Ford to Make Cars Smarter With Google Prediction

Ford Motor Co. wants to use the Google Prediction API to predict driver
behavior and use that input to make cars perform better.

The company's goal is to use cloud-based storage and computing to collect
and process information about how drivers use their vehicles. Accessing
those resources over a wireless network, a vehicle could automatically
change how it performs, according to Ford.

The Google Prediction API, released last year by Google Labs, is designed
for giving an application information about past user actions so it can
better predict future behavior. The API (application programming interface)
uses pattern-matching and machine learning and is designed to be simple to
use, according to Google. On its information page about the API, Google said
it could help a website recommend products to regular visitors, detect what
a user would consider spam e-mail, and predict how much a user would be
likely to spend on a given day.

Ford is extending this technology from the Web to the road. On Tuesday at
the Google I/O conference in San Francisco, Ford will present one conceptual
case for the use of this technology in a plug-in hybrid electric vehicle. If
the owner started the car on a weekday morning, the car could use its recent
driving history to predict that this will be a trip to the office. An
automated voice would ask the driver to confirm this, and if the driver
said, "Yes," the car could prepare to handle the route to the office
intelligently.

In the example presented at Google I/O, the hybrid could favor its gasoline
engine on the first part of the trip to set aside battery power for an
electric-only section of road that it knew was coming up later on the route.
But knowledge about location, routes and driver behavior could also help a
car in other ways, according to Ryan McGee, a technical expert at Ford. For
example, the vehicle could shift to full electric power while close to the
driver's home, in order to reduce noise and pollution. A car with access to
the predictive tool could also use different techniques to save energy for
later, depending on an individual's style of driving, McGee said.

A large amount of information about a driver's preferences and driving
behavior could be stored in a profile in the cloud and then accessed when
that driver got in the car, according to Ford. The company emphasized that
drivers would have to opt in to such a service so none of this data would be
collected unless they agreed to it. The profile would also be encrypted to
protect the information from unauthorized use.

Ford doesn't expect smarter cars to start changing drivers' behavior by
throttling back their acceleration or limiting their speed to save energy,
McGee said. Rather, the idea is to help the car program itself to adapt to
each driver's style.

The company won't discuss any plans for car features or services based on
this technology, and due to logistical issues, its software demonstration at
Google I/O doesn't even use an actual vehicle. But McGee said products based
on this type of intelligence might reach the market some time after 2015.