Company Using Video Game to Program Driverless Trucks?
Driverless trucks are a fast-growing technology, poised to change the freight transportation industry forever.
In the U.S., companies have already taken autonomous vehicles on successful test runs. Big names are investing in the movement to move freight without a human in the driver’s seat, and it’s expected that more driverless vehicles will be seen on the roads very soon.
There are always questions about just how safe an unattended machine is, especially when it must operate based on its own capacity to make decisions. Even though some driverless trucks have a driver (and in some cases an engineer) along for the ride, the idea of a computer piloting the rig is an unnerving thought for many.
The ability driverless commercial trucks have to complete trips is one thing, but gaining public acceptance is another matter entirely. Drivers who will share the roads with these vehicles are concerned, as are safety advocates who want to ensure the machines are designed with care.
U.S. companies are making their own progress when it comes to ushering in the autonomous revolution, and they’re using a variety of different technologies. But a startup in South Korea is using something very unique to train its truck artificial intelligence systems – the popular PC game “Euro Truck Simulator 2.”
Mars Auto has spoken about the program they plan to test on the road, stating they’d like to test it in a game first.
The decision may seem like an odd one, but Mars Auto isn’t the first company to take this approach. Many companies in the self-driving field use a game known as “The Open Racing Simulator” to develop their AI effectively.
The premise is simple – if a program can learn to analyze risks in a simulation, it should be able to perform the same task in a real-world situation. Using cameras and sensors to gain information on the road, the AI program can process nearby hazards and adjust accordingly.
Mars Auto isn’t stopping their tests here – their next step is to build their own game, creating a simulation of radar and sonar signals to help train their self-driving truck software even further. While deep learning has a long way to go, test programs like these can help it develop quickly.