Self-driving cars are trained to operate with extreme caution, but driving situations arise where autonomous vehicles must maneuver at high speeds to avoid collisions. Can these vehicles, equipped with tens of thousands of dollars worth of high-tech sensors and programmed for "normal" driving, make these decisions faster than a human?
The answer may lie with the engineers at Stanford University. They created a neural network that allows driverless cars to perform high-speed, low-friction maneuvers just like race car drivers. Once they perfect the system, they will driverless cars had capabilities beyond human, as it is 94 percent of traffic accidents attributable to human error.
The researchers believe that this is an important step in improving the ability of autonomous vehicles to avoid accidents.
"We want our algorithms to be as good as the best qualified drivers – and hopefully better," he said Nathan Spielberg, a graduate student in mechanical engineering at Stanford and lead author of an article recently published in the journal Science Robotics. "Our work is motivated by increasing traffic safety, and we want autonomous vehicles to work in many scenarios, from normal driving on asphalt to high-speed, low-friction driving in ice and snow." The team used a type of artificial intelligence algorithm called a neural network, which is based on the neural networks in our brains, to create a self-checking system. Neural networks are a type of machine learning where programmers build models that process vast databases of data and look for patterns. These networks are used to power the "brains" of the autonomous vehicle, usually high-performance graphics processors stored in the trunk of each vehicle that control the decision-making process.
The Stanford team trained a neural network with data from 200,000 movement patterns, including test runs on slippery surfaces such as snow and ice. They then took their system to Thunderhill Raceway in the Sacramento Valley to test it. The Stanford team used two cars in their tests: an autonomous one Volkswagen GTI and autonomous Audi TTS.
They say they are encouraged by the results, but stress that their neural network system does not yet perform well enough in different environments and conditions. But they say that through extensive testing and an increase in the driving database, they will also solve the last problems that separate them from a system that will perform better than a human driver.
Source:
theverge.com
