From Local Learning to Heatmap-Assisted Coordination in Slotted LoRaWAN

LoRaWAN networks are limited by random-access collisions, fairness degradation, and scarce downlink resources, making frequent ACK-based Reinforcement Learning (RL) difficult to scale. This paper introduces a heatmap-assisted slotted LoRaWAN scheme where end-devices (EDs) perform lightweight RL-based slot learning, while the network server (NS) periodically broadcasts a compact heatmap hint (HM) indicating low-activity slot regions. The HM acts as a global bias rather than an explicit schedule, preserving distributed operation and LoRaWAN compatibility while reducing dependence on frequent feedback. Theoretical analysis as well as simulation and experiment assessment show that the proposed method improves packet reception ratio (PRR) over Aloha, Slotted Aloha, CSMA, and Slotted CSMA. Under sparse feedback, HM-assisted RL maintains performance close to pure RL while requiring $4-5\times$ fewer ACK transmissions, showing that compact NS guidance can improve scalability under strict LoRaWAN downlink constraints.