How much safer would driving be if robots replaced humans on the roads? NY Times looks at this story.
It has been hard to estimate because fully autonomous cars are not yet available to test. Google says that its driverless cars have logged more than 700,000 miles without an accident caused by the car, and that its cars do not do unsafe things that people do, like sharply accelerating or braking.
But two studies by researchers at Virginia Tech — H. Clay Gabler, a professor of biomedical engineering, and Kristofer D. Kusano, a research associate — suggest how much safer robot cars might be. They found that even cars that are not fully autonomous but that automate some of the most dangerous aspects of driving could have as big an effect as seatbelts have had.
The studies, which were sponsored in part by Toyota Motor, analyzed the crashes, injuries and fatalities that could have been prevented by cars thatalert drivers when they drift out of their lane or correct the car’s course, and those that sense an impending collision and automatically brake. They used a representative sample of real-world crashes nationwide and simulated what would have happened had the automation been in place.
They found that lane-departure warning systems would have prevented 30.3 percent of the crashes caused by lane drifting, and 25.8 percent of the injuries. Rear-end and collision warning systems and automatic brakingwould have prevented only 3.2 percent to 7.7 percent of crashes, but would have reduced their severity. The number of people injured or killed would have declined in the range of 29 to 50 percent, the researchers concluded.
Cars with no human involvement at all, like those Google is making, would theoretically take even more of the human error out of driving. They have other drawbacks, though, like the specter of robot error.