Is there a mathematical solution for preventing fly-aways?
My understanding is that the following is what causes them:
- Accel z goes negative due to ship vibration versus an actual drop in altitude
- The ship corrects by adding power
- The added power causes a larger vibration-induced negative z
- The ship corrects the pseudo altitude drop by adding more power
- A fly away occurs
My understanding of mathematical solutions are:
- A filtered or weighted z only diminishes (does not solve) the effect.
Has anyone tried a "significant z", z / s, where s is the moving average of the variation of z?
- when the IMU isn't vibrating, s is low so the magnitude of z / s is high
... z is significant
... z can be trusted for use in altitude control
- when the IMU is vibrating, s is high so the magnitude of z / s is low
... z / s has less effect
... and will not caused a flyaway
Replies
Fantastic article! Thanks. Now if we can just get a developer to implement. My question was has significant z been tried. Apparently so, by NASA.
"One of the jobs of the preprocessing component is to determine each sensor's noise estimate, and this requires a removal of any real signal trend ... after the Fourier transform based highpass filter has been applied. ... A moving window that spans ... worth of the most recent data for each sensor is used to characterize the noise level every ... The noise levels of all sensors of given type are then averaged together to characterize each population of sensors. Whenever a sensor's noise is greater than some factor (e.g., 3 or 5) times the population noise, the sensor is rejected as defective. A sensor is also rejected whenever a health/status message reports a fault. All sensor data that passes noise and health/status checks is forwarded to the map calculating tomography component.
I don't remember the name of the algorithm but I did a google search on "nasa resolve conflicting multiple sensor data", the following paper http://www.meshoxford.com/2011_01_20_SensorNettingAlgorithmPaper_fi... seems promising.
In this case, the same concept applies, but instead of location, read rate (Hz) would need to vary between the IMUs to handle harmonic issues that can hide input at specific frequencies.
In that way if two IMU saw the accel-z but one didn't, it it likely that the acceleration is in line with the read rate of the one that was blind to it.
that would work, except the IMU doesn't know that it's ascending. it thinks it is falling in a downdraft or other natural (non-pilot control reason). you've told the FU that you want to hold elevation, and it's trying to do that.
but i see where you are going with this. simply have a parameter where the pilot can limit max non-stick control before the flight so if it starts to fly away, it at least won't be at full throttle. a pilot could gauge the weather and choose the limit on the parameter.
There is another way of looking at this problem and that is by comparison to "norms". A norm or benchmark is used to establish acceptable levels of performance of a product. An individual experience can change someone's perception of what is normal, so we see different extremes of experience being stated in this thread, sometimes passionately, as the absolute truth.
However, once you start talking about an entire industry, the individual experience becomes a single point in a very big statistical set. And the bigger the industry, the bigger the data set and the more that previously insignificant few percent of outliers start looking like a major problem.
Once drones become a fully fledged industry, not just the new kid on the block, the norms will change significantly. Surprisingly, the biggest push for change comes not from the FAA, but from the insurance companies. The insurance companies have a vested interest in not paying out claims, so they are the single biggest contributor to the standards that we all take for granted in every other established industry, like air travel or motor cars or oil refineries.
We find ourselves at the point of transition from a hobby into a professional industry. The norms will change, and there is an actuary right now calculating the cost of "drone fly-aways" to his insurance company. New standards will be imposed and drone manufacturers will have to prove that their products meet these standards. Human error will not be accepted as an excuse for something going wrong. They will demand that solutions are found or else… no insurance cover! Many companies in the drone sector are already asking for ISO9001 certification from their suppliers. Such fun ;)
LaserDev
A good point on yet how another financial industry service provider interferes with technological development! :-(
Legalized gamblers called insurance companies need to take a back seat IMHO...the focus should be on implementing standards and systems to avoid risks completely, NOT to make money from them.
Classic example is cars. We easily have the technology to adopt a much better rail based system that avoids the necessity for every automation system included in a "Google car", bar your awesome range measuring LIDAR products, and would make personal travel safe, faster and more efficient, and not cost 100,000's of lives a year on our perverted road system which is still based on the horse drawn cart from thousands of years ago. We can do much better IF we can overcome the faulty stimulus for decisions, caused by a corrupted financial system.
In this case a possible remedy is to use sensory information that drones do not currently have, but that currently gives pilots an advantage. That is vision.
Most drones are currently flying around "blind" in that there frame of reference is purely the result of susceptible sensory inputs. None of which have the capacity to determine attitude, position or orientation by visual cues (remember the old IR LED horizon sensor?). We're constantly flying drones IFR...it's silly. I firmly believe that the next step is to incorporate visual data into the FC method, which will result in a comparable, if not overall far superior FC system. Thanks to mobiles the mobile processing power and camera hardware is there, we just need the software.
Who's up for doing visual flight control?
Regards
+1. This is also my intuition that visual camùera input is the only way out via the top.
As long has you have a FC that accepts inputs and runs proper code (like pure gyro mode - without accelerometer trashing a not needed horizon) and as long as you have a pilot that can fly RC you will have no flyaways at all. If RC contact breaks do Barometer autoland (with drift) or shut down motors so no flyaway in that case as well. It can't get any simpler that this. Acc-Z jumbo-mumbo will take you nowhere. If you blame all flyaways on vibration alone you will first have to have a specific onboard sensor that just picks up vibration (maybe a little piezo-plate?) and is not disturbed by accelerations. Just my point of view.
@Crashpilot1000
Although I agree that the pilot can prevent flyaway, however, as the UAV progresses and automous flying becomes a norm, the UAV must have the self-correcting functions in place in case of component failure.. my 0.02.
That's why the industry is pushing to image/video recognition because that can solve the problem. Currently we are in the gap between marketing hype of what future technology could do (projected to this point in time ofcourse for the sales) and what the average FC can do.
-
1
-
2
-
3
-
4
-
5
of 17 Next