Last month, Ford unveiled its hybrid research vehicle, a Ford Fusion designed to help test automated driving, to a crowd of international reporters at its headquarters in Dearborn, MI. Last week, Ford announced it will partner with researchers at Stanford and MIT to continue its work developing automated vehicle technology.
Greg Stevens, global manager for driver assistance and active safety research at Ford, says the goal is to teach computers the habits of exceptional human drivers. Stanford and MIT have research strengths that match the challenges Ford is trying to address before it can bring automated vehicle technology to market, Stevens adds.
At Stanford, researchers will focus on how a vehicle might use sensors to “see” around objects to predict what’s in a car’s path. “Automated driving means constantly checking sensors, but we also want to be aware of what we can’t see,” Stevens says. “The ability to peek around a truck in front of you is useful because if the truck slams on its breaks, it’s good to know the lane change area is clear.”
Ford’s goal is to develop an algorithm that will teach a vehicle’s computer what to do if sensors are blocked. Stevens says Stanford has a lot of expertise in researching vehicle dynamics and the habits of expert drivers. Stanford talent has been tapped for automated vehicle research before. The team behind Stanford’s entry in the DARPA Urban Challenge, led by Xconomist Sebastian Thrun, also assisted Google with its driverless car initiative.
At MIT, researchers will focus on predicting the behavior of other vehicles on the road as well as pedestrians. “Sensors on the vehicle can see where objects around the vehicle are, but they also need to be able to predict where they’ll be in the near future to plan a path to avoid them,” Stevens says. “As humans, we do that all the time because we’ve built up models in our head. We want to build those models in computers using advanced algorithms.”
Essentially, there are three main issues Ford wants to examine in partnership with the Stanford and MIT researchers: the natural limitations of vehicles, teaching the car’s onboard software to pick up on cues from other drivers, and learning to detect situations like an approaching exit ramp so the software can predict why a neighboring car might be suddenly trying to switch lanes.
“You find that teaching computers to do things that computers are good at is not that hard,” Stevens explains. “It’s teaching computers to do things that drivers are good at that’s complicated.”
Stevens expects that some Ford models will offer automated parking and lane-change assistance within five years as part of its overall Blueprint for Mobility. Active park assist, which uses sensors to help drivers angle into parallel parking spots, is already available in about a dozen models. Within five years, the car will be able to handle all the steering, shifting, and braking needed to park anywhere.
Soon after that, Ford plans to introduce remote parking assistance technology, where a driver would be able to pull up into the driveway, step out of the vehicle, press a button on the key fob, and let the car figure out how to squeeze into that tight
