Transfer Learning-Enabled IoT System for Continuous Prediction of Vehicle CO2 Concentration

Transfer Learning-Enabled IoT System for Continuous Prediction of Vehicle CO2 Concentration

In this paper, we present the design, implementation, and deployment of an IoT-based system for machine learning (ML)-based real-time prediction of CO 2 exhaust concentrations in road vehicles, by means of transfer learning. The system offers a scalable and cost-effective solution for monitoring and...

Full description

Saved in:
Bibliographic Details
Published in:2023 IEEE International Smart Cities Conference (ISC2) pp. 1 - 7
Main Authors: Al Selek, Mohammad, Tena-Gago, David, Alcaraz-Calero, Jose M., Wang, Qi
Format: Conference Proceeding
Language:English
Published: IEEE 24-09-2023
Subjects:
Artificial neural networks
CO2 prediction
Computational modeling
Edge Computing
Embedded system
IoT
Microcontrollers
Performance evaluation
Predictive models
Real-time systems
Transfer learning
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this paper, we present the design, implementation, and deployment of an IoT-based system for machine learning (ML)-based real-time prediction of CO 2 exhaust concentrations in road vehicles, by means of transfer learning. The system offers a scalable and cost-effective solution for monitoring and predicting CO 2 exhaust concentrations from a vehicle fleet. Moreover, the system is specifically designed to be installed in a low-powered microcontroller unit (MCU) to collect and process data from a CO2 sensor, as well as to perform ML-based emission prediction without a sensor. To accomplish this, we have developed an artificial neural network (ANN) model able to generate training labels to further train subsequent ANN models via transfer learning mechanisms. The resulting models are sent to the IoT devices installed in the vehicle to perform the on-board predictions. Empirical experiments show very promising results in terms of high prediction accuracy and satisfactory transfer learning loop execution time.
ISSN:2687-8860
DOI:10.1109/ISC257844.2023.10293394