Date of Award

2026

Document Type

Open Access Dissertation

Degree Name

Doctor of Philosophy in Mechanical Engineering-Engineering Mechanics (PhD)

Administrative Home Department

Department of Mechanical and Aerospace Engineering

Advisor 1

Darrell Robinette

Advisor 2

Jeffrey Naber

Committee Member 1

Jungyun Bae

Committee Member 2

Scott Wagner

Abstract

The transportation sector currently accounts for nearly 30% of global energy consumption, necessitating urgent advancements in vehicle efficiency to meet Net Zero targets. Leveraging connectivity and automation, this dissertation proposes and validates methodologies to reduce the energy consumption of light-duty vehicles at both fleet and individual levels.

First, a validation framework is developed to bridge the “simulation-to-real world” gap in Cooperative Automated Vehicle (CAV) research. Moving beyond virtual simulations, the study establishes a methodology for physically validating centralized control architectures via a custom Cellular V2X network. By synchronizing vehicle-powertrain models with physical test vehicles, the framework successfully orchestrates complex arterial maneuvers on a closed track. Experimental results confirmed the platform's robustness against real-world uncertainties, such as actuator delays and localization noise, validating strategies that realized cohort-wide energy savings of up to 19% and travel time reductions of 11%.

Second, the research investigates the energy savings potential of heterogeneous vehicle cohorts on limited-access highways. Unlike traditional homogeneous platooning studies, this work addresses the complex dynamics of mixed-traffic consisting of diverse vehicle types (sedans, minivans, and pickup trucks) and powertrain architectures (Internal Combustion, Hybrid, and Electric). Through on-road testing, the study quantifies the impact of vehicle ordering on aerodynamic drag. The results demonstrate that optimal cohort configuration can yield up to 10% energy savings compared to running independently. An optimization framework is introduced to assess the "cost of reconfiguration"—the energy expended to shuffle vehicles into an optimal order. Using a System of Systems (SoS) simulation environment, the study establishes that this reordering cost is negligible (typically < =0.45% of trip energy), thereby validating the viability of dynamic platoon reconfiguration for long-distance travel.

Third, the dissertation shifts focus to the individual vehicle level, developing a Deep Reinforcement Learning (DRL) framework for the predictive energy management of multi-mode Hybrid Electric Vehicles (HEVs). Focusing on the complex Generation II Chevrolet Volt powertrain, a Soft Actor-Critic (SAC) agent is trained to optimize powertrain operating modes and torque splitting in real-time. The framework utilizes a Long Short-Term Memory (LSTM) network for state prediction, allowing the controller to handle hybrid action spaces that combine discrete mode selection with continuous control targets. This Reinforcement Learning approach allows for adaptive energy management strategies that can outperform rule-based controllers across diverse real-world driving cycles.

Creative Commons License

Creative Commons Attribution-Share Alike 4.0 License
This work is licensed under a Creative Commons Attribution-Share Alike 4.0 License.

Download

Share

COinS