Toyota Reinvesting in Collaborative Safety Research

Working with a host of research partners, Toyota is continuing its investment in generating safety data, which it shares widely.

Over the last decade Toyota’s CSRC has collaborated with a host of research partners, including the National Advanced Driving Simulator at the University of Iowa. (Toyota)

To adapt to a swiftly changing mobility ecosystem, Toyota announced on April 27 a new five-year, $30-million investment in its Ann Arbor, Michigan-based Collaborative Safety Research Center (CSRC). Created in 2011, the CSRC focuses on foundational safety research, and the $85 million apportioned over its first ten years funded research including the factors that lead to distracted driving, and developing tools and testing procedures related to advanced driver-assistance systems (ADAS).

Since its inception, CSRC has completed 85 research projects with more than 30 different institutions, the output of which Toyota shares globally in the name of shared safety. “The foundation of CSRC is built on collaborations to tap outstanding safety researchers and institutions throughout the country,” said Danil Prokhorov, director of Toyota’s CSRC and Future Research Department (FRD). “We’ll also continue to publish our CSRC research discoveries for others, to help foster societal benefit.”

The nine new projects will engage the expertise of: University of Massachusetts – Amherst; Children’s Hospital of Philadelphia; University of Michigan Transportation Research Institute; Massachusetts Institute of Technology; University of Iowa; Virginia Tech; and Indiana University – Purdue University Indianapolis. The newly funded research will involve a wide spectrum of safety issues, including:

  • driver training
  • vulnerable road user (VRU) interactions
  • SAE Level-2 “non-driving” activities
  • post over-the-air (OTA) feature-update education
  • intersection evasive-action trends
  • predicting bicycle and e-scooter behavior
  • considerations in human anatomical variation when assessing post-crash injuries

“We are aligning with the next emerging trends,” explained Jason Hallman, senior manager for the CSRC. “Talking about safety systems integrations, we're really asking what future crashes will remain and what new opportunities will be presented as we have an evolution and a convergence of active and passive safety systems. We're also asking what new tools are needed as automation becomes a bigger component of driving today. Particularly, as we think about future automated vehicles and technology, what tools do we need to evaluate passive safety.”

Mental models

To help characterize the types of diverse research funded and supported by the CSRC, Toyota presented some of the findings completed in its last five-year funding program, including a fascinating study on driver “mental models.” Kicked off in 2017, the multi-year project studied the interaction between vehicle safety systems and the mental images users of the technology generate in response to the technology.

“In this research, the concept of mental model was very broadly applied to include not only drivers' understanding of how the driver-assistance technology works, but how drivers interact with and relate to the systems,” noted psychology Ph.D. James Jenness, senior researcher in Westat's Transportation Research Group. “This included knowledge of driver-assistance functional capabilities and limitations, understanding of the underlying technology, subjective feelings about the technology, trust in the systems and the drivers’ metacognition about their own understanding – how confident are they that they know everything they need to know about how the systems work?”

Jenness noted some of the goals of the research included forming queries about how driver mental models form, how they change over time as drivers get more exposure to the technology, and how drivers’ mental models differ. “We conducted focus groups to learn how drivers typically think and talk about their driver-assistance systems,” Jenness said. “Some of the candidate factors were complexity, functional accuracy, anthropomorphism and emotional surveillance.”

“One of our findings is that analogs or analogies are a good tool for learning about drivers’ mental models,” he continued. “Self-generated or self-selected analogs, such as ‘elderly aunt’ or ‘teacher’ are an easy way for drivers to describe how they think about their vehicle technologies while revealing characteristics of their mental models. We found that learning about driver-assist systems happens most intensely during the first 10 weeks of ownership, but many drivers continued to learn new things over six months or more.”

The study identified five types of learners, with Jenness noting the most interesting being the “misinformed” group. “They don't know much about their systems, but they tend to be confident that they do know everything, and yet they have less interest in using the technology,” he pointed out. “They were more likely to identify with an analog such as robot, while the expert and skilled learners were more likely to think of their systems as an assistant.”

Jenness said design personas based on cognitive learning and personality variables may be useful for understanding drivers' mental models of vehicle technology. This approach, he noted, could be beneficial for supporting consumer education and user-interface (UI) design. “By better understanding drivers' mental models,” he said, “designers can create systems that are more intuitive or easily learnable.”