(A) study on scalability enhancement for internet of things in LoRa networks

Abstract

The demand for Internet of Things (IoT) services such as smart instrumentation, wearable devices, parking systems, sharing systems, office or factory monitoring, industrial IoT, and etc is rapidly increasing with the wide adoption in the world. As a number of IoT devices are connected to a network with an increase in services, IoT devices are expected to increase rapidly around the world. As the interest in the IoT increased, network technologies for the IoT have been developed. Existing cellular networks or wireless network technologies have high bandwidth and consume a lot of power, so they are not suitable for IoT systems that accommodate a large number of devices and require low power consumption. In order to solve this problem and accommodate IoT services, low-power wide-area network technologies such as LoRa, Sigfox, and NB-IoT have been developed. LoRa developed by Semtech, which is a founding member of the LoRa Alliance, is one of the most promising LPWAN technologies that provide higher data rate than the one provided by SigFox, and longer-range connectivity than the one of NB-IoT. LoRa is a technology that operates in an unlicensed band, and as the number of devices increases, the scalability of the network significantly decreases. In order to solve this problem, a study is conducted to improve the scalability by considering the physical characteristics of LoRa. In addition, the data collected in the IoT environment is measured too frequently and there is a large amount of redundant data, causing a data redundancy problem. This problem not only causes network performance degradation, but data is collected inefficiently. Therefore, a study is conducted to improve the scalability while ensuring the error of the data through a cross-layer design that considers the data collected from the physical layer and the application layer.

LoRa can transmit packets through a chirp spread spectrum (CSS) modulation scheme that is resistant to interference and can be transmitted to a long distance. Depending on the chirp rate, it is divided into 6 spreading factors (SF) from 7 to 12. IoT device using SF 12, which has a high chirp rate, has a low receiver sensitivity due to a long channel occupancy time, and can be transmitted to a long distance. A device with a high SF can transmit a long distance, but due to the long channel occupancy time, the probability of a collision is very high compared to a low SF. This problem causes a scalability issue of the entire LoRa network, and IoT devices that use high SF to communicate with the distant devices have a very low reception probability. To solve this issue, we propose a novel fair and scalable relay control (FSRC) scheme that considers the physical characteristics of LoRa. The proposed FSRC scheme promotes relay operation with low SF to improve the success probability for distant end-devices (EDs) and the fairness of success probability for each SF region. To achieve this, a theoretical framework for designing the relay operation is analytically developed by considering a practical LoRaWAN MAC protocol as an analytical model. The proposed FSRC scheme encompasses a selective relay operation by considering both the signal-to-noise ratio and signal-to-interference ratio and the receive signal strength indicator value for the location-unaware relay selection strategy. Using this model, a genetic algorithm based relay control strategy is proposed to maximize both coverage probability and minimum success probability for all SF regions by controlling the relay parameters, such as source-relay region and source-relay ratio. Our numerical analysis validates the effectiveness of the proposed FSRC scheme under various parameters in terms of the minimum success probability of each SF region, coverage probability, and fairness.

In addition, a study is conducted to improve the scalability of the LoRa network through a deep learning-based cross-layer design scheme. Based on the characteristics of the physical layer of LoRa and the data actually collected in the application layer, a model for improving scalability is proposed by incorporating it into a deep learning model. In the application layer, a multi-modal based autoencoder model is proposed to cluster devices based on collected multi-modal data. After clustering is performed based on the characteristics of each device from the autoencoder model, a predictive model is generated for each cluster. For the prediction model, we propose a Global LSTM model that is pre-trained based on data from all devices, and a model that quickly learns a Local LSTM model for each cluster based on the data of each cluster and the Global LSTM model. Based on the error rate of the predicted value and the actual measured value through the results of the prediction model, a LoRa device transmission period control scheme is proposed that increases by one transmission period if the error rate is less than the threshold value, and initializes the initial transmission period when the error rate is greater than the threshold value. The determined periodic value is transmitted to each device through the gateway. Therefore, it is possible to reduce the probability of collisions caused by interference by transmitting data at a lower intensity than before in the LoRa network. Finally, through the proposed cross-layered design scheme, the scalability of the LoRa network can be increased while allowing the error of the data.

The proposed both relay control scheme and the deep learning-based cross-layered design scheme significantly improve the collision problem occurring in the LoRa wireless environment and consequently solve the scalability problem of the LoRa network. Through this dissertation, it is expected that IoT service providers will be able to accommodate more IoT devices, and IoT service users will be able to obtain higher user satisfaction than before by providing a higher packet success rate.