Inside Unmanned Systems

FEB-MAR 2018

Inside Unmanned Systems provides actionable business intelligence to decision-makers and influencers operating within the global UAS community. Features include analysis of key technologies, policy/regulatory developments and new product design.

Issue link: https://insideunmanned.epubxp.com/i/945861

Contents of this Issue

Navigation

Page 58 of 67

ENGINEERING. PRACTICE. POLICY. 59 February/March 2018 unmanned systems inside If all that is needed next is a rough idea of where the unmanned vehicle is, a coarse-positioning system the researchers have developed runs the sequences of images past a matching algorithm. "The vehicle will have sur- veyed the mine beforehand, and during operations, it will match the sequences of images it sees with this reference map, just like a person would," Milford said. "This system has an accuracy down to a few meters." If one wants a better idea of where the unmanned vehicle is, an accurate- positioning system the researchers are developing "uses slightly more sophis- ticated visual-processing techniques," Milford said. "It could give you accura- cy down to a few centimeters in terms of position." The coarse-positioning system is in- tended to optimize mine sites by help- ing people track mobile assets, while the accurate-positioning systems could enable autonomous operations of unmanned vehicles. "Not all com- mercial clients are interested in both capabilities—the coarse-positioning system could use much cheaper cam- era systems and computer systems, since it's much easier to accomplish than the accurate-positioning sys- tem," Milford said. Testing So far Milford and his colleagues have carried out three tests with their systems on unmanned vehicles in several underground mining sites in Australia many hundreds of meters deep. "The first tests were just with four-wheel-drive vehicles, but we've recently beg un exper iments w ith load-haul-dump trucks that were 10, 20, 30 tons in size One unexpected problem the re- searchers ran across during testing were the low roofs of some mine tun- nels. "The roof can be so close to the camera that any small movements can completely change what the camera is looking at," Milford said. "Imagine you walk up to the wall of a room until you're six inches away and you can only see one patch of the wall, and then you step one foot to the right and now you have a completely different view. For a camera-based positioning system to work, what the sensor sees has to over- lap with what it has seen before in its mapping run, and if the vehicle, say, goes over a bump that it didn't before, that can change what the camera sees and the system may not necessarily think it is following the same path that it was before." The researchers are now developing techniques to compensate for this problem. The researchers hope camera-based navigation systems may prove cheap- er, more reliable or more f lexible than current alternatives. Still, while they hope cameras can become the main sensors of underground unmanned vehicles, they noted such vehicles would still likely employ lasers in some capacity. What This Means for the Industry " The primar y advantage of lasers is that they are almost infallible in terms of things like collision avoid- ance," Milford said. "A laser is also potentially useful in other mining- related applications, such as in dig- ging, where one can help measure the three-dimensional volume of a mass of earth." "I f ind it very exciting that their brain-inspired algorithm can gener- ate state of the art results as good or better than other existing algorithms," said Jeff Krichmar, a professor of cog- nitive science and computer science at the University of California at Irvine. He suggested using this biologically inspired algorithm makes sense un- derground "since their algor ithm doesn't rely on GPS and sensors that wouldn't work in this harsh environ- ment. It's very impressive." "It is a great success story for our f ield of neurorobotics," K richmar added. "It shows that a brain-in- spired a lgor ithm can outper for m existing solutions. This suggests that brain-inspired algorithms could po- tentially be used to solve other real- world problems." These systems are under devel- opment as part of a two-year col- laboration between the Queensland University of Technology, industry giant Caterpillar, technology trans- fer organization Mining3 and the Queensland government. "The coarse-positioning system has been reasonably well-prototyped and tested, and our partners are talking about possibly commercializing it," Milford said. "The accurate-position- ing system is firmly in the research and development stage. It should take another six to 12 months to mature it up before handing it off to an industry partner." "IT IS A GREAT SUCCESS STORY FOR OUR FIELD OF NEUROROBOTICS. IT SHOWS THAT A BRAIN-INSPIRED ALGORITHM CAN OUTPERFORM EXISTING SOLUTIONS." Jeff Krichmar, professor of cognitive science and computer science, University of California at Irvine

Articles in this issue

Archives of this issue

view archives of Inside Unmanned Systems - FEB-MAR 2018