Autonomous cars rely on data rather than knowledge of the real world like human drivers so it is essential to provide the machine learning systems with vast amounts of training data. Waymo uses "fuzzing," which adds random variations to the same simulation to test and train the machine learning system's response to confusing situations. There are currently close to 400 people using Waymo's autonomous vehicles as taxis in Arizona.
There is a big gap however, between the driving environment in Phoenix, Arizona and cities like San Fransisco or Boston. Being able to address the unexpected situations that arise from more chaotic or less temperate places is the next hurdle for Waymo and other autonomous driving companies.
Click here to read the full article.