Daniel J. Fagnant

Operations of Shared Autonomous Vehicle Fleet for Austin, Texas, Market

The emergence of automated vehicles holds great promise for the future of transportation. Although commercial sales of fully self-driving vehicles will not commence for several more years, once these sales are possible a new transportation mode for personal travel promises to arrive. This new mode is the shared autonomous (or fully automated) vehicle (SAV), combining features of short-term, on-demand rentals with self-driving capabilities: in essence, a driverless taxi. This investigation examined the potential implications of the SAV at a low level of market penetration (1.3% of regional trips) by simulating a feet of SAVs serving travelers in the 12-mi by 24-mi regional core of Austin, Texas. The simulation used a sample of trips from the regions planning model to generate demand across traffic analysis zones and a 32,272-link network. Trips called on the vehicles in 5-min departure time windows, with link-level travel times varying by hour of day based on MATSIMs dynamic traffic assignment simulation software. Results showed that each SAV could replace about nine conventional vehicles within the 24-mi by 12-mi area while still maintaining a reasonable level of service (as proxied by user wait times, which averaged just 1 min). Additionally, approximately 8% more vehicle miles traveled (VMT) may be generated because of SAVs ability to journey unoccupied to the next traveler or relocate to a more favorable position in anticipation of its next period demand.

Shifts in Long-Distance Travel Mode Due to Automated Vehicles: Statewide Mode-Shift Simulation Experiment and Travel Survey Analysis

Although vehicle automation technology has experienced rapid gains in recent years, little research has been conducted on the potential impacts of self-driving vehicles on long-distance personal travel, a major area of travel growth in the United States. Automated vehicles (AVs) offer flexible trip time and origin–destination pairings at travel time costs perceived to be lower; thus, AVs have the potential to dramatically change how travelers pursue long-distance tours. This study analyzed travel surveys and then developed a statewide simulation experiment of long-distance travel to anticipate the impact of AVs on long-distance travel choices. The research explored the Michigan State 2009 Long-Distance Travel Survey and estimated a long-distance trip generation model and a modal-agnostic long-distance mode-choice model. These models were applied in a statewide simulation experiment in which AVs were introduced as a new mode with lower perceived travel time costs (via lowered values of travel time en route) and higher travel costs (to reflect the initially high price of complete vehicle automation). This experiment highlighted the potential shifts in mode choices across different trip distances and purposes. For travel of less than 500 mi, AVs tended to draw from the use of personal vehicles and airlines equally. Airlines were estimated to remain preferred for distances greater than 500 mi (43.6% of trips greater than 500 mi were by air, and 70.9% of trips greater than 1,000 mi were by air). Additionally, at certain AV travel time valuations, travel cost was not a significant factor. The findings showed that as the perceived travel time benefits from hands-free travel rose, monetary costs became less important.

A direct-demand model for bicycle counts: the impacts of level of service and other factors

Transportation planning in the US has traditionally focused on automotive traffic but is increasingly turning towards a multimodal approach in order to accommodate all users. This shift in focus is particularly crucial for cyclists, as 630 were killed and 51 000 injured on America’s roadways in 2009 alone. Unfortunately, most municipalities do not conduct comprehensive bicycle counts to determine where cyclists are riding, though some do seek a scatter of spot counts. This investigation uses Seattle metropolitan area cyclist-count data from 251 locations to develop a direct-demand model for estimating peak-period cyclist counts based on trip-generation and attraction factors (such as site-based population and employment densities), as well as cycling-relevant roadway conditions. Roadway-condition variables were chosen from the 2010 Highway Capacity Manual (Chapter 17, Transport Research Board, Washington, DC) on urban street segments (including factors like traffic volumes and bike-lane width), as well as other physical features, like bridge presence and access to bicycle trails. Model results show greatest practical significance for intersections within the City of Seattle and curb-lane width (both of which are associated with higher counts) and roadway speed (which is associated with lower counts). The model was implemented in the community of Shoreline, Washington, just north of Seattle, to demonstrate its applicability. We examined both segment-based levels of service for cyclists and expected intersection-based counts. Model application findings indicate that segment-based levels of service show little correlation with expected counts, though such information still serves a valuable purpose in terms of evaluating cyclists’ comfort levels (which may also have potential correlation with cyclist safety), a worthwhile goal in itself. As such, these two models may be used in combination when targeting new cycling infrastructure and improvements.

Motorcycle Use in the United States: Crash Experiences, Safety Perspectives, and Countermeasures

Motorcycles are an important form of personal transportation used by many Americans. They provide an enjoyable recreational opportunity for many and a convenient, functional mode of transportation for others. However, U.S. motorcycle crash rates are 68% higher (per vehicle-mile traveled [VMT]) than those for other vehicles, and their fatal crash rates are more than 28 times higher. This investigation examines the riding and crash experiences and safety perceptions and recommendations of 238 U.S. motorcyclists. Top rider recommendations to other motorcyclists are to wear a helmet, to avoid riding under the influence (of alcohol or drugs), and to obtain motorcycle training. Ordered probit model results for helmet-use prediction suggest that those who ride on a daily or weekly basis and have received formal motorcycle training are more likely to wear a helmet. Other results show that rider training is most common among those who wear a helmet, ride on a weekly or monthly basis, and ride to and from work, school, and/or errands. Negative binomial model results for crash count experiences suggest that long-distance riders and those with criminal convictions are at greater risk of crashing than others, whereas those who have not received motorcycle training are less likely to have been involved in a crash at some point in their riding history.

Outcome of Transportation Projects Under Uncertainty

Budget constraints and competing opportunities demand thoughtful evaluation of projects before investment. Significant uncertainty surrounds travel choices, demographic futures, project costs, and model parameters. The impact of this uncertainty was explored with the use of hundreds of sensitivity test runs conducted across 28 random parameter sets to evaluate highway capacity expansion and toll project scenarios in Austin, Texas. The effects of different parameter sets on project benefit–cost ratios, crash counts, emissions, traffic volumes, and tolling revenues were examined in detail. Linear regression results showed that link capacity, link performance parameters, and their covariation were key to the results, followed by the elasticity of demand, trip growth rates, and values of travel time.

Who Rides and Who Pays: Comprehensive Assessment of Motorcycling Costs and Benefits in the United States

A comprehensive benefit–cost assessment of motorcycle use is offered, along with the characteristics, behaviors, and attitudes of motorcycle riders. U.S. motorcyclists are at high risk of crashing, with rates 24 times higher than those of passenger car and light-duty truck drivers per mile traveled. However, motorcycles require only one-quarter the parking space of a car and can reduce congestion by doubling network capacity (in vehicles per hour). Most motorcycles have high fuel economy, but low seating capacities render them little or no better than most cars and some light-duty trucks (assuming average vehicle occupancies) when compared in terms of fuel consumption per rider (or vehicle occupant) mile traveled. Motorcycles emit fewer grams of many polluting emissions but more volatile organic compounds and carbon monoxide than most cars if a catalytic converter is not installed. Noise impacts are a serious issue, with an inconsistent patchwork of regulations applied across states and localities. Results of a survey of current and former U.S. motorcyclists indicate that most respondents ride motorcycles for recreational purposes and ride in groups, but about one-half also ride for more mandatory (less discretionary) purposes and about 40% also ride solo. Fewer than one-third of respondents have formal motorcycle training, and helmet use is lowest among current riders who do not own a motorcycle. Engine size appears to be rising, and respondents support policies that combat operating a vehicle under the influence (e.g., ignition interlock devices for offenders). Regression models illuminate key factors and marginal effects on motorcycle riding and ownership rates.

Beyond Single Occupancy Vehicles: Automated Transit and Shared Mobility

It is commonly accepted that Automated Transit will still be as relevant as it is now, if not more so, even when fully-automated vehicles become a reality. We need to develop a consensus on how vehicle automation will transform and perhaps disrupt the traditional transit systems, what new and different types of market-driven and publicly-run frameworks will emerge, and how we should invest our limited public resources. The two day session on Automated Transit and Shared Mobility Track (ATSM) during the 2015 Automated Vehicle Symposium (AVS) explored implications for the changing roles of transit and shared mobility as vehicle automation progresses. This chapter not only documents the main ideas presented during the symposium, but also supplements certain ideas with further discussions and clarifications after the conference.

Welfare Analysis Using Logsum Differences Versus Rule of Half: Series of Case Studies

Logsum differences and rule-of-half (ROH) calculations are two methods for estimating consumer surplus in transport economics. As a traditional and relatively straightforward (and potentially more robust) procedure, ROH is widely used in project investment and policy analysis, and much of the literature agrees that logsums are somewhat superior to ROH when user benefits are valued, at least when true travel behaviors stem from random-utility maximization with Gumbel error terms. This paper explores the differences in both methods through a careful review of literature and many case study results. The comparison of ROH and logsum methods relies on three specifications in order of increasing complexity: binary logit, multinomial logit, and nested logit models under a variety of settings and scenarios. This work offers a closer look at three numerical examples and concludes that the difference between ROH and logsum solutions rises with increases in travel times or costs and with changes in parameters. The monetized differences in logsums is usually smaller than the ROH solution for welfare changes under most situations and gives a more exact result for consumer surplus than ROH (which assumes a linear demand relationship with respect to cost). Larger coefficients on affected variables (such as travel time and cost) in the random-utility expressions tend to increase differences between logsum-and ROH-based estimates. Such findings should be of interest to policy makers and planners when they are developing transportation planning and land use models and interpreting their results for more accurate and rigorous and behaviorally defensible project evaluations.

Anticipating Roadway Expansion and Tolling Impacts: Toolkit for Abstracted Networks

Transportation investments are nearing

Preparing a nation for autonomous vehicles: opportunities, barriers and policy recommendations

1,000 per capita annually in the U.S., and the Highway Trust Fund has been depleted. Such significant investments and budget-constrained contexts demand careful decision-making and thoughtful cost-benefit analysis. A toolkit has been developed for comprehensive assessment of network expansion and pricing projects with only project expenditures, link attributes, and traffic counts as required inputs. The toolkit uses a self-contained travel demand model to predict future and alternative scenario traffic volumes, speeds, crash counts, emissions and toll revenues, while providing project-summary measures, including net present value and benefit/cost ratios. The toolkit seeks to provide early assessment of major projects along abstracted networks, using hundreds of coded links (rather than thousands), providing results in a matter of minutes (rather than days). This paper describes the model and develops two case study sites, each with several alternative scenarios. The first examines capacity expansion projects along a highly congested link on the periphery of Austin, Texas, while the second focuses on strategies to reduce traffic in central Austin (through tolling and capacity reduction projects). Toolkit results show which projects merit further consideration by summarizing and monetizing impacts across scenarios.