PathFinder: Advancing Path Loss Prediction for Single-to-Multi-Transmitter Scenario

2025-12-16

AI Summary

Existing deep learning-based path loss prediction methods suffer from three key limitations: passive environmental modeling, reliance on a single-transmitter assumption, and an inherent bias toward in-distribution generalization. To address these challenges for 5G and smart city applications, this paper proposes a novel multi-transmitter path loss prediction paradigm. We introduce an active wireless propagation modeling framework featuring: (i) decoupled building and transmitter feature encoding, (ii) mask-guided low-rank attention, and (iii) transmitter-oriented hybrid data augmentation. Furthermore, we release S2MT-RPP—the first benchmark for single-to-multi-transmitter extrapolation. Experiments demonstrate that our method reduces prediction error by 32.7% in multi-transmitter scenarios and improves robustness under cross-building-density distribution shifts by 41.5%, significantly outperforming state-of-the-art approaches.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Machine Learning: Multi-instance/Multi-view Learning
• Intelligent Robots: Motion and Path Planning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Radio path loss prediction (RPP) is critical for optimizing 5G networks and enabling IoT, smart city, and similar applications. However, current deep learning-based RPP methods lack proactive environmental modeling, struggle with realistic multi-transmitter scenarios, and generalize poorly under distribution shifts, particularly when training/testing environments differ in building density or transmitter configurations. This paper identifies three key issues: (1) passive environmental modeling that overlooks transmitters and key environmental features; (2) overemphasis on single-transmitter scenarios despite real-world multi-transmitter prevalence; (3) excessive focus on in-distribution performance while neglecting distribution shift challenges. To address these, we propose PathFinder, a novel architecture that actively models buildings and transmitters via disentangled feature encoding and integrates Mask-Guided Low-rank Attention to independently focus on receiver and building regions. We also introduce a Transmitter-Oriented Mixup strategy for robust training and a new benchmark, single-to-multi-transmitter RPP (S2MT-RPP), tailored to evaluate extrapolation performance (multi-transmitter testing after single-transmitter training). Experimental results show PathFinder outperforms state-of-the-art methods significantly, especially in challenging multi-transmitter scenarios. Our code and project site are available at: https://emorzz1g.github.io/PathFinder/.

Problem

• Improves radio path loss prediction for multi-transmitter scenarios
• Addresses poor generalization when training and testing environments differ
• Overcomes limitations of passive environmental modeling in current methods

Innovation

• Disentangled feature encoding for active building and transmitter modeling
• Mask-Guided Low-rank Attention focusing on receiver and building regions
• Transmitter-Oriented Mixup strategy for robust training in multi-transmitter scenarios