Radar-gebaseerde intelligente detectie voor de volgende generatie binnenhuissystemen: van veiligheid tot gezondheid

Modern indoor environments are becoming increasingly populated with wireless devices as the global smart home market rapidly expands (expected to rise from USD 216.3 billion in 2025 to USD 489.2 billion by 2035). In this context, ensuring electromagnetic safety is a critical prerequisite for enabling future wireless power transfer systems. Conventional approaches maintain electromagnetic radiation below safety thresholds at all 2D positions in a room, leading to inefficient power delivery and limited adaptability. This PhD thesis introduces a radar-assisted framework that enables person-aware and intelligent control of indoor electromagnetic exposure while simultaneously extending radar functionality toward contactless health monitoring. The goal is to develop a multifunctional radar-based platform that provides both safe wireless powering and real-time physiological sensing in indoor environments. The first part of this work investigates how a Frequency-Modulated Continuous Wave (FMCW) radar can be integrated into a distributed wireless power transfer system to monitor and mitigate electromagnetic radiation in the presence of people. This approach allows for higher electric field amplitude in areas without the person, thus optimizing overall power utilization within the system. A hybrid estimation model combining simulations and measurements is proposed to estimate the electric field amplitude without explicit channel information. Conservative correction factors are introduced to account for worst-case conditions. Experimental results demonstrate that the system successfully maintains electric field amplitude below a predefined threshold across various 2D locations of persons. Moreover, the experiments show that the system can improve how efficiently the available power is used. This is measured through the power utilization ratio, which represents the ratio between the transmitted power and the maximum available power. A higher value means the system uses the available energy more effectively. In the experiments, the system achieved a power utilization ratio of over 50%. In addition, this approach is extended to multi-person and three-dimensional (3D) environments. Since an advantage of using machine learning is realizing a real-time system, which is crucial for indoor applications, three machine learning models are evaluated, namely random forest, support vector machine, and neural network. Simulation results highlight the superior performance of the neural network model, demonstrating its ability to effectively capture the complex nonlinear characteristics of indoor propagation environments, with only approximately 6% of its predictions exceeding the predefined safety threshold. Experimental results show that neural network-based wireless power transfer can maintain electric field amplitude below a defined threshold for multiple indoor experimental scenarios including various persons' heights. In addition, the proposed approach outperformed the Maximum Ratio Transmission (MRT) approach in terms of Radio Frequency (RF)-RF transmission efficiency in 21.43% of the measurements conducted with multiple people present in the testbed. In the second part, the same radar is used for contactless monitoring of vital signs such as breathing and heart rate. Radar-based vital sign detection already exists, but this work introduces an important improvement. It enables reliable monitoring over a wide range of body angles using compact, low-power hardware. This is achieved through a signal processing pipeline, a localization algorithm, and a beamforming step. The system first determines where each person is located. It then focuses the radar beam in the right direction and extracts the tiny chest movements, even when people sit at different angles from the radar. In our experiments, a commercial Multiple-Input Multiple-Output (MIMO) Millimeter (mm)-wave FMCW radar is employed to monitor vital signs within a range of 1.15-2.3 m and an angular span of -44.8 degree to +44.8 degrees, achieving an average difference of 1.5 and 0.06 beats per minute (BPM) relative to the reference for heart rate and breathing rate, respectively. Furthermore, a machine learning-based approach is proposed to detect abnormal respiration patterns by leveraging data from different radar types. It includes FMCW and ultra-wideband (UWB) sensors, covering diverse target angles, distances, and setups. Utilizing optimized classical feature engineering and hyperparameter tuning, the proposed method classifies normal and abnormal respiratory behaviors. It shows a test accuracy of 98.17%, demonstrating the robustness and adaptability of radar-based vital sign monitoring across varying indoor conditions. The concept is further extended to multi-person monitoring, where additional algorithms are implemented to suppress random body movements and inter-subject interference, allowing estimation of vital signs even in dynamic scenarios. Experimental validation with two subjects across 13 spatial configurations (up to 2.4 m and ±45 degrees) achieved Mean Squared Error (MSE) of 1.88 BPM for respiration and 6.94 BPM for heart rate. The analysis showed that increasing angular separation from 15 degrees to 75 degrees reduced respiration error from 3.73 BPM to 2.06 BPM, while minor motions caused errors below 1 BPM and larger movements up to 2.6 BPM. In summary, this PhD thesis presents a unified radar-assisted system that integrates electromagnetic safety management and health monitoring within an indoor wireless power transfer system. By combining radar sensing, hybrid field estimation, and data-driven intelligence, the proposed framework lays the foundation for safe, adaptive, and health-aware wireless environments. This is a crucial step toward next-generation smart homes and person-aware wireless systems.