From IEEE Spectrum:
One of the ways in which robots are just starting to get really useful is with hauling aerial cargo. Last year, the optionally-manned KMAX made its first autonomous cargo delivery in Afghanistan, and since it can fly as many missions as you have fuel to keep it going, it's definitely a safer and more efficient way to get supplies to troops, especially in dangerous areas.
To move cargo around, helicopters (autonomous or otherwise) often carry stuff slung beneath them on long ropes, and as you can probably imagine, said cargo often ends up doing all sorts of swinging about, especially if the helicopter that's carrying it has to maneuver. Researchers from the University of New Mexico have been developing algorithms that allow robots to compensate for motion-induced swinging of suspended loads, and testing them out on real live quadrotors.
Essentially, what the quadrotor is doing here is dynamically adjusting its trajectory to damp out the swinging motion of its cargo (think tacos). It's sort of like an upside-down version of pendulum balancing, with maybe a little bit of this insane hinged stick balancing thrown in for good measure. Next, the researchers plan to see if they can get their algorithms to work on platforms that are less balanced (and more realistic), which (they say) should be "an important step towards developing the next generation of autonomous aerial vehicles."
Trajectory Generation for Swing-Free Maneuvers of a Quadrotor with Suspended Payload: A Dynamic Programming Approach, by Ivana Palunko, Rafael Fierro, and Patricio Cruz from the University of New Mexico, was presented last month at the 2012 IEEE International Conference on Robotics and Automation in St. Paul, Minn.
[ MARHES ]
Comments
hao
John, there is a huge difference between "crane" applications, where wind is the primary reason for swing and more agile flights, like longer distances, where the acceleration/decelleration of the object and the control become much more complicated. "Pilot induced oscillatios" is a big facor there as well as the aerodynamical properties of the lifed object.
hmmmmmm...... i could have sworn i did this for a senior project.... :)
From what I have been told, payload swinging in practical applications (helicopter lift) does not primarily come from self-induced movements, but are generated by side winds.
Roy, thanks for the link!
If you got the position of the load and the position/orientation of the copter you can easily calculate these angles. The other possibility would have been a mechanical sensor for the load angles.
@Marc S, Jack
The load position appears to be obtained from the VICON system, although the algorithm was developed using the two swing angles of the load.
http://marhes.ece.unm.edu/images/4/41/2012_ICRA_IPalunko_Swing_free...
- Roy
Acceleration info from the payload seems useful. I used to do construction surveying and we would hang a plumb bob down to the next floor and mark a point, but the plumb bob had to be held steady. When you first let it rip it would swing all over the place so you had to make big opposing arm motions and then it would be finger adjustments...and mark. The bigger the swings, the faster the motions to counter it, so acceleration of that payload was part of the visual/sensory calculations we would make.
Can't see any difference in the 480p video. Any real solution requires acceleration information from the payload.
Would be interesting to know if the load position is also measured with the Vicon...
Anyway, autonomous load transportation using a small helicopter has been demonstrated outdoor years ago:
http://youtu.be/J3gmzn0bSa4