Synchronisation spatio-temporelle et traitement du signal pour les jumeaux numériques à énergie nulle dans les systèmes de communication par rétrodiffusion ambiante

Indoor localization is a critical challenge for many modern applications such as logistics, Industry 4.0, and smart buildings. Traditional positioning technologies like GPS, Wi-Fi, or Bluetooth require active infrastructure, which is often costly, power-hungry, and difficult to deploy at scale. To address these limitations, this thesis explores a novel indoor localization system based on Zero-Energy Devices (ZEDs) that leverage ambient backscatter. These passive tags operate by modulating ambient 4G signals to transmit their identifiers, without emitting any signal themselves. The first part of this work introduces a localization framework using ZEDs detected by a smartphone. Assuming perfect synchronization, we simulate a realistic office building and demonstrate that detecting the set of tag IDs can enable reliable room-level localization using a simple geometric inference algorithm. Next, we focus on the detection of ZED signals under real-world conditions. We propose a non-coherent detector based on the Neyman-Pearson criterion, optimized to balance detection probability and false alarm. Unlike traditional coherent receivers, our approach does not require phase tracking or full knowledge of the ambient signal. We validate the detector's performance both in simulation and on a hardware testbed, demonstrating robust detection of binary-coded ZED signals under typical interference and noise. We then address the challenges of multi-tag detection, where several ZEDs may transmit simultaneously. Unlike previous studies that only consider single-tag scenarios, our analysis includes signal collisions and proposes detection algorithms for overlapping tag transmissions. We model realistic channel conditions and evaluate the performance in terms of bit error rate, scalability, and detection reliability in multi-tag environments. Finally, we design and evaluate a coding scheme based on short Low-Density Parity-Check (LDPC) codes for tag ID transmission. By mapping each ZED identifier to a codeword, and using joint decoding over multiple bits, our method significantly reduces decoding errors compared to repetition or parity codes, especially in low-SNR regimes. Together, these contributions provide a comprehensive system design for passive indoor localization using ambient backscatter. The thesis addresses the key challenges of synchronization, detection, interference handling, and robust ID decoding. Results show that ZED-based localization is not only feasible but also scalable and energy-efficient, paving the way for future large-scale deployments in smart environments.