Autonomous Lightweight Scheduling in LoRa-based Networks Using Reinforcement Learning

The Aloha-based channel access of LoRa-enabled devices is a challenging task due to the high potential for significant packet collisions. This paper proposes a Reinforcement Learning (RL) approach, wherein each end-device (ED) autonomously learns how to transmit data in time slots within a fixed time frame in order to alleviate collisions. The proposed approach offers an autonomous lightweight scheduling method eliminating the gateway's computational requirements for calculating comprehensive schedules. Comparative simulations conducted using the ns-3 network simulator against the Pure and Slotted Aloha approaches demonstrate significant improvements in packet delivery ratio. The results indicate that in a network with 300 EDs and a time frame of 200 seconds, RL approach achieves a delivery ratio of over 95%, showcasing a notable improvement of around 20 percentage points compared to Pure Aloha and 17 percentage points compared to Slotted Aloha.