Divide and Code: Efficient and Real-time Data Recovery from Corrupted LoRa Frames

Due to power limitations and coexistence in ISM bands, up to 50% of the Long Range (LoRa)-frames are corrupted at low signal strengths (≈ -115dBm) and the built-in redundancy schemes in LoRa-Wide Area Network (LoRaWAN) cannot correct the corrupted bytes. To address this, higher Spreading Factors (SF) are used resulting in wasted energy, increased traffic load, and highly compromised effective data rate. Our on-field experiments showed a high correlation in the corruption of close-by frames. We propose a novel Divide & Code (DC) scheme for LoRaWANs as an alternative to using higher SF. DC pre-encodes LoRa payloads using lightweight and memoryless encoding. After receiving a corrupted frame, DC uses a combination of most probable patterns of errors, Time Thresholds (TT), and splitting of payloads into subgroups for batch processing to recover frames effectively and maintain low complexity and timely operation. By implementing DC on our LoRa-testbed, we show it outperforms vanilla-LoRaWAN and Reed-Solomon codes in decoding and energy consumption. Our schemes decode up to 80.5% of corrupted payloads on SF10 by trying only 0.03% of all patterns of error combinations. TT keeps processing times below 2 ms with only minor reductions in the decoding ratio of corrupted payloads. Finally, we showcase that introducing 30% redundancy with DC results in minimum energy consumption and high decoding ratio at low SNRs.