Safe Transportation Research & Education Center

Public acceptance remains a major barrier to autonomous vehicle (AV) deployment, yet most perception research relies on hypothetical surveys or small controlled experiments with AV-naive respondents. This cross-sectional study surveys 811 San Francisco Bay Area residents with real-world AV exposure (376 robotaxi passengers and 435 non-passengers: pedestrians, cyclists, and drivers) in one of the world’s most mature deployment environments. We ask whether trust levels among AV-exposed residents differ from national baselines, whether trust varies by interaction mode, and whether experience is associated with differential recalibration of specific safety concerns. We find that 61% trust AVs to be safe, 4.7 times the 13% national rate. Trust varies sharply by mode: passengers report 82% trust versus 43-45% among non-passengers. Even non-riders far exceed national levels, consistent with ambient exposure beyond ridership alone contributing to higher trust, though residential selection cannot be ruled out. Concern about system malfunction is significantly lower among passengers (45%) than pedestrians and cyclists (63%), consistent with recalibration through direct performance feedback, whereas concern about unpredictable human behavior remains constant across all modes ( ∼ 55%). These patterns suggest that experience is associated with attenuation of technology-related concerns while human-behavior concerns persist: a divergence with implications for communication strategies and infrastructure design as AV deployment expands.

The safe deployment of autonomous vehicles (AVs) depends on the ability of automated driving systems (ADS) to handle rare, complex, and safety-critical “edge cases.” This study develops a novel framework for identifying and analyzing such scenarios using the National Highway Traffic Safety Administration (NHTSA) ADS crash dataset. Two complementary approaches are applied. First, large language models (LLMs) are used to directly analyze crash narratives, identifying edge cases through high-risk keyword and phrase detection and anomaly-based rarity analysis. Second, LLMs are employed to extract structured variables from narrative fields, which are then analyzed using hierarchical clustering to systematically isolate unusual crash groups. Edge cases are characterized by a higher prevalence of unusual crash partner behaviors, non-motorist involvement, roadway anomalies, and disengagement of AV systems, highlighting their distinct and atypical nature. The findings underscore the importance of focusing AV evaluation on rare, high-risk scenarios that challenge ADS performance. The study advances AV safety research and can provide a foundation for refining testing protocols, safety standards, and regulatory frameworks to better capture the operational limits of AVs.

This report investigates jaywalking laws in connection with traffic safety, racial equity, and street design, focusing on California. It traces the concept of "jaywalking" to an early 20th-century auto industry campaign to shift safety responsibility from drivers to pedestrians. By analyzing national and California pedestrian injury and fatality data (2009–2022) alongside California Racial and Identity Profiling Act (RIPA) police stop data (2018–2022), the study describes demographic disparities in both pedestrian crashes and law enforcement of jaywalking. It also documents recent legislative efforts in California and other states and cities to decriminalize or reform jaywalking enforcement. Findings show that pedestrian fatalities reached a 40 year high in 2022, with California’s rates consistently exceeding the national average. Significant racial and economic disparities exist: Black pedestrians experience fatality rates multiple times those of White pedestrians, and lower-income neighborhoods suffer disproportionately. RIPA data further reveal that jaywalking-related police stops disproportionately affect Black pedestrians. These disparities are likely driven by the built environment—such as wide arterials and sparse crosswalks—which incentivizes mid-block crossings, particularly in under-invested communities. The report also examines California’s Assembly Bill (AB) 2147 (2022), which partially decriminalized jaywalking by limiting enforcement to cases of "immediate hazard." It concludes by recommending continued monitoring of enforcement and safety data to track AB 2147’s impact, alongside collecting built environment data to better contextualize racial and economic disparities in pedestrian outcomes.

The Vehicle Weight Safety Study provides supporting analysis for the California Transportation Commission’s study on therelationship between vehicle weight and road user injury and roadway degradation required by Assembly Bill (AB) 251, which was signed by the Governor in October 2023. To inform the work of the CTC, this report summarizes trends of road user injuries and fatalities in California and potential factors contributing to these trends (Chapter 2); summarizes trends in vehicle weight, size, and height for registered vehicles in California (Chapter 3); documents the landscape of policy solutions focused on vehicle size that might address California’s road user injuries and fatality challenge (Chapter 4); analyzes the impact of potential weight-based fees on consumer vehicle purchasing behavior (Chapter 5); and, analyzes the relationship between shifts in passenger vehicle weight and degradation of road infrastructure (Chapter 6).

In the US, speed limit setting (SLS) procedures have historically relied on driver-behavior-based methods, such as the 85th percentile speed, which are considered objective and allow for consistent application. However, this approach has notable shortcomings, including drivers’ tendency to underestimate their speeds, speed creep, and insufficient consideration of vulnerable road users, which may conflict with the Safe System Approach and Vision Zero initiatives endorsed by the USDOT (US Department of Transportation). In contrast, context-sensitive approaches, which classify roads based on roadway typologies, have been developed in countries like New Zealand, Sweden, the Netherlands, and Australia. While effective, these approaches have largely been applied outside the US, leaving many US roads with speed limits that may not fit their surroundings or adequately address pedestrian and cyclist safety. Drawing on New Zealand’s One Network Framework, we developed a US-based, context-sensitive roadway classification framework for urban and suburban areas that incorporates “Place,” which captures surrounding land uses and locational contexts, and “Movement,” which relates to the road’s transport function. Using nationally available data from the Smart Location Database (SLD) and the Highway Performance Monitoring System (HPMS), we evaluated our roadway classification framework through internal reviews by our research team and external interviews with state-level practitioners, uncovering both opportunities and challenges in adopting a context-sensitive SLS approach in the US. Our findings demonstrate the feasibility of creating an objective context-sensitive roadway classification in the US and offer insights for developing new speed-limit guidance aligned with the Safe System framework.

Bicycle level of service (BLOS) is an essential performance measure for transportation agencies to monitor and prioritize improvements to infrastructure, but existing measures do not capture the nuance of facility differences on the state highway system. However, with the advancements in virtual reality (VR) technology, a VR bicycle simulator is an ideal tool to safely gather user feedback on a variety of bicycling environments and conditions. This research explored the benefits and limitations of using a VR environment to assess individuals’ bike infrastructure preferences. We conducted a bicyclist user experience survey in person on SafeTREC’s VR bicycle simulator and online and compared the results. The online survey consisted of showing participants pairs of VR videos of biking scenarios and asking them to choose the one that they preferred. To validate the online survey responses, we conducted in-person experiments with a VR bike simulator using the same pairs of videos. Our analysis indicates that 63 percent of the responses were consistent while a smaller percentage of responses (37 percent) changed after the simulator ride due to better perception provided by the simulator virtual environment. The outcome of this study helped to validate the online survey responses of the study.

The escalating number of injuries and fatalities among cyclists is a pressing safety concern. In the United States, communities are actively seeking strategies to boost cyclist safety, with some states implementing bike-specific policies, such as stop-as-yield laws, to support cyclists. Stop-as-yield laws allow cyclists to treat stop signs as yield signs. The laws are not yet widely implemented, and their potential safety impact is a subject of debate among transportation experts and advocates. This study investigates how stop-as-yield laws can positively or negatively affect safety and provides insights and guidelines for California policymakers and safety practitioners if the law passes in California. We collected cyclist data from five states that have enacted stop-as-yield laws—Idaho, Arkansas, Oregon, Washington and Delaware—and data from some of their contiguous states without such legislation. Using an observational before-after study with comparison groups at the state level, the research examined changes in cyclist crash frequencies after the laws were implemented. Additionally, a random-effects negative binomial regression model with panel data was employed to estimate a law’s overall impact. The results did not indicate a significant change in cyclist crashes among the states with stop-as-yield laws.

In 2018, SafeTREC conducted a survey on transportation mobility issues among older adults in California. A follow-up survey planned for 2020, just as the COVID-19 pandemic changed life for all residents, was redesigned to assess mobility needs and changes during the Shelter-in-Place order and focused on COVID-19 impacts. Results indicate that the COVID-19 pandemic and subsequent Shelter-in-Place order have had a major impact on senior mobility. Communications for many were restricted to phone, email, texts, social media and video chats. Among those with a medical problem, just over 60% called a doctor or nurse line or went to a doctor’s office, while 11.2% went to an emergency room, and 8.6% did nothing. A total of 8% of respondents said they had run out of food or other important items during the Shelter-in-Place order. Rates of exercise outside the home dropped 20% between January and June 2020, and while over 60% sought outside entertainment in January 2020, by June 2020, nearly 70% accessed their entertainment online at home. Almost 80% of working respondents feared spreading or contracting COVID-19 because of their work or related transportation. Almost 20% felt a lack of companionship or closeness sometimes or often. Over 30% were worried about their current or long-term finances. A total of 84.5% strongly agreed or agreed that the Shelter-in-Place order was necessary. None of the respondents to the follow-up survey were diagnosed with COVID-19, and 88.2% were not concerned about risk of exposure from any member of their household.

Investigating the dynamics behind the likelihood of vehicle crashes has been a focal research point in the transportationsafety field for many years. However, the abundance of data in today's world generates opportunities for deepercomprehension of the various parameters affecting crash frequency. This study incorporates data from many differentsources including geocoded police-reported crash data, curbside infrastructure data and socio-demographic data for thecity of San Francisco, CA. Findings revealed that the GFMNB model provides a better statistical fit than the FMNB andNB model in terms of AIC and log likelihood, while the NB model outperformed both mixture models in terms of BIC dueto model complexity of the latter. Among the signicant variables, TNC pick-ups/dropoffs and duration of parked vehicleswere positively associated with segment-level crashes.

The mobility needs of an aging population is one of the most substantial challenges facing California in the coming decades. The number of residents age 65 and older is expected to double between 2012 and 2050, and the number age 85 and above is expected to increase by over 70% between 2010 and 2030. Declines in physical function related to age may reduce mobility options dramatically. A survey of 510 residents age 55 and older in Contra Costa County was conducted to determine mobility patterns and limitations related to age and other factors. Results of the survey indicate that a majority of seniors are car dependent. However, some older adults miss important activities due to mobility limitations associated with increasing age, poorer health, living alone, not having a licensed driver in the household, and having a disability. Mobility options are also limited in some geographic areas and demographic groups. Importantly, older adults want to “age in place.” Based on these findings and those in related studies, the travel options and the quality of life for older adults, now and in the future, can be greatly enhanced if efforts are made to develop mobility solutions beyond use of private vehicles. The findings support the recommendations of recent regional plans such as the Coordinated Public Transit–Human Services Transportation Plan (2018), adopted by the Metropolitan Transportation Commission (MTC) of the San Francisco Bay Area, which recommends supporting a range of mobility options centered around shared mobility and accessibility for populations at risk for limited mobility.