CoRa: A Collision-Resistant LoRa Symbol Detector of Low Complexity

TL;DR

CoRa tackles LoRa symbol decoding under severe collisions by introducing a Bayesian symbol detector that relies on two waveform-based features, PMD and HPD, extracted from dechirped symbols. The method does not require peak detection or precise symbol boundary information, and it integrates with existing synchronization and decoding blocks from a state-of-the-art receiver. Evaluations on real-world CIC and TnB datasets and LTE-ETU simulations show CoRa achieving up to 29% higher decoding performance than TnB and up to 11.53x throughput over a baseline, while maintaining a modest $O(N\log N)$ complexity with a roughly $3\times$ overhead. This demonstrates a practical, collision-tolerant LoRa demodulation approach suitable for deployment in dense IoT networks and mobile environments.

Abstract

Long range communication with LoRa has become popular as it avoids the complexity of multi-hop communication at low cost and low energy consumption. LoRa is openly accessible, but its packets are particularly vulnerable to collisions due to long time on air in a shared band. This degrades communication performance. Existing techniques for demodulating LoRa symbols under collisions face challenges such as high computational complexity, reliance on accurate symbol boundary information, or error-prone peak detection methods. In this paper, we introduce CoRa , a symbol detector for demodulating LoRa symbols under severe collisions. CoRa employs a Bayesian classifier to accurately identify the true symbol amidst interference from other LoRa transmissions, leveraging empirically derived features from raw symbol data. Evaluations using real-world and simulated packet traces demonstrate that CoRa clearly outperforms the related state-of-the-art, i.e., up to 29% better decoding performance than TnB and 178% better than CIC. Compared to the LoRa baseline demodulator, CoRa magnifies the packet reception rate by up to 11.53x. CoRa offers a significant reduction in computational complexity compared to existing solutions by only adding a constant overhead to the baseline demodulator, while also eliminating the need for peak detection and accurately identifying colliding frames.

Figures (9)

• Figure 1: Spectrogram of a LoRa symbols (top), downchirp signal (middle) and dechirped symbols used for demodulation (bottom).
• Figure 2: Illustration of a LoRa symbol subjected to collision, showing both the spectrogram (left) and frequency spectrum (right). The baseline LoRa demodulator selects the peak at $f_1$, in contrast to the true peak location in $f_m$.
• Figure 3: System architecture of $\text{CoRa}$ . The blue box highlights the symbol detector block, which is our main contribution
• Figure 4: Derivation of the Half-Period Discriminator (HPD), showing the spectrum of a dechirped symbol (X), the spectrum of the signal with a phase inversion in the second half (Y), the masked phase-inverted signal (Z) and the output of the HPD feature. The figure shows the location of the interference peaks (orange markers) as well as the true peak (green marker)
• Figure 5: Discrete (200x200 samples) posterior probability $P\left[C_{k}\left(\mathbf{x}\right) \mid \mathbf{ph}_{k}\left(\mathbf{x}\right)\right]$, as a function of PMD ($p_k$) and HPD ($h_k$) features, estimated from simulated data. This probability is subsequently used to compute the Bayesian classifier $P\left(T_{k}\left(\mathbf{x}\right)\right)$.