Google shocked the world this weekend by announcing that not only was it developing robot car technology, but that its fleet of autonomous cars had already racked up 140,000 miles driving experience. As described in the New York Times, seven converted Toyota Prius’ use laser range finders, cameras, radar, inertial sensors, and high-detail maps to autonomously drive while humans sit behind the wheel and monitor software. While robotic cars have made leaps forward in the past decade, spurred on by DARPA’s Grand Challenge competition, Google’s accomplishment stands heads and shoulders above the rest. The search engine giant’s announcement has fueled enthusiasm across the blogosphere for the technology, and many are hoping for the first time that robot cars could be nearer than we think. They will be disappointed. Google’s venture into autonomous cars may be an epic win, but automotive regulation and government bureaucracy will raise a wall of fails in the future. Video of the Google car is available below. Celebrate the success while you can – the technology may be getting better, but society is not prepared to use it.
Google did many things very right in developing their autonomous car program. Foremost was the gathering of some of the most brilliant minds in robotic driving, as tested by DARPA’s Grand Challenge. Sebastian Thrun, head of the project, was one of the leads in the Stanford Racing Team when it won DARPA’s challenge. He also headed Google’s StreetView project. Before Chris Urmson was ‘on leave’ from Carnegie Mellon to work for Google, he developed the autonomous vehicles that brought the university victory at the Grand Challenge. Michael Montermerlo (who got his PhD in robotics from Carnegie) was the software lead for Stanford’s racing team. Anthony Levandowski made news a few years ago by developing PriBot, a Toyota Prius that drove itself through San Francisco. He also worked on autonomous motorcycles at UC Berkeley. Google’s current robot car seems like a next generation version of PriBot. All in all, Google had just 15 engineers in their robot car project, but they chose the best. That’s why a stealth project could be developed and quickly outperform so many other robotic car endeavors. Brilliant strategy, no doubt about it. You can see Urmson behind the wheel in the Google robotic Prius in the video from NY Times below.
Robert Scoble, aka Scobleizer, actually caught the Google car on video back in January, but didn’t know what it was at the time.
The Google car looks great, and it’s performed well, but it’s likely many years from reaching the masses. From the NY Times:
The self-driving car initiative is an example of Google’s willingness to gamble on technology that may not pay off for years, Dr. Thrun said. Even the most optimistic predictions put the deployment of the technology more than eight years away.
It will simply be a matter of time before autonomous cars have the range of capabilities needed to replace human drivers. Yes, the Google fleet drove down Lombard Street’s curves, it handled the wind on the Golden Gate Bridge, and it dared the cliffs of the Pacific Coast Highway. Yet it hasn’t shown that it can defend itself around drunk drivers, dodge children dashing into the street, or notice that the bicyclist next to the road is signaling to cross into its lane. Human drivers face these problems all the time, even if they regularly make mistakes (1.2 million lives lost each year according to the World Health Organization). Google’s system currently relies on a lead car making detailed maps of the route ahead of the autonomous car’s passing. Robot vehicles still need semi-controlled situations to succeed. Years will pass before the controls needed are whittled down to match the variety of scenarios human drivers face all around the world every day. Even then, such systems will need to be tested and retested, made much more reliable than the computers we use (and crash) on our desktops today.
But let’s look ahead to that time, maybe a decade from now maybe much less, when robot vehicles can perform as well as humans. Already, we’ve seen how the Stanford team is developing a system that can race up Pike’s Peak. When robots can defeat rally car drivers the world will be suitably impressed. I’m sure there will be many exhibitions on NASCAR and Formula tracks everywhere highlighting their skill. It won’t matter much. Even once the robots are ready to drive in real world situations, I still think it will take many more years before we actually see automated cars on the road.
For all that was done right, Google did one major thing wrong – they approached the robot car like it was only a technology challenge. It’s not. It’s also a social-legal one. Those 1.2 million people who lose their lives to car accidents each year mostly have other humans to blame. Drivers are held accountable for the machines they control. Who is accountable for an autonomous vehicle? The Google project had manned test drives, and various means for the human to quickly grab control. Not because the car was making mistakes. It never caused a single accident, though it was rear-ended by a human driver. No, the Google robot cars needed to be manned because California state law, not to mention our sense of scientific ethics, demands a human be responsible for a potentially lethal activity.
When robots are ready to drive for us, there will still be accidents. Much fewer, one hopes, but millions in damages and thousands of lives lost all the same. Who will answer for that loss? The company that designs the robot’s software, the car manufacturer who installed it, or the driver who believed that they didn’t need to pay attention because their car was driving itself?
Toyota just spent millions repairing and recalling cars that occasionally had sticking accelerator pedals. They face ongoing lawsuits, and are likely to be confronted by more, blaming them for collisions. That’s just a single instance of a faulty piece of automotive technology. When human drivers cannot control their cars, the manufacturers face enormous legal consequences. When then will we want to pursue a vehicle that takes away the responsibility of driving from humans? No one could face the legal burden, no matter how safe their autonomous cars could be.
That doesn’t mean that robotic cars will never arrive. I just think they’ll appear in small steps. More stealth, less hype. Already we have systems in place that make driving easier, while never removing humans from the equation. Think about the automated systems already in your car: automatic transmissions, airbags, and anti-lock brakes. We’re adding more all the time. Vehicles have camera systems surrounding their cars to help with parking and to avoid collisions (Google uses commercial versions in their robot car). Some new cars automatically engage the brakes if they detect a slowing or stopped object ahead, and more companies will be adding these accident avoidance systems in the future. Technologies that ‘enhance’ the human driver, or ‘increase safety’ help sell cars, and edge us close to autonomy. The goal is to get people to be safer drivers, to provide automated systems to aid us when we’re about to make a mistake, not to take over the responsibility of driving completely. 100% autonomous vehicles are a legal nightmare…but 50%, 75%? That could be done, maybe sooner than we think, and with happy results. It doesn’t take a fully robotic car to save thousands of lives each year.
If I were to predict how Google’s autonomous car project would really affect our lives, I would point to all the possible applications it could enable that don’t involve the dream of robotic vehicles. Advanced laser range finding and radar sensors can be integrated into modern cars to help with anti-collision braking controls, or to create a warning system for drivers. Highly detailed maps could change the way we drive. Google’s Prius has a voice announce when you approach a crosswalk, or near a turn. Imagine a GPS guidance system that gave you 100% accurate help, and warned you of complex dangers like children that play nearby. There are many different ways in which the Google autonomous car projects could help us drive better.
…and yes, one of those ways will be, eventually, the adoption of fully robotic cars. I do believe that 100% autonomous vehicles will arrive, it will simply take longer than we think. Cars will become more and more helpful, removing more and more of the risks of driving, until automated systems are standard safety features for driving. From there we will make the leap to robotic cars. But there will be legal battles, social mores will have to be changed, and it’s likely to bankrupt at least one major car manufacturer in the process. Years after the robots are ready to drive, we’ll be ready to let them. For now we can applaud Google, and go back to our normal lives. Autonomous driving is not near.