Off-campus UMass Amherst users: To download campus access dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.

Non-UMass Amherst users: Please talk to your librarian about requesting this dissertation through interlibrary loan.

Dissertations that have an embargo placed on them will not be available to anyone until the embargo expires.

Author ORCID Identifier


Open Access Dissertation

Document Type


Degree Name

Doctor of Philosophy (PhD)

Degree Program

Computer Science

Year Degree Awarded


Month Degree Awarded


First Advisor

Shlomo Zilberstein

Subject Categories

Artificial Intelligence and Robotics


The path to building adaptive, robust, intelligent agents has led researchers to develop a suite of powerful models and algorithms for agents with a single objective. However, in recent years, attempts to use this monolithic approach to solve an ever-expanding set of complex real-world problems, which increasingly include long-term autonomous deployments, have illuminated challenges in its ability to scale. Consequently, a fragmented collection of hierarchical and multi-objective models were developed. This trend continues into the algorithms as well, as each approximates an optimal solution in a different manner for scalability. These models and algorithms represent an attempt to solve pieces of an overarching problem: how can an agent explicitly model and integrate the necessary aspects of reasoning required to achieve long-term autonomy?

This thesis presents a general hierarchical and multi-objective model called a policy network that unifies prior fragmented solutions into a single graphical decision-making structure. Policy networks are broadly useful to solve numerous real-world problems. This thesis focuses on autonomous vehicle (AV) problems: (1) route-planning with multiple objectives; (2) semi-autonomy with proactive transfer of control; and (3) intersection decision-making for reasoning online about any number of other vehicles and pedestrians. Formal models are presented for each of the distinct problems. Solutions are evaluated using real-world map data in simulation and demonstrated on a fully operational AV prototype driving on real public roads. Policy networks serve as a shared underlying framework for all three, enabling their seamless integration as parts of an overall solution for rich, real-world, scalable decision-making in agents with long-term autonomy.