Ford, Google Partner To Boost Car Efficiency

Using the Google Prediction API and cloud services, Ford could optimize the performance of its hybrid electric vehicles.
Google could help your car operate more efficiently if experiments by Ford researchers come to fruition.

Ford and Google spoke about the work they have been doing together in a presentation at the Google I/O conference Tuesday. Ford is experimenting with using the Google Prediction API to apply predictive analytics to maximizing the efficiency of a vehicle--particularly a hybrid or electric vehicle designed to wring the most performance out of the least investment of gas or electric power.

"We're really talking about making your car behave better with some knowledge of how it was used in the past," Ryan McGee, an architecture and algorithm supervisor at Ford who works closely with the hybrid and electric vehicle programs, said in an interview. Ford is not making a product announcement, but has proven the concept with some of its prototype vehicles, he said. "We think the technology could be ready in the four to eight year timeframe, but that's not a plan, more of a feasibility judgment."

Travis Green, Google prediction product manager, said this is the same sort of machine learning software used for things like determining what movies you might like, based on past behavior. Given a set of historical data with known outcomes, the Prediction API can identify patterns and predict "that the car, in these conditions, should behave this way."

Ford already offers cloud-based services through its Ford SYNC in-vehicle information system, but to date it's been applied mostly for entertainment, navigation, and traffic services "to empower the driver," Johannes Kristinsson, a system architect for Ford Research and Innovation said in a statement. "This technology has the potential to empower our vehicles to anticipate the driver's needs."

In one scenario outlined by Ford engineers, the vehicle might ask the driver "Are you going to work?" and take his or her answer into account, in addition to monitoring the actual movements of the car and real-time information such as traffic conditions. Google's data centers would provide the cloud computing power to do the analysis on that data, and then recommend an optimized route that includes factors such as taking advantage of electric vehicle-only lanes. Beyond providing navigational guidance, these analytics would feed changes in drive train performance, as controlled by an on-board computer. A plug-in hybrid gas-electric car might use more gas and conserve electric power during one part of the trip so as to be able to operate in electric-only mode in those EV-only lanes.

Because Google cloud computing provides more computing power and data storage capacity than could reasonably be packed into a car, the Google Prediction API should be able to produce a much superior optimization model for vehicle performance, taking into account historical data for traffic patterns and engine performance.

McGee said most of the intelligence for deciding how the vehicle operates will stay within the vehicle, which is where Ford's software development effort will be concentrated. "We take the data and put it up in the cloud, and the Prediction API builds the model for us based on that data. Our focus is really, given this information, how do we make the best use of it," he said.

The application may be able to figure out the user's route and destination on its own much of the time, but not all the time, Ryan said. "Customers typically have patterns about how they go home to work, and come home to work. How I go to work may be the same every day, but where I go to lunch may be different." So the Ford software engineers are playing with concepts where the vehicle might prompt the user to choose among two or three predicted destinations to see if it has guessed right, he said.

In the wake of recent controversies over location tracking in Apple and Google phones, Ford's press release took care to say the company will respect user privacy in the way it uses location and vehicle data in its application design.