Advances, Systems and Applications
From: Data transmission reduction formalization for cloud offloading-based IoT systems
Data type | System | Transmission reduction method |
---|---|---|
Based on prediction model | ||
 Temperature, Humidity, Light, Infrared, and Voltage | Sensor nodes and a central workstation | A synchronized prediction model on both sides, transmits only the values that surpasses a predefined error threshold [22] |
 Temperature, humidity, and light intensity | Between sensors and the cluster head | Both sides are using the same prediction model to predict the next data. If the prediction is different from the real captured data, the captured data will be sent to the cluster head node [23] |
 Humidity, temperature, light, and voltage values | Between a sensor node and the gateway in a WSN for environmental monitoring | The sensor and the gateway predict (using Least Mean Square algorithm) the future value simultaneously based on the past measurements [24] |
 Current production capacity, the volume of produced bottles, the volume of broken bottles, temperature and pressure | Between smart industrial machines and the server monitoring the system | If the new captured data is equal to the previous one or has a linear behavior, it can be deducted by the server [26] |
 Temperature, humidity, light, voltage values | Sensor nodes and cluster heads in a WSNs | Least-Mean-Square (LMS) algorithm to predict the next value. The sensor node sends the captured values until the cluster head is able to predict them with an acceptable error [25] |
Based on classifiers | ||
 Movement of the person | Wearable sensor networks (on a human body with a cluster head) | Classifiers (machine learning algorithms) that allow sensor nodes to decide if current sensor readings have to be transmitted to the cluster head or not [10] |
 Vehicles’ locations | Mobile IoT devices and an edge server | A machine learning model trained on the importance of data, each captured data is classified as important or not and only the data of importance in tra c flow prediction will be sent to the server [27] |
Based on data compression | ||
 Images | Between low power IoT devices and the cloud | Proposed content-awareness and task-awareness compression method using tiny machine learning models [8] |
 Temperature, humidity, and light intensity | Between cluster heads and the cloud | Data compression done at the cluster head to benefit from the spatio-temporal correlation of the data collected by the sensors of the cluster [23] |
 Images | visual sensors in WSN | Generating the optimal representation for each pixel value [28] |
Based on heuristics | ||
 Temperature, light, humidity | Application-specific IoT networks | Machine learning model in the sensor to calculate the number of humans in a room, send it to the cloud/edge server if it is different from the previous one [31] |
 Images | Visual sensor networks (camera nodes) | Eliminate near-duplicate images: near-duplicate clustering, seed image selection then delete all the rest images in the cluster [30] |
 Images | From a surveillance camera in an agricultural context to a server | Send the first captured image, then, whenever a change is detected only the changed part is sent. An image reconstruction phase is performed on the server’s side [29] |