1/11/2017 | 2 MINUTE READ

No Way—Waymo! Google Accelerates Self-Driving Vehicle Program

Facebook Share Icon LinkedIn Share Icon Twitter Share Icon Share by EMail icon Print Icon

Google parent Alphabet Inc.’s new Waymo unit is taking its self-driving vehicle program to the next level. After more than seven years of research, development and testing, the technology has reached an “inflection point” that promises to bring fully autonomous driving much closer to reality, asserts Waymo CEO John Krafcik.

Share

Facebook Share Icon LinkedIn Share Icon Twitter Share Icon Share by EMail icon Print Icon

Google parent Alphabet Inc.’s new Waymo unit is taking its self-driving vehicle program to the next level.

After more than seven years of research, development and testing, the technology has reached an “inflection point” that promises to bring fully autonomous driving much closer to reality, asserts Waymo CEO John Krafcik. The company (formerly Google Project X) has announced several new initiatives and partnerships in recent weeks, and the rate of advancement is accelerating.

Waymo has accumulated 2.5 million miles of autonomous driving tests and another 1 billion computer-based “virtual” test miles involving special and difficult driving conditions. This month it will begin tests of its next-generation autonomous driving system on public streets in Arizona and California.

The new round of tests will involve at least 100 specially equipped Chrysler Pacifica minivans fitted with Waymo’s most advanced technologies and software control system. This includes eight multi-sensor vision modules and an ultra-high-definition camera. The array provides a 360° view around the vehicle. A separate radar system complements the cameras with its own 360° view.

The sophisticated array helps a vehicle to “see” clearly in rain, fog, snow and darkness, Waymo says, including detecting and interpreting such subtleties as a bicyclist’s hand gestures. It also can determine which way stationary pedestrians are facing and thus anticipate which way they may move, according to the company.

The vehicle’s next-generation 3-D lidar system alone can identify an object the size of a football helmet 200 yards away. Krafcik says the device costs about $7,500—one-tenth that of the pricey lidar system Google used on its first-generation self-driving car in 2009.

Waymo’s next-generation system uses artificial intelligence to integrate input from the various sensors. The result, Krafcik says, give the vehicle a far sharper and more sophisticated look at its surroundings than was possible with earlier Google self-driving vehicles. The technology has helped Waymo cut the number of mistakes—logged as “disengages” in which test drivers must assume control—made during autonomous driving by 75% to 0.2 times per 1,000 miles.

Waymo says its system also is better protected against cyber attacks because it operates mostly offline. The Waymo vehicle only connects to the cloud briefly when it needs such information as traffic updates before logging off again. In addition, the ability to operate on its own for long periods makes Waymo’s technology more reliable, Krafcik claims, noting that this would allow a vehicle to continue to function if an external communication network is disrupted or breaks down.

But Krafcik reiterates that Waymo’s aim is to market its integrated hardware and software system, not to build its own automated cars. The goal, he says, is to build a better driver rather than a better car.

In addition to Fiat Chrysler, Waymo is talking with Honda about a potential partnership under which the carmaker would provide vehicles to Waymo to equip with its technology. Honda R&D engineers based in Silicon Valley and Japan would work closely with Waymo engineers in Michigan and California, similar to the co-location of Waymo and Fiat Chrysler engineering teams in Michigan.

Thus far, Waymo has designed and built most of the sensors and computers used in its automated vehicles. Having the company’s hardware engineers and software developers work together on the entire system has helped drive costs down and improve operating efficiency, Krafcik says. But to make the platform affordable for affordable and accessible for consumer applications, he concedes, likely will require a production partner.