An automated decision making framework for modern vehicles CO(2) emissions using multi modal engine telemetry and feature interpretability.

Saraswat, SK; Abdullah, M; Habelalmateen, MI; Vivek, V; Sahu, PK; Pant, R; Sharma, S; Shojonov, M; Madaminov, B · 2026 · PubMed Central

DOI: 10.1038/s41598-026-42137-3

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the challenge of accurately predicting vehicle CO₂ emissions, a critical issue given that transportation is the second-largest source of greenhouse gases. The authors propose an automated decision-making framework that integrates multi-modal engine telemetry with explainable machine learning to model the complex, nonlinear relationships between vehicle attributes and emissions. The research aims to improve prediction precision and convergence stability while providing interpretability to identify dominant emission drivers, thereby supporting low-carbon vehicle design and urban mobility planning. The methodology utilizes a dataset of 7,385 vehicles from the Government of Canada, spanning seven years and including features such as engine displacement, cylinder count, fuel type, transmission, and fuel consumption metrics. The core model is a Multi-Layer Perceptron (MLP) neural network, optimized using two nature-inspired metaheuristic algorithms: the Horned Lizard Optimization Algorithm (HLOA) and the Giant Armadillo Optimization (GAO). These optimizers were selected to overcome limitations of traditional methods like premature convergence and sensitivity to initialization. Feature engineering involved Recursive Feature Elimination (RFE) to select the most predictive variables, followed by interpretability analysis using SHAP and Class Activation Mapping (CAM). The results demonstrate that the GAO-enhanced MLP achieved superior predictive performance, yielding an R² of 0.9881 and a Root Mean Square Error (RMSE) of 6.478. Feature analysis identified fuel consumption in liters per 100 km and miles per gallon as the highest predictors of CO₂ emissions, followed by fuel type. Correlation analysis confirmed strong positive relationships between engine size, city fuel consumption, and CO₂ emissions. The study highlights that GAO provided effective global exploration during early search stages, while HLOA enhanced local refinement, resulting in robust model generalization. The significance of this work lies in its integration of high-accuracy prediction with model transparency. By combining metaheuristic optimization with interpretability techniques, the framework offers a reliable tool for forecasting emissions and understanding their determinants. The authors conclude that this approach can inform intelligent transportation systems, facilitate real-time emission monitoring, and support environmentally conscious policy-making and infrastructure development.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• telematics crash prediction