Balancing reliability and energy efficiency in LoRa networks using reinforcement learning

The Long Range (LoRa) are wide area networks that combine long coverage and low energy consumption for the connected end devices. The low energy consumption is largely due to the simplicity and low overhead of Pure Aloha, which LoRa uses as a default medium access protocol. However, serious scalability issues arise when the number of interconnected devices increases, with the main effect being the increase in collisions and the decrease in reliability of transmissions. The Long range wide area network protocol defines two types of transmissions in the uplink, which are the confirmed and the unconfirmed ones, without describing any rules on how to use them. Confirmed transmissions improve reliability but also increase the energy consumption of the end devices, while the exact opposite being the case when using unconfirmed transmissions. In this paper, a reliability balancing using reinforcement learning mechanism is proposed, whose main function is to adjust the reliability of the LoRa network according to the required quality of services provided. This adjustment will be carried out by determining the ratio between the confirmed and unconfirmed transmissions achieving the minimum energy consumption. The effectiveness of the proposed mechanism was evaluated in a series of simulation scenarios, where the proposed mechanism managed to combine the reassurance of the required packet loss rate with the minimum energy consumption in the end devices.