Optimizing Collision-Aware ADR for High-Density LoRaWAN Deployments: A Multi-Parameter Tuning Study

Abstract

LPWANs, especially LoRaWAN, are commonly employed in large IoT deployments. The Adaptive Data Rate (ADR) scheme enhances LoRaWAN by modifying SF and transmit power. But ADR usually depends only on the device-specific links and does not consider the effects of collisions or the capture effect at the network layer, making it less effective in dense scenarios. The present paper presents a CA-ADR technique to handle collisions in dense LoRaWAN deployments. In contrast to other ADR schemes, CA-ADR is an algorithm running at the network layer. Parameter selection is formulated as a multi-objective optimisation problem, which minimises cost based on multiple criteria, including collision probabilities and energy consumption. The collision evaluation also incorporates a capture-aware model based on signal-to-interference ratio thresholds, enabling a more realistic assessment of packet reception under overlapping transmissions. The proposed approach is evaluated using an NS-3 LoRaWAN simulation framework. First, Baseline, standard ADR, and default CA-ADR are compared for network sizes from 10 to 500 devices with a 300~s transmission period. Then, 144 CA-ADR weight configurations are explored in the dense 500-device scenario. Results show that CA-ADR provides controllable energy--reliability trade-offs. High-reliability configurations achieve more than 99\% Last 4 PDR with an average transmit power between 13.51 and 13.60~dBm over 10 independent random seeds