In the future, self-driving cars could orient themselves on the road surface if gps was not available. Researchers at the southwest research institute (swri) have developed a system called ranger, which photographs the floor of the road while driving and recognizes it the next time. It consists of a camera with light-emitting diodes on the underbody of the vehicle and a computer. Ranger was exhibited at the xponential drone show in new orleans this week. "Gps is the achilles’ heel [autonomous vehicles]", said ryan d. Lamb in the conversation with our site, "we wanted to find a different approach."
Lamm heads the research department for applied sensor technology at swri. Under his agide, the researchers kristopher kozak and marc alban developed the ranger and found the matching algorithms. In test drives of various self-driving cars with deactivated gps, ranger was able to control the vehicles with an accuracy of two centimeters. That is measured in the direction of travel; at right angles to the direction of travel, the range of fluctuation is higher.
Leaves do no harm
The system is amazingly robust. It even works with rainwater depths of up to five centimeters, or when leaves or snow cover the road surface by up to 70 percent. The ground may be made of asphalt, concrete or cobblestones, and also unpaved roads ("dirt roads") are not a problem as long as they are not muddy. The researchers give 130 km / h as the maximum speed for a sufficiently fast location determination.
When the road surface is caught for the first time, the cameras take 60 pictures per second. However, not the complete images are saved in the ranger, only features classified as significant. About one mbyte of storage space per kilometer and lane is sufficient for this. When used continuously for recognition and location, ranger takes ten pictures per second.
Because uniform lighting significantly increases the detection rate, leds illuminate the street floor. They draw an average of ten watts. "The lighting accounts for the largest part of the energy consumption", told lamb. An intel i7 cpu is calculating in the prototype, but: "it doesn’t need that. The aim is to use a mobile cpu or gpu, or overlain the above-up to the on-board computer of the vehicle."