/>
X

Robot helicopters flying low among obstacles

According to New Scientist, engineers at Carnegie Mellon University (CMU) have modified an unmanned commercial civilian helicopter to fly fast and low while avoiding obstacles such as buildings, trees or power lines. The unmanned aerial vehicle (UAV) from Yamaha has been adapted to integrate a sensing system able to see obstacles -- and to avoid them. The article said that 'the helicopter's eye is a custom-built 3D laser scanner, which sweeps an oval path ahead of the 3.5-metre long craft. The scanner can detect objects as hard to see as power lines from 150 metres away.' But read more for interesting discoveries about this project...
zd-defaultauthor-roland-piquepaille.jpg
Written by Roland Piquepaille, Inactive on

According to New Scientist, engineers at Carnegie Mellon University (CMU) have modified an unmanned commercial civilian helicopter to fly fast and low while avoiding obstacles such as buildings, trees or power lines. The unmanned aerial vehicle (UAV) from Yamaha has been adapted to integrate a sensing system able to see obstacles -- and to avoid them. The article said that 'the helicopter's eye is a custom-built 3D laser scanner, which sweeps an oval path ahead of the 3.5-metre long craft. The scanner can detect objects as hard to see as power lines from 150 metres away.' But read more for interesting discoveries about this project... (Update: November 16, 2008) CMU professor Sanjiv Singh who led this project sent me some interesting details about how his research was reported. You'll find some of his comments below.

CMU low-flying robot helicopter #1

You can see above a picture of the helicopter platform used by the CMU researchers to test collision avoidance while flying autonomously in between two poles 10 meters apart. The rotor span of this helicopter is 3.4 meters. (Credit: CMU)

CMU low-flying robot helicopter #2

And you can see above the helicopter used by the CMU researchers for their experiments, a Yamaha RMax, which has a payload of 29 kilograms besides fuel. (Credit: CMU) This project has been led by CMU professor Sanjiv Singh.

Here is another quote from the New Scientist article about the vision tools used by this helicopter. "The helicopter uses two navigation strategies. First, a long-range planning algorithm uses an existing 3D map to work out a general course that avoids large obstacles like buildings and trees. That map can be preloaded, or built up by the helicopter as it explores a new area. When the aircraft flies a route, its scanner looks out for other obstacles. As these appear, a local planning system takes over and plots a detour. The UAV can fly between two obstacles with only around 3 metres clearance on each side."

This research work has been published in The International Journal of Robotics Research under the name "Flying Fast and Low Among Obstacles: Methodology and Experiments" (Volume 27, Number 5, Pages 549-574, May 2008)

Here is the abstract. "Safe autonomous flight is essential for widespread acceptance of aircraft that must fly close to the ground. We have developed a method of collision avoidance that can be used in three dimensions in much the same way as autonomous ground vehicles that navigate over unexplored terrain. Safe navigation is accomplished by a combination of online environmental sensing, path planning and collision avoidance. Here we outline our methodology and report results with an autonomous helicopter that operates at low elevations in uncharted environments, some of which are densely populated with obstacles such as buildings, trees and wires. We have recently completed over 700 successful runs in which the helicopter traveled between coarsely specified waypoints separated by hundreds of meters, at speeds of up to 10 m s-1 at elevations of 5-11 m above ground level. The helicopter safely avoids large objects such as buildings and trees but also wires as thin as 6 mm. We believe this represents the first time an air vehicle has traveled this fast so close to obstacles. The collision avoidance method learns to avoid obstacles by observing the performance of a human operator."

Now, here starts the interesting stuff. The same team of researchers wrote a paper named "Flying Fast and Low Among Obstacles" which was included in the Proceedings of the International Conference on Robotics and Automation of April 2007. And please take a look at the abstract. Do you see a difference with the May 2008 paper? Besides "10 meters/sec" in 2007 being replaced by "10 m s-1" in 2008 and a longer title, there is none that I can see. Apparently, this is how research works are increasingly published these days. [Note: I know that 'cut and paste' is an easy process, but what do you think about this publishing process? Drop me a note to tell me what you think.]

[(Update: November 16, 2008) When I wrote that the ICRA and the IJRR papers had similar abstracts, this was true. But Singh ans his colleagues wrote a 6-page paper for the ICRA conference and submitted a 26-page paper to the peer-reviewed IJRR journal, a process that took a year before acceptation. Here are some comments from CMU professor Sanjiv Singh: "In general in our field, it is acceptable for authors to publish in a conference proceedings and then spend a longer time producing an extensive paper that is a superset for a journal. Journal papers go through a much longer process of vetting and review by our peers. In this case, we submitted both papers simultaneously but you will notice from the date of the journal paper that it took almost a year for the reviewers to be satisfied with our work to pass muster for a journal." My apologies to Singh and his colleagues...]

Anyway, here is a link to the full 2007 paper (PDF format, 8 pages, 692 KB), from which the above pictures have been extracted. Here is one excerpt from the conclusions. "We have developed a first-of-a-kind capability suited for UAVs that avoids obstacles of various sizes and shapes while flying close to the ground at significant speeds. In our experiments, the uninhabited helicopter started with no prior knowledge of the environment, having been provided only a list of coarse waypoints separated by up to hundreds of meters. The straight line path between the waypoints often intersected obstacles. While we regularly ran collision avoidance close to buildings, trees and wires between 4-6m/sec, the system was tested at speeds above 10 m/s. To accomplish these results our system uses a fast avoidance method that stays away from obstacles intelligently coupled with an online planning method that suggests a direction of travel."

And here is what the researchers planned to do in the future -- in 2007. "We intend to address a few issues in future work. Our current method is not built to scale with significant increases in speed and avoids the obstacle with the margin irrespective of the speed. Ideally the reaction should depend on speed to react earlier to obstacles if the speed is fast and later if the speed is slow. Another issue is the tradeoff between sensitivity to small obstacles and an excessive reaction to large objects close by (such as in an urban canyon) even if they are not in the path of the vehicle."

Finally, here is another link to short movies about collision avoidance demos.

Sources: Kurt Kleiner, New Scientist, November 10, 2008; and various websites

You'll find related stories by following the links below.

Related

He flew American Airlines, she flew United. For both, the unthinkable happened
screen-shot-2022-06-30-at-10-14-36-am.png

He flew American Airlines, she flew United. For both, the unthinkable happened

Business
CERN is firing up its Large Hadron Collider at record energy levels, in search of dark matter
cern-photo-202011-145-2.jpg

CERN is firing up its Large Hadron Collider at record energy levels, in search of dark matter

Innovation
Giant data breach? Leaked personal data of one billion people has been spotted for sale on the dark web
close-up-of-a-womans-hands-typing-on-a-keyboard-in-the-dark.jpg

Giant data breach? Leaked personal data of one billion people has been spotted for sale on the dark web

Security