A U-M study found that training self-driving car algorithms on near-miss events, rather than just crashes, significantly improves safety performance. Researchers say this approach could accelerate the development of safer autonomous vehicles and help build public trust in the technology.
