Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Automotive Engineering

Committee Chair/Advisor

Beshah Ayalew

Committee Member

Zoran Filipi

Committee Member

Yunyi Jia

Committee Member

Pierluigi Pisu


Energy is an inherently limited resource for two reasons: the sources utilized to obtain energy are limited; and the methods used most often to convert energy also generate pollutants that the planet has a limited capacity to absorb or compensate for without adverse effects. Efficiently utilizing energy is, therefore, an important topic. As around 16% of the energy consumed in the United States is by light vehicles, a significant impact can be made if vehicle energy consumption is reduced. The advent of connected and automated vehicles offers unprecedented opportunities to optimize their behavior with respect to energy consumption through trajectory planning and coordination schemes. Due to recent advancements in computational capabilities and general-purpose non-linear solvers, model predictive control has become a promising tool for planning and control tasks, as it explicitly and intrinsically can capture vehicle motion and energy models while optimizing over a finite horizon. Further, communication technologies, such as dedicated short-range communication or 5th generation cellular broadband networks, allow for distributed yet coordinated control of automated vehicles. Taking advantage of these technologies, this dissertation proposes model-based planning approaches to coordinate connected and automated vehicles in an energy-aware fashion. We look in detail at two use cases of particular interest: 1) distributed maneuver planning of connected and automated vehicles on multi-lane roads, and 2) energy-aware coordination of multi-vehicle teams off-road including a recharging host.

Over 30% of vehicle miles traveled each year in the United States occurs on multi-lane highways and such roads are expected to be the first realized operation domain for automated vehicles, motivating the first use case. The structured environment of roadways allows for the use of simpler models, however, the potentially dense and highly dynamic environment necessitates appropriate considerations for coordinating vehicle flows. As such, we develop a distributed model predictive control framework that coordinates traffic through a distributed speed harmonization algorithm. Connected and automated vehicles are, however, expected to operate within mixed, human driven and automated, traffic for the foreseeable future. Therefore, we also develop methods for predicting the movement of unconnected neighboring vehicles. Then through extensive traffic micro-simulations, we show the potential of the proposed framework to reduce fuel consumption up to 30% depending on the level of connected and automated vehicle penetration.

We progress to the second use case within off-road environments. As off-road vehicles typically operate within resource constrained environments, it is imperative to ensure on-board energy reserves are not prematurely exhausted. Due to the structure of on-road driving, it is possible to simply minimize accelerations in order to reduce energy consumption. However, off-road domains do not share the same simplifying characteristics, motivating the use of more detailed vehicle models. As such, we develop a control oriented model that includes tire-terrain interaction and elevation change, in order to directly model energy as an output. As off-road environments lack a previously defined path (roadway) to follow, we develop a global energy-aware planner to generate the reference to be tracked by the local model predictive controller. Both a terminal cost-to-go to the goal and reference tracking integration of the global planner and local model predictive controller are tested. Although, the terminal cost-to-go method realizes reduced energy consumption, it is ultimately not computationally feasible, making the reference tracking method more desirable. The global planner is developed further to incorporate probabilistic energy constraints and additional turning effort on deformable terrains. The resulting high-confidence trajectory planning framework is able to increase the success rate from 46% with nominal global planning to 100%.

There is the potential that a large-scale mission (such as those seen in search and rescue, military, agricultural, or other domains) requires a team of multiple vehicles to be completed. Further, in spite of our best efforts with high-confidence trajectory planning, range beyond what is capable with the vehicles' on-board energy reserves may be required. From a cost and logistical perspective it is infeasible to build out a network of static charging stations to replenish energy reserves during the mission, motivating the incorporation of a mobile charging host within the vehicle team. To coordinate the workers (team members assigned tasks to complete) with the charging host, we solve a centralized charging rendezvous planning problem, posed as a mixed-integer problem, on-board the charging host. Through extensive Monte-Carlo simulations, we evaluate the scalability of the charging rendezvous planner and then the robustness of the complete closed loop system (charging rendezvous planning in conjunction with distributed trajectory planning).

Author ORCID Identifier


Available for download on Saturday, August 31, 2024