Edge-IoT Driven Smart Air Quality Monitoring Using the Proposed EdgeAQ-Net Distributed Deep Learning Model
TL;DR
An Edge-IoT-based smart air quality monitoring system based on the proposed EdgeAQ-Net distributed deep learning architecture that combines edge computing and IoT-based environmental sensors to allow real-time analysis and prediction of air quality to be performed at the network edge.
Abstract
Ambient air quality has been greatly impaired by rapid urbanisation and industrialisation, which is a serious threat to the overall health of the population and the sustainability of the environment. Traditional air quality devices use centralised cloud-based systems, which can be characterised by a high level of latency, bandwidth consumption, and inability to scale to large volumes of real-time sensor data. In addition, most of the current monitoring frameworks do not have intelligent edge-level analytics that can conduct distributed learning among heterogeneous Internet of Things (IoT) devices. In order to overcome these shortcomings, this paper suggests an Edge-IoT-based smart air quality monitoring system based on the proposed EdgeAQ-Net distributed deep learning architecture. The proposed solution combines edge computing and IoT-based environmental sensors to allow real-time analysis and prediction of air quality to be performed at the network edge. The EdgeAQ-Net model is a hybrid of convolutional feature extraction and the use of bidirectional long short-term memory (BiLSTM) layers, which are effective in identifying spatial-temporal patterns of air pollutant data. Also, a lightweight distributed learning system is deployed to provide collaborative model updates among edge nodes and minimise communication costs. Benchmark urban air quality datasets with various pollutant indicators, including PM2.5, PM10, NO2, and CO, are used to validate the performance of the experimental model with respect to conventional machine learning and deep learning frameworks, including SVM, Random Forest, and normal LSTM. The accuracy of the proposed framework is 97.84, which is higher than the baseline models, and the inference latency is also 32% lower, and network bandwidth consumption is also 28% less. The findings indicate that the developed EdgeAQ-Net framework is very effective in increasing the efficiency and scalability of real-time air quality monitoring in the distributed smart city setting.
