Enabling Reliable and Scalable LoRa with Collision Resolution and Centralized Radio Access

Numerous Internet of Things (IoT) applications depend on energy-efficient, low-maintenance devices that transmit short messages over long distances. These devices often operate from batteries or harvested energy and must remain connected in locations where dense infrastructure deployment is impractical. Low-power wide-area networks (LPWANs) address these requirements by trading high throughput and low latency for long range, low cost, and low energy consumption. Among LPWAN technologies, LoRa is widely used because its chirp spread spectrum modulation enables robust communication at very low signal-to-noise ratio (SNR), while simple uplink operation keeps end-device complexity and cost low. Despite LoRa's ability to operate at very low SNR, uplink frames near the coverage limit may still be received close to the gateway sensitivity threshold. In this operating regime, noise and interference increase the risk of missed detections and decoding failures, while clock and carrier-frequency inaccuracies in low-cost radios degrade the synchronization and, ultimately, the demodulation performance. Because LoRaWAN uses largely uncoordinated uplink access, the probability of packet collisions increases with the number of active devices. Reception failures and collisions trigger retransmissions, which increase end-device energy consumption and channel load. As IoT adoption continues to accelerate, these effects increasingly threaten the reliability, scalability, and overall performance of LoRaWAN and comparable IoT networks. Given the large installed base of LoRa end nodes, improving reliability and scalability through device modifications or channel-access changes is impractical. In this thesis, we investigate how receiver-side signal processing at gateways and centralized processing across multiple receivers can improve LoRa uplink coverage, reliability, and network scalability while preserving compatibility with existing end nodes. We first establish an interoperable LoRa physical-layer model and software-defined radio implementation that reproduce the low-level conventions required for communication with commercial transceivers and provide a reproducible baseline for LoRa receiver research. On this basis, we address missed detections and synchronization errors through a preamble detection method that is robust to rapid variations in noise and interference power, together with a synchronization algorithm that explicitly accounts for sampling frequency offset. We then treat packet collisions as a structured multi-user reception problem, showing that overlapping LoRa frames can be reliably recovered through successive interference cancellation when combined with iterative soft-decision decoding (SDD). To exploit receiver diversity beyond independent gateway processing and enable a centralized allocation of baseband processing resources, we develop a cloud radio access network (C-RAN) architecture for LPWANs. We implement and validate the proposed software-