Voyage, the self-driving car spin-out from Udacity, is open-sourcing its approach to autonomous driving safety. This comes at a time when autonomous driving programs are under intense scrutiny following two fatal crashes — one involving Tesla’s Autopilot and the other involving one of Uber’s self-driving cars in Tempe, Arizona. Meanwhile, Voyage has successfully deployed five Level 4 self-driving vehicles in retirement communities in California and Florida.
Dubbed Open Autonomous Safety, the initiative aims to help autonomous driving startups implement better safety-testing practices. Companies looking to access the documents, safety procedures and test code can do so via a GitHub repository.
“Each and every autonomous vehicle startup today has to define their own safety programs, and we think that is dangerous,” Voyage CEO Oliver Cameron tweeted earlier today.
Version one includes scenario testing, functional safety, autonomy assessment and a testing toolkit. Later this year, OAS will release driver training material, additional scenarios and fault injection code and tests.
Here’s a quick breakdown of what the above currently entails:
- Scenario testing: Looks at fundamental questions, like how self-driving cars behave around pedestrians and when cars back out of driveways.
- Functional safety: Helps to ensure safety without a driver present.
- Autonomy assessment: Validates whether or not car is moving in the right direction “and how we know that we are solving the right problems,” Cameron wrote in a blog post.
- Testing toolkit: A library of traffic, roadway and vehicle assets.
“When it comes to safety, we believe open is better. At Voyage, we welcome contributions to improve OAS, like any other open source project,” Cameron wrote in a blog post. “The purpose of this effort is to promote an elevated standard of safety in the autonomous vehicle industry, increasing public trust through transparency.”