Yahoo Autos This Self-Driving Truck Learned How to Drift to Make Autonomous Cars Safer
Technology like Tesla's Autopilot may have consumers thinking otherwise, but autonomous cars have a long way to go before they can replace good old human drivers. Among other issues, computers still aren't the smoothest of drivers. Nor are they very good on loose surfaces. But technology developed at Georgia Tech may help to fix that.
Students on the school's AutoRally team have built a 40-inch-long, one-fifth-scale pickup truck that can hit a top speed of 60 mph. They then outfitted it with rotation sensors on each wheel, two front-facing cameras, a GPS sensor, an inertial measurement unit, and a quad-core computer with 32 gigabytes of RAM and an Nvidia graphics processor. All that hardware allows the truck to read the road ahead and make incredibly quick decisions based on the data it collects.
But instead of attempting to calculate a single trajectory, the Georgia Tech truck calculates more than 2500 possible trajectories for its next few feet of travel, and then computes a weighted average of all of them. The result is an autonomous truck with the ability to drive smoothly around the track and learn as it goes.
Oh, and it has the ability to stay under control even in a power slide.
But while the idea of a tiny autonomous rally truck is super cool, the bigger deal is how this technology could allow an autonomous vehicle to stay in control even if it loses traction. There are still plenty of challenges self-driving cars will have to overcome before autonomous cars can do their jobs without human help, but this technology shows what's possible when you teach a robot to drift.
