ILA-CSMA: Hybrid Sensing and Adaptive Fair Backoff for Large-Scale LoRa Networks.

Dense Long Range (LoRa) networks suffer from packet loss when many end devices contend for the same unlicensed channel. Channel activity detection (CAD) can miss weak or cross-spreading-factor (cross-SF) transmissions, while a uniform carrier sense multiple access with collision avoidance (CSMA/CA) backoff rule ignores the different time-on-air (ToA) costs of SF7-SF12 packets. To address these two coupled problems, this paper proposes an interference-limit-aware CSMA protocol (ILA-CSMA). ILA-CSMA first combines CAD with an instantaneous received signal strength indicator (RSSI) test derived from the residual interference tolerance of the selected spreading factor, and then scales the contention window according to normalized ToA. The protocol is implemented in the Framework for LoRa (FLoRa), an OMNeT++-based LoRa network simulator, and is evaluated for networks with 100-2000 nodes. Compared with Pure ALOHA, Slotted ALOHA, standard CSMA/CA, and two ablation variants, ILA-CSMA improves dense-network access by jointly reducing hidden collisions and airtime imbalance. In the 2000-node case, it increases the packet delivery ratio (PDR) by about 20 percentage points relative to standard CSMA/CA, keeps the Jain fairness index (JFI) above the 0.85 reference line, reduces the energy consumed per successful packet to 22% of the standard CSMA/CA value, and reduces conditional average packet delay from 18.5 s to 8.2 s. These results show that interference-aware sensing and ToA-aware backoff can improve large-scale LoRa access under the evaluated simulation conditions.