Author ORCID Identifier

https://orcid.org/0009-0008-2905-7218

Semester

Summer

Date of Graduation

2025

Document Type

Dissertation

Degree Type

PhD

College

Statler College of Engineering and Mineral Resources

Department

Lane Department of Computer Science and Electrical Engineering

Committee Chair

Natalia Schmid

Committee Co-Chair

Arvind Thiruvengadam

Committee Member

Matthew Valenti

Committee Member

Katerina Goseva-Popstojanova

Committee Member

Rasik Pondicherry

Committee Member

Sasanka Katreddi

Abstract

Heavy-duty trucks constitute only a modest fraction of on-road vehicles, yet their intensive duty cycles and high fuel demands yield a disproportionately large share of transportation fuel use and greenhouse gas emissions. Addressing this imbalance requires data-driven tools that capture the realities of fleet operation and translate complex telemetry into actionable insight.

This dissertation introduces a unified machine-learning framework that operates exclusively on high resolution time-series data collected from fifty-nine diesel trucks deployed across Southern California. It begins by constructing a multi-modal feature space that blends statistical summaries of key engine signals, static vehicle descriptors, and Mel-Frequency Cepstral Coefficients, thereby providing a compact yet expressive encoding of temporal and structural behavior. An unsupervised clustering pipeline based on Ward-linkage agglomerative clustering then partitions the fleet into semantically coherent operational segments, revealing groupings such as long-haul freight, urban delivery, and construction service trucks.

Within these segments, the study investigates several predictive architectures, including feed-forward neural networks, Long Short-Term Memory models, and a hybrid design that couples LSTM-derived temporal embeddings with Gradient-Boosting regression. Empirical evaluation shows that aligning model training with the discovered behavioral clusters substantially improves fuel economy prediction accuracy, particularly under high-variance duty cycles, while preserving the interpretability required for operational decision making. Beyond methodological innovation, the proposed framework supports practical benefits: fuel-use forecasting, scheduling and maintenance planning, and establishes a scalable foundation for anomaly detection and emissions reporting as environmental regulations tighten. Collectively, the work advances intelligent transportation analytics by bridging complex real-world telemetry and sustainable fleet management, charting an interpretable and extensible path toward lower emissions and higher operational efficiency in heavy-duty trucks.

Share

COinS