THE ROLE OF EDGE COMPUTING IN ENHANCING DATA PROCESSING EFFICIENCY

Abstract

The gadgets that make up the Internet of Things (IoT) collect a mountain of data from people and their surroundings, yet they frequently run on a tight budget. In order to make educated judgments and precise forecasts, machine learning algorithms analyze this kind of sensor data. With billions of devices sending data at once, networks can get overwhelmed, rendering conventional cloud data processing useless for Internet of Things applications. This research looks at the latest developments in models, architectures, hardware, and design requirements for machine learning deployment on edge devices and cloud networks with limited resources. Popular Internet of Things (IoT) devices that use edge intelligence include Beaglebone AI, SparkFun Edge, SparkFun, Google Coral Dev Board, SparkFun, Arduino Nano 33 BLE Sense, and Jetson from NVIDIA, among many more. Utilizing bespoke AI frameworks like TensorFlow Lite, OpenEI, Core ML, Caffe2, and MXNet, these devices are capable of running ML and DL workloads, including object identification and gesture detection. Many deep learning algorithms are used by these devices, including YOLOv4, ResNet-50, distributed edge computing, and long short-term memory (LSTM). In addition, we used decision tree, random forest, and support vector machine classifiers to sift through a thousand articles published in the last year on "ML in IoT" in IEEE Xplore in search of new research topics and potential projects. Topics such as big data, cloud computing, multimedia, privacy, security, quality of service, and activity recognition were popular, while important areas of focus included healthcare, transportation, smart cities and homes, agriculture, and assisted living. Some of the main obstacles to implementing edge machine learning include protecting sensitive user data on edge devices, finding an efficient way to manage edge node resources through distributed learning architectures, and finding a balance between machine learning's energy demands and the limited energy available on edge devices.