It is the best of environments and the worst of environments for designers of autonomous off-highway vehicles. Freight haulers and some agricultural equipment are in some ways easier to automate than passenger cars, but many construction machines pose extremely vexing challenges.
Mining companies have been using autonomous trucks for a few years, with a number of users saying efficiency gains average 15-20% over conventional trucks. In many ways, hauling freight from mines to ports or rail depots is far simpler than autonomously driving trucks on highways.
“The mine might be 20 miles from the rail head,” said Dan Williams, ADAS and Autonomy Director at ZF Commercial Vehicle Technologies. “The driving path is very predictable—they run the same route every day, often in a remote environment where there aren’t a lot of intersections and they don’t have to deal with many drivers or pedestrians.”
Passenger cars usually drive forward from point to point, so most of their sensors face forward, with comparatively few facing rearward. But any construction vehicle equipped with a backup alarm requires extensive sensing technologies on the front and back to provide autonomy.
“While off-highway machines are trying to get from one spot to another, they’re also performing work functions,” said Chris Woodard, Business Development Manager for Autonomous Machines, Danfoss Power Solutions. “You may have a wheel loader that’s transporting material or an excavator that’s digging. Or, all the work may be happening behind the machine, like planting or tilling. These all require different sensor systems and configurations. In short, the flexibility and adaptability of autonomous systems for the off-highway market is much higher than in the automotive industry.”
Another difference is that while off-highway vehicles may need more sensors, they may not need the high resolutions used for autos. For many tasks in construction and agriculture, distances are only a few meters and fewer objects are important, so the high-resolution cameras used in cars aren’t necessary.
“We’ve worked with lower-resolution cameras as well as different filtering technologies—infrared, for example—to pick out important pieces of data, as well as high-resolution cameras that can pick out the finer details of objects” Woodard said. “The trade-off is processing requirements, which impacts the processor cost.”
“We’re working with different types of cameras and backend processing for different tasks, so there isn’t a ‘one size fits all’ solution. The type of application can determine which configuration works best. For example, a lower-resolution camera is used for sprayers or mechanical weed removal to differentiate between a plant or a weed simply based on the area it takes up,” he added.
However, Woodard noted that as automotive volumes for high-resolution cameras rise, it’s likely that prices will fall to levels that make high-resolution cameras viable for off-highway systems. That will increase computing requirements, but processor prices are also likely to decline over time.Continue reading »