Google’s AutoM8s

by Dave Michels

I try to keep this blog telecom focused, but sometimes other tech news insists on inclusion.

In the hi-tech book Daemon, by Daniel Suarez, villains use a variety of connected technologies to create havoc and control wealth. One of those tools were AutoM8s – impressive unmanned cars. In one pursuit, a pack of AutoM8s coordinate their activities to protect a VIP passenger with high speed coordinated defensive moves as well as offensive strikes to take-out other vehicles and create a trail of obstacles to block the pursuers. These high-speed robotic cars were tightly coordinated (inches apart, perfect timing) at a level impossible for human drivers to mimic.

That was fiction.

Last week, Google unveiled the AutoM8 prototype, A secret fleet of self driving cars. Finally shedding some light on a recent Schmidt quote: “It’s a bug that cars were invented before computersYour car should drive itself. It just makes sense.”

Really? Really!

According to the New York Times, Google equipped seven cars that drove a total of 140,000 miles with minor human intervention. The cars drove up to 1000 miles without human intervention. The cars were equipped with multiple sensors, cameras, and lasers that streamed information to a data center for real-time analysis for subsequent computer controlled instructions. Each vehicle was also equipped with two humans; an emergency standby driver and a software engineer. The only accident was one of the Google vehicle’s got rear-ended.

I recently lost a 93 year old relative and we were discussing all the things that changed in her lifetime. We spoke about world wars, computers, microwaves, jet travel, space travel, and so much more, but what really impressed the kids was cars. Technically the automobile goes further back, but mass production caught hold in the 1920s. The modern automobile is an amazing device, yet outright primitive when sitting in gridlock traffic. So eloquently captured in the opening scene of OfficeSpace – the man with the walker moving faster than traffic.

Google points out 1.2 million people are killed every year in road accidents. A number they think can be cut in half with technology. Robotic driving can also reduce energy consumption, time, congestion, and parking shortages. 1.2 million is a big number, though they didn’t mention how many people are full time employed as drivers – truckers, taxi drivers, delivery, etc. Nor did they mention why the number could be cut in half instead of eliminated.

There is certainly merit to robotic driving. Humans really are terrible drivers. The problem is the amount of inputs are far greater than we can really manage (though few admit it). At any given time we are expected to have our eyes on the road as well as the mirrors as well as know the current speed limit and actual speed plus other dashboard feedbacks, all while estimating the trajectory of other moving objects (vehicles, pedestrians, bicycles, animals, etc.), and communicating intentions (signals, eye contact, the bird) and that’s the minimum – throw in passengers (a crying baby), a hamburger, a CD to change, a phone call to make, blinding sunlight, a gorgeous pedestrian, mud or snow, a dirty windshield, a personal problem, and countless other distractions and the 1.2 million deaths seems like a small number. Conversely, computers can manage all these inputs. I think my car checks the status on my ABS breaks something like 30 times a second? The challenge is the ever elusive goal of teaching a computer to see.

Computer vision is a difficult problem to solve. Motion, temperature, and many other stimuli are easy to sense, but teaching a computer the difference between a small dog and a big cat is non trivial. Accurately identifying road conditions and hazards at high speeds gets near impossible. At least I thought.

Ten years ago, in-car navigation was thrust upon an unsuspecting public – an absolutely inconceivable and marvelous invention (I am still amazed). Then the mobile Internet took it to a newer level with real time traffic information. It turns out we know quite a bit about the roads, ramps, and highways. We know things like speed limits, congestion, even construction in some cases. Video recognition is dramatically improving – lane markings, stop light status, other vehicles, are all pretty easy to recognize. Enhanced sensors can identify where the shoulder starts and road ends better than the driver, or the exact distance between cars. Factor in the notion of cloud computing and 4G speeds – vast amounts of data and processing cheaply available and the Google project moves into the realm of non-fiction.

“Cars are where black-and-white televisions were in terms of their level of possibility or electronic development,” Toyota’s Peter Evans says. “We’ve still got the automotive equivalent of colour, plasma, flat screens and 3D to go, in terms of electronics, in the car. “Think about a modern fighter aeroplane, which is about extending the capability of the pilot, and that’s where the car is going,” Evans says, citing more advanced active cruise-control and lane-departure systems that will be capable of keeping a car tracking on a freeway with minimal or no driver intervention.

Mercedes, Volvo, Lexus even GM are already offering various forms of computer assistance – self parking cars, automatic braking, and more. But is this technology really going to come from the auto makers… probably not. Robot driving is a software challenge – and it was very interesting to me to see that Google (of all companies) is developing this. From the Google blog:

So we have developed technology for cars that can drive themselves. Our automated cars, manned by trained operators, just drove from our Mountain View campus to our Santa Monica office and on to Hollywood Boulevard. They’ve driven down Lombard Street, crossed the Golden Gate bridge, navigated the Pacific Coast Highway, and even made it all the way around Lake Tahoe. All in all, our self-driving cars have logged over 140,000 miles. We think this is a first in robotics research.

Robotic driving really does pose a huge disruptive threat – one that is probably overdue. Safety is the motivator but the social ramifications are huge. Instead of airport parking, just send the car home. Traffic jams would be eliminated. Speeds could be dramatically increased, texting while “driving” will be cool again.

Though don’t get too excited – it won’t be happening anytime soon. This type of disruptive change will come gradually at best – with a fair amount of controversy. To this day, commercial airplanes still have pilots, but the military uses less and less of them. The concerns from the public and driving related unions would be significant – and that’s if the technology works.