Network resource optimization with reinforcement learning for low power wide area networks

Park, Gyubong; Lee, Wooyeob; Joe, Inwhee

doi:10.1186/s13638-020-01783-5

Cited by 33 publications

(27 citation statements)

References 16 publications

(18 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The researchers in [ 14 ] presented the comparison of LPWAN standards and focused mainly on the LoRa standard. In [ 15 ], authors improved the performance of the LoRa network in terms of throughput by allocating appropriate resources. Further allocation of the same parameters to multiple nodes may increase the loss ratio in the densely populated environment.…”

Section: Related Workmentioning

confidence: 99%

Edge Based Priority-Aware Dynamic Resource Allocation for Internet of Things Networks

Ali

Qureshi

Mustafa

et al. 2022

Entropy

View full text Add to dashboard Cite

The exponential growth of the edge-based Internet-of-Things (IoT) services and its ecosystems has recently led to a new type of communication network, the Low Power Wide Area Network (LPWAN). This standard enables low-power, long-range, and low-data-rate communications. Long Range Wide Area Network (LoRaWAN) is a recent standard of LPWAN that incorporates LoRa wireless into a networked infrastructure. Consequently, the consumption of smart End Devices (EDs) is a major challenge due to the highly dense network environment characterised by limited battery life, spectrum coverage, and data collisions. Intelligent and efficient service provisioning is an urgent need of a network to streamline the networks and solve these problems. This paper proposes a Dynamic Reinforcement Learning Resource Allocation (DRLRA) approach to allocate efficient resources such as channel, Spreading Factor (SF), and Transmit Power (Tp) to EDs that ultimately improve the performance in terms of consumption and reliability. The proposed model is extensively simulated and evaluated with the currently implemented algorithms such as Adaptive Data Rate (ADR) and Adaptive Priority-aware Resource Allocation (APRA) using standard and advanced evaluation metrics. The proposed work is properly cross validated to show completely unbiased results.

show abstract

Section: Related Workmentioning

confidence: 99%

Edge Based Priority-Aware Dynamic Resource Allocation for Internet of Things Networks

Ali

Qureshi

Mustafa

et al. 2022

Entropy

View full text Add to dashboard Cite

show abstract

“…Tal problema impediu que o desempenho da proposta fosse avaliado em redes mais densas. Park et al propuseram a utilizac ¸ão da técnica de aprendizado por reforc ¸o profundo para ajustar o fator de espalhamento e a potência de transmissão [Park et al 2020]. A área na qual os dispositivos finais foram dispostos era de apenas 1,5 quilômetros quadrados e a quantidade de dispositivos foi de apenas 30.…”

Section: Trabalhos Relacionadosunclassified

“…Até o momento, a literatura dispõe de trabalhos que ou contam com o mecanismo padrão do LoRa, chamado de Adaptive Data Rate (ADR) [Cuomo et al 2017], ou com abordagens que utilizam técnicas de aprendizado de máquinas diferentes das utilizadas neste trabalho. No caso, essas abordagens são mais limitadas [Yatagan and Oktug 2019], por focarem apenas no fator de espalhamento, ou mais energeticamente custosas para os dispositivos finais, por usarem aprendizado por reforc ¸o [Ta et al 2019, Park et al 2020. Além de expandir a avaliac ¸ão de propostas baseadas em aprendizado de máquinas supervisionado que con-figuram apenas o fator de espalhamento, este trabalho propõe a configurac ¸ão em conjunto do fator de espalhamento e potência de transmissão a partir de modelos de aprendizado de máquinas supervisionado.…”

Section: Introduc ¸ãOunclassified

Economia de Energia através da Configuração Inteligente de Parâmetros de Camada Física em Redes LoRa

Filho

Campista²

2022

Anais Do XL Simpósio Brasileiro De Redes De Computadores E Sistemas Distribuídos (SBRC 2022)

View full text Add to dashboard Cite

A tecnologia LoRa (Long Range) é caracterizada por comunicações em longa distância e forte resiliência a interferências. No LoRa, a modulação é ajustada para permitir maiores taxas de transmissão de dados em função do limiar de sensibilidade de recepção e distância de comunicação. O fator de espalhamento e a potência de transmissão, por sua vez, estão diretamente relacionados com o consumo de energia, influenciando no desempenho da rede. Este artigo propõe o uso de técnicas de aprendizado supervisionado para seleção simultânea do fator de espalhamento e potência de transmissão. Essa abordagem se diferencia da literatura já que configura de forma supervisionada dois parâmetros ao invés de apenas um, o fator de espalhamento. Através de simulações de uma rede LoRa, diferentes técnicas de aprendizado são avaliadas. Simulações comparam o desempenho da proposta com o protocolo LoRaWAN tradicional e o estado da arte com seleção inteligente do fator de espalhamento. Os resultados mostram que a proposta é capaz de prever o sucesso da transmissão e ajustar os referidos parâmetros de modo a reduzir o consumo de energia, sem prejuízo da vazão e da taxa de entrega de pacotes.

show abstract

“…This method was suitable for very low data rates and very long transmission. Thus, the energy efficiency was improved but the throughput was decreased by up to 15% [ 5 ]. Dimitrios Zobras analyzed the same SF transmissions in different time slots and different SF transmissions in parallel by allocating SFs to nodes while maintaining the duty cycle [ 6 ].…”

Section: Related Workmentioning

confidence: 99%

“…Increasing the number of ENs for the same SF increases the probability of collision which decreases the total throughput [ 4 ]. The encoding method and transmission power need to change as the distance between GW and EN increases [ 5 ]. LoRaWAN cannot be used for large data payloads and is limited to 100 bytes.…”

Section: Introductionmentioning

confidence: 99%

Optimizing the Performance of Pure ALOHA for LoRa-Based ESL

Khan

Aamir

et al. 2021

Sensors

View full text Add to dashboard Cite

(1) Background: The scientific development in the field of industrialization demands the automization of electronic shelf labels (ESLs). COVID-19 has limited the manpower responsible for the frequent updating of the ESL system. The current ESL uses QR (quick response) codes, NFC (near-field communication), and RFID (radio-frequency identification). These technologies have a short range or need more manpower. LoRa is one of the prominent contenders in this category as it provides long-range connectivity with less energy harvesting and location tracking. It uses many gateways (GWs) to transmit the same data packet to a node, which causes collision at the receiver side. The restriction of the duty cycle (DC) and dependency of acknowledgment makes it unsuitable for use by the common person. The maximum efficiency of pure ALOHA is 18.4%, while that of slotted ALOHA is 36.8%, which makes LoRa unsuitable for industrial use. It can be used for applications that need a low data rate, i.e., up to approximately 27 Kbps. The ALOHA mechanism can cause inefficiency by not eliminating fast saturation even with the increasing number of gateways. The increasing number of gateways can only improve the global performance for generating packets with Poisson law having a uniform distribution of payload of 1~51 bytes. The maximum expected channel capacity usage is similar to the pure ALOHA throughput. (2) Methods: In this paper, the improved ALOHA mechanism is used, which is based on the orthogonal combination of spreading factor (SF) and bandwidth (BW), to maximize the throughput of LoRa for ESL. The varying distances (D) of the end nodes (ENs) are arranged based on the K-means machine learning algorithm (MLA) using the parameter selection principle of ISM (industrial, scientific and medical) regulation with a 1% DC for transmission to minimize the saturation. (3) Results: The performance of the improved ALOHA degraded with the increasing number of SFs and as well ENs. However, after using K-mapping, the network changes and the different number of gateways had a greater impact on the probability of successful transmission. The saturation decreased from 57% to 1~2% by using MLA. The RSSI (Received Signal Strength Indicator) plays a key role in determining the exact position of the ENs, which helps to improve the possibility of successful transmission and synchronization at higher BW (250 kHz). In addition, a high BW has lower energy consumption than a low BW at the same DC with a double-bit rate and almost half the ToA (time on-air).

show abstract

Network resource optimization with reinforcement learning for low power wide area networks

Cited by 33 publications

References 16 publications

Edge Based Priority-Aware Dynamic Resource Allocation for Internet of Things Networks

Edge Based Priority-Aware Dynamic Resource Allocation for Internet of Things Networks

Economia de Energia através da Configuração Inteligente de Parâmetros de Camada Física em Redes LoRa

Optimizing the Performance of Pure ALOHA for LoRa-Based ESL

Contact Info

Product

Resources

About