Self-Organized Low-Power IoT Networks: A Distributed Learning Approach

Abstract

Enabling large-scale energy-efficient Internet-ofthings (IoT) connectivity is an essential step towards realization of networked society. While legacy wide-area wireless systems are highly dependent on network-side coordination, the level of consumed energy in signaling, as well as the expected increase in the number of IoT devices, makes such centralized approaches infeasible in future. Here, we address this problem by selfcoordination for IoT networks through learning from past communications. To this end, we first study low-complexity distributed learning approaches applicable in IoT communications. Then, we present a learning solution to adapt communication parameters of devices to the environment for maximizing energy efficiency and reliability in data transmissions. Furthermore, leveraging tools from stochastic geometry, we evaluate the performance of proposed distributed learning solution against the centralized coordination. Finally, we analyze the interplay amongst energy efficiency, reliability of communications against noise and interference over data channel, and reliability against adversarial interference over data and feedback channels. The simulation results indicate that compared to the state of the art approaches, both energy efficiency and reliability in IoT communications could be significantly improved using the proposed learning approach. These promising results, which are achieved using lightweight learning, make our solution favorable in many low-cost low-power IoT applications.

Deriving and Updating Optimal Transmission Configurations for Lora Networks

“…. Note that this assumes that an underlying Poisson distribution governs the packet generation process -which is a fairly common assumption in Low-Power Wide Area Networks (LPWAN) networks [4], [12], [13], [15], [18], [19]-. We extend the procedure of [5] by, instead of averaging all the observed inter-arrival times for a node i, considering a moving average that assigns exponentially larger weights to recent samples.…”

LoRa-MAB: A Flexible Simulator for Decentralized Learning Resource Allocation in IoT Networks

“…Recent work on distributed learning for radio resources allocation in LoRaWAN had recourse to Multi-Armed Bandit (MAB) problem [13], [14]. Each end-device is considered as an intelligent agent that chooses a given SF and/or channel to improve the success transmission rate [13] or the reliability and energy-efficiency tradeoff [14], through an adequate reward process. In [13], the authors assumed that all end-devices use the same SF and adopted the stochastic MAB algorithm to determine the frequency selection.…”

“…However, such an assumption is impractical in reality due to the mutual coupling between multiple intelligent end-devices. The work in [14] has presented stochastic and adversarial based distributed learning algorithms for resource allocation in an IoT network. However, the capture effect and inter-SF interference are overlooked.…”

LoRa-MAB: Toward an Intelligent Resource Allocation Approach for LoRaWAN

“…Recent work on distributed selection of radio resources in LoRaWAN had recourse to the Multi-Armed Bandit (MAB) problem [6], [7]. Each end-device is considered as an intelligent agent that chooses a given SF and/or channel to minimize its cumulative regret in comparison with the best fixed allocation that renders the highest reward.…”

“…However, such an assumption is impractical in LoRa network due to the mutual coupling between multiple intelligent end-devices. The work in [7] has explored adversarial MAB for resource allocation in an IoT network. However, the capture effect and inter-SF interference were not taken into consideration.…”