Abstract:A Machine Learning (ML) network based on transfer learning and transformer networks is applied to wave propagation models for complex indoor settings. This network is designed to predict signal propagation in environments with a variety of objects, effectively simulating the diverse range of furniture typically found in indoor spaces. We propose Attention U-Net with Efficient Networks as the backbone, to process images encoded with the essential information of the indoor environment. The indoor environment is defined by its fundamental structure, such as the arrangement of walls, windows, and doorways, alongside varying configurations of furniture placement. An innovative algorithm is introduced to generate a 3D environment from a 2D floorplan, which is crucial for efficient collection of data for training. The model is evaluated by comparing the predicted signal coverage map with ray tracing (RT) simulations. The prediction results show a root mean square error of less than 6 dB across all tested scenarios, with significant improvements observed when using a Double U-Net structure compared to a single U-Net model.

What problem does this paper attempt to address?

This paper attempts to solve the problem of accurately predicting millimeter - wave signal propagation in complex indoor environments. Specifically, the paper focuses on how to effectively simulate and predict signal propagation patterns in indoor environments containing various furniture and other obstacles. Traditional methods such as Ray Tracing (RT) are accurate but computationally costly and time - consuming, while empirical models have difficulty dealing with complex and changeable indoor layouts. ### The core challenges of the problem include: 1. **Complex electromagnetic wave - obstacle interactions**: When electromagnetic waves encounter obstacles such as walls and furniture, reflection, refraction and diffraction occur, which make the signal propagation pattern complex. 2. **High computational cost**: Traditional ray - tracing methods require a large amount of computational resources and time, especially when considering diffraction and multiple reflections. 3. **The influence of furniture arrangement**: The presence of furniture has a significant impact on the signal coverage pattern, especially in the high - frequency band (such as 28 GHz). Ignoring the furniture arrangement will lead to large prediction errors. 4. **The need for dynamic environments**: Modern indoor environments (such as smart healthcare, smart homes, etc.) require real - time feedback and optimization, which places higher demands on computational efficiency. ### The method proposed in the paper: To solve the above problems, the paper proposes a machine - learning model based on transfer learning and Transformer networks, especially using the dual - U - Net structure and EfficientNet as the backbone network. This model aims to improve signal propagation prediction in the following ways: - **Data - driven ML model**: Encode the indoor floor plan with furniture arrangement into a matrix and use it as input to output the corresponding radio map. - **Transfer learning**: Use the pre - trained EfficientNet model for fine - tuning to reduce the amount of data and time required for training. - **Dual - U - Net architecture**: The first stage generates a preliminary signal coverage prediction, and the second stage performs refined processing on this basis to improve the prediction accuracy. - **Attention mechanism**: Introduce an Attention Gate to emphasize the areas in the environment that have an important impact on signal propagation. ### Main contributions: 1. Propose a signal propagation model suitable for indoor environments with furniture arrangements, and construct a database containing furniture arrangements and their corresponding RT - simulated coverage. 2. Develop a MATLAB tool for constructing indoor environments that can be used for RT simulation, and share the processed environment and simulation result database. 3. Propose a dual - U - Net architecture with EfficientNet as the backbone. This model performs well in complex indoor environments, especially in signal propagation modeling. Through this method, the paper not only improves the accuracy of signal propagation prediction, but also significantly reduces the computational time and resource requirements, thus providing a more feasible solution for practical applications.

Transfer Learning and Double U-Net Empowered Wave Propagation Model in Complex Indoor Environment

EM DeepRay: an Expedient, Generalizable, and Realistic Data-Driven Indoor Propagation Model

Constructing Adaptive Indoor Radio Maps for Dynamic Wireless Environments

Fast Selection of Indoor Wireless Transmitter Locations With Generalizable Neural Network Propagation Models

A Radio Wave Propagation Modeling Method Based on High-Precision 3-D Mapping in Urban Scenarios

Deep Attention-Based Network Combing Geometric Information for UWB Localization in Complex Indoor Environments

Robust UWB Indoor Localization for NLOS Scenes via Learning Spatial-Temporal Features

A Machine Learning Based Channel Parameter Prediction Method Utilizing Ray Tracing Simulation Dataset

A Modeling Method of Pre-training and Fine-tuning for Non-uniform Indoor Environment

Diffraction and Scattering Aware Radio Map and Environment Reconstruction using Geometry Model-Assisted Deep Learning

Improving Deep Learning-Based UWB LOS/NLOS Identification with Transfer Learning: An Empirical Approach

Spatial Signal Strength Prediction using 3D Maps and Deep Learning

Transfer learning for metamaterial design and simulation

Deep Learning Architecture Based Approach For 2D-Simulation of Microwave Plasma Interaction

Lightweight Hybrid CNN-ELM Model for Multi-building and Multi-floor Classification

Indoor Scene Recognition: An Attention-Based Approach Using Feature Selection-Based Transfer Learning and Deep Liquid State Machine

Transfer Learning for CSI-based Positioning with Multi-environment Meta-learning

Deep Learning for Ultra-Wideband Indoor Positioning

Rf-based fingerprinting for indoor localization: deep transfer learning approach

Non-autoregressive transformer with fine-grained optimization for user-specified indoor layout

Machine Learning-Based 3D Channel Modeling for U2V mmWave Communications