which integrate physical knowledge (e.g., empirical models or propagation constraints) into the learning process, remain underrepresented yet exhibit a narrower range of prediction errors under spatially aware evaluation protocols. The review also identifies a validation gap: 77% of studies rely on random train–test splits, which can underestimate prediction error by up to 3-dB relative to spatial cross-validation (SCV) or leave-one-scenario-out (LOSO) schemes. FR3 is the most underrepresented frequency segment in the reviewed corpus, with only five studies (7%) addressing this band and no coverage of urban macro-cell scenarios for Pure ML or PGML; this finding is interpreted as a coverage gap rather than a basis for definitive FR3-specific performance claims. Based on a structured gap analysis, the review derives a research agenda prioritizing controlled PGML-versus-Pure ML benchmarking, expansion of FR3 measurement campaigns, and the exploration of physics-based loss constraints. These findings provide a strategic foundation for data-efficient and physically grounded network planning in emerging 6G deployments.