Dissertation
Adaptive behavioral planning for more social automated vehicles
University of Iowa
Doctor of Philosophy (PhD), University of Iowa
Spring 2021
DOI: 10.17077/etd.006061
Abstract
Driving automation technology is progressing at a rapid pace, and highly automated vehicles promise to be a fixture on public roadways soon. An open area of research in this domain involves how to design the behavior of highly automated vehicles best to assimilate and comingle with human drivers. The solution to this problem involves research on a broad swath of topics, including human perception and cognition, vehicle sensing technologies, social interaction, and computer science. This thesis takes a human factors-based approach to solving this problem.The theoretical basis of the model hinges on the idea that driving behavior is limited by the avoidance of visual cues relating to the perception of risk. A control theoretic model for this risk-perception-based model was proposed and validated against a naturalistic dataset of driving trajectories. The model converged to all test cases and was accurate to within the tolerance of vehicle sensing technologies in 91.7% of test cases.
A novel model of automated driving decision-making was then proposed using the previously described model. This model is unique in that it does not estimate the strategic value of the other vehicle’s actions, but instead estimates the behavioral limitations of the other drivers based on observed risk-perception parameters. This model was found to provide statistically significant improvement for the automated vehicle compared to control cases where the model was not present.
Finally, the impact of the proposed model on the traffic characteristics was measured based on the induced trajectories of the subject vehicle and other vehicle. The proposed model was found to have a significant impact on traffic flow and throughput, measures of traffic performance that represent impacts on systemic traffic characteristics.
This work could have an impact on the way automated vehicles designed for mixed-traffic environments are built. The model-based approach means that a smaller volume of data is required to train the model, which allows for a high level of adaptability. Additionally, this approach means that the parameters utilized by the model are highly interpretable, which could aid in wide acceptance of this method. Ultimately this thesis represents a first step in a new approach to designing automated driving systems that are more able to negotiate the complex social interactions requisite in the driving task.
Details
- Title: Subtitle
- Adaptive behavioral planning for more social automated vehicles
- Creators
- T Zachary Noonan
- Contributors
- Daniel V McGehee (Advisor)Stephen Baek (Committee Member)Venanzio Cichella (Committee Member)Chris Schwarz (Committee Member)Chao Wang (Committee Member)
- Resource Type
- Dissertation
- Degree Awarded
- Doctor of Philosophy (PhD), University of Iowa
- Degree in
- Industrial Engineering
- Date degree season
- Spring 2021
- DOI
- 10.17077/etd.006061
- Publisher
- University of Iowa
- Number of pages
- xi, 101 pages
- Copyright
- Copyright 2021 T Zachary Noonan
- Language
- English
- Description illustrations
- illustrations (some color)
- Description bibliographic
- Includes bibliographical references (pages 93-101).
- Public Abstract (ETD)
As vehicle automation becomes increasingly prevalent on public roadways, an important area of research will be designing automation that can make decisions to effectively intermingle with human drivers. Typical approaches to this problem involve abstract value-based models of rational decision-making. This project seeks to provide an alternative approach to designing a model of automated decision-making in interactions with other drivers. First, a model was proposed based on drivers’ approach to a perceptual risk boundary. This model was verified with naturalistic driver trajectory data. Then, the empirically derived parameters of said model were utilized to produce a novel model of automated decision-making called adaptive behavioral planning. It was demonstrated that the proposed model could improve outcomes of simulated interactions between drivers. Finally, the impact of the proposed system on the traffic system was analyzed using proxy measure of traffic effectiveness.
- Academic Unit
- Industrial and Systems Engineering
- Record Identifier
- 9984097278002771
Metrics
59 File views/ downloads
50 Record Views