Wesley Kumfer

Investigation into the Role of Rational Ethics in Crashes of Automated Vehicles

Traffic engineers, vehicle manufacturers, technology groups, and government agencies are anticipating and preparing for the emergence of fully automated vehicles into the American transportation system. This new technology has the potential to revolutionize many aspects of transportation, particularly safety. However, fully automated vehicles may not create the truly crash-free environment predicted. One particular problem is crash assignment, especially between automated vehicles and nonautomated vehicles. Although some researchers have indicated that automated vehicles will need to be programmed with some sort of ethical system in order to make decisions on how to crash, few, if any, studies have been conducted on how particular ethical theories will actually make crash decisions and how these ethical paradigms will affect automated vehicle programming. The integration of three ethical theories—utilitarianism, respect for persons, and virtue ethics—with vehicle automation is examined, and a simple programming thought experiment is used to demonstrate the difficulty in selecting and implementing different ethical decisions. A simple crash scenario is introduced; an automated vehicle must choose between three crash types on the basis of a randomly assigned ethical theory. The results of the experiment indicate that in specific crash scenarios, utilitarian ethics may reduce the total number of fatalities that result from automated vehicle crashes, although other ethical systems may be useful for developing rules used in machine learning. The experiment demonstrates that understanding rational ethics is crucial for developing safe automated vehicles.

Effects of Demographic and Driver Factors on Single-Vehicle and Multivehicle Fatal Crashes: Investigation with Multinomial Logistic Regression

Human error is often considered the leading cause of motor vehicle crashes. Although some research has been conducted to assess the influence of human factors, full driver impacts on crashes are rarely analyzed, especially on a large scale in the United States. This study sought to identify the driver behavior and demographic factors that affected the likelihood of a multivehicle or single-vehicle fatal crash. A multinomial logistic regression framework, including odds ratios, was used to analyze the variables from several states. A tiered model approach was adopted to find the variable effects for combined, urban, rural, undivided urban, divided urban, undivided rural, and divided rural data sets. Each model produced different significant demographic or driver variables, many being unique or contradictory to the expected results of other research. Gender, often seen as a major contribution to crash outcome, was significant only for the full and urban models and likely not an important variable for determining crash outcomes in rural areas. Distracted driving and failing to make avoidance maneuvers were notably significant across various roadway types. Contrary to other studies, age, licensure, restraint use, and driving at certain times of the day were not found to be significant factors for either single- or multivehicle fatal crashes. Last, some previous conclusions about the number of occupants were refuted. These results may help safety analysts improve crash analysis and prevention methods.

A Human Factors Perspective on Ethical Concerns of Vehicle Automation

This paper presents a knowledge synthesis of ethical questions for the application of rational ethics theories to human factors in vehicle automation. First, a brief summary of ethical concerns related to transportation automation and human factors is presented. A series of theoretical questions are then posed for different levels of vehicle automation. Particular concerns relating to the Principle of Utility and the Principle of Respect for Persons are highlighted for low levels of automation, high levels of automation, and full automation through the use of theoretical scenarios. Although some recommendations are drawn from these scenarios, the primary purpose of this paper is to serve as a starting point to encourage discussion and collaboration between human factors professionals, engineers, policymakers, transportation officials, software programmers, manufacturers, and the driving public regarding realistic goals for automated vehicle implementation.

Traffic queuing at unsignalized crosswalks with probabilistic priority

Probabilistic yielding behavior is often observed at unsignalized crosswalks, but its impacts on the traffic flow in terms of traffic capacity and associated vehicular delay have not been examined before. The uniqueness of this problem is that neither traffic nor pedestrian flow holds the absolute priority as normally assumed in existing literature. Based on queuing theory, this paper developed traffic capacity and delay formulas considering the probabilistic priority. The proposed capacity equation reduces to the classic formula when the yielding rate equals one. The service time distributions for both queuers and non-queuers were derived, and M/G2/1 queuing model was applied to determine queue length and vehicular delay. Stochastic simulations showed that the proposed capacity and delay formulas precisely match the simulation results. For ease of practical applications, simpler queuing formulas including M/G/1, M/M/1 and M/D/1 were also examined. Noting that the differences between M/G2/1 and M/G/1 model are only marginal, especially for low yielding rates, we recommend the M/G/1 model for practical applications. In addition, when the pedestrian volume is relatively low, the M/M/1 model is also applicable due to its sufficient accuracy and simplicity.