5 ms·

> very steady hand I don't know the first thing about helicopters, but the negative feedback circuits that keep things steady in many applications, from power

by ableal 8y ago

> very steady hand I don't know the first thing about helicopters, but the negative feedback circuits that keep things steady in many applications, from power supply voltages to Segways, could help with that. It's a very simple basic concept that has been around for many decades, even before analog electronics, never mind digital devices: https://en.wikipedia.org/wiki/Negative_feedback https://en.wikipedia.org/wiki/Negative_feedback (I repeat I do not know if they are available in helicopters. It may well be the pilot doing it by hand the hard way.)

rzzzt 8y ago
Search-and-rescue helicopters are the most likely to have that capability: https://en.wikipedia.org/wiki/AgustaWestland_AW139#Design https://en.wikipedia.org/wiki/AgustaWestland_AW139#Design https://en.wikipedia.org/wiki/Eurocopter_HH-65_Dolphin#Design https://en.wikipedia.org/wiki/Eurocopter_HH-65_Dolphin#Desig...
halbritt 8y ago
Computerized stability control systems exist, are quite prevalent, and have a thriving open source community. Check out: https://github.com/betaflight/betaflight/wiki https://github.com/betaflight/betaflight/wiki Granted, it's for hobby scale drones, but the same principles apply.
- dfox 8y ago
  The physics involved in multi-rotor drones (or even counter rotating dual rotor toy-level RC helicopters) is significantly different from single rotor+stabilizer helicopters in that there is at least in theory point where the thing is dynamically stable.
  - bri3d 8y ago
    For what it's worth many toy RC helis exist with collective pitch tail rotor setups and full feedback stabilization - it started with tailrotor yaw hold automation and now they can stabilize in all axes. Check out something like Blade mCPx
    - gjhan 8y ago
      The dynamics, which are not just the rigid body dynamics plus the aerodynamics of the vehicle but also include many structural modes of the rotorhead and blades, are not scale invariant. An RC helicopter is proportionally stiffer in certain degrees of freedom by an order of magnitude (in rotor angular velocity and thrust:weight ratio e.g.), and so can ignore dynamical factors that dominate the control and structural sizing for manned helicopters.
      - halbritt 8y ago
        Do you think a PID control would would fail to be able to accommodate those dynamic factors?
  - namibj 8y ago
    I thought quadrocopters were so unstable that flying one without a PD controller, or better a full PID, would lead to a crash within seconds. A Helicopter, possibly only of the counterrotating coaxial variant, should not be affected by this instability, I think.
    - halbritt 8y ago
      That's correct. The controllers use full PID loops for each axis.
  - halbritt 8y ago
    The physics involved in multi-rotor drones and fixed wing aircraft are quite different. However, there is control software that works on either. https://github.com/iNavFlight/inav/wiki https://github.com/iNavFlight/inav/wiki The underlying differences in how they fly aren't that relevant. You have a six-axis IMU, various controls that influence how the craft behaves in those axes and various PID loops that influence the controls. I don't really understand the downvotes. Obviously, having a type certificated auto-pilot for a commercial helicopter is a world apart from the flight control system for a hobby-scale drone, but the gap between them has nothing to do with the technology and everything to do with the process of type certification. My point is that it's pretty impressive what these software projects are able to achieve with sub $100 MCUs. Some of them now have 32khz control loops with 32khz IMU sampling. Anyone that's ever peeked under the hood of a commercial auto-pilot would be amazed by that.
kragen 8y ago
I disagree with most of your comment, but explaining why will require several paragraphs of background. You can read Maxwell's On Governors from 1868 and see that, although it explores their math in some detail, it doesn't attempt to generalize beyond braking or otherwise slowing down steam-engines: https://www.maths.ed.ac.uk/~v1ranick/papers/maxwell1.pdf https://www.maths.ed.ac.uk/~v1ranick/papers/maxwell1.pdf So, although devices with negative feedback had been built for centuries — and many natural or simple artificial objects can be usefully analyzed in terms of feedback — the general idea of negative feedback as such seems to date from the 1920s, and apparently Black discovered its usefulness in 1927. The idea was so outlandish at the time that his patent was initially rejected as describing a useless invention — why would you want to reduce the gain of an amplifier? It wasn't public until 1933. After Black's invention, which upgraded FDM for long-distance telephony (the "carrier system") from three voice channels per wire up to nine, then dramatically more, and permitting long-distance circuits with a sequence of 34 repeaters, rather than, say, three. Bennett's "A history of control engineering, 1930–1955" discusses this history in detail; find it at your favorite library. At the same time, in 1922, Minorsky formalized PID control, which gets into negative feedback from a different angle, that of automatic ship steering. Black's, Minorsky's, and earlier negative-feedback designs were not generalized into a general theory of negative feedback until the work of Nyquist and others in the 1930s — most prominently, Wiener! Nowadays, explicitly designed negative feedback is far more ubiquitous than your comment suggests. By far the most common analog circuit component, for example, is an op-amp, which is almost invariably applied with negative feedback; also, though, negative feedback is central to nearly any kind of homeostasis, optimization, or stable static mechanical equilibrium. So, for example, we have negative feedback in TCP bandwidth usage, in glucemia, and in your toilet tank. Parkinsonian tremors — the opposite of your "very steady hand" — are now being analyzed as oscillations in a negative feedback system involved in human motor control, though the theory is not yet fully mainstream in neuroscience. As for "digital devices", those date back at least to Leibniz; they are far older than analog electronics, which in turn are a couple of decades older than the concept of negative feedback.
- ableal 8y ago
  I agree with your detailed disagreement ;-) But I get the feeling that a large part of the audience here is not that aware of those possibilities - I was just pointing in that general direction.
colanderman 8y ago
> It's a very simple basic concept I mean, sure, the general concept of feedback is "simple" and applicable to helicopters, in the same way that addition is "simple" and applicable to planetary orbital dynamics. But even basic control theory – what you'd need to actually stabilize a dynamic system as complex as a helicopter – fills a textbook or two, and the field is still evolving. [1] Not to mention that the whole concept of "hovering" is damn difficult for a machine to quantify. You'd need LIDAR and/or visual tracking systems to accurately hold your place in all six dimensions well enough to keep a worker next to a powerline, and only in the past decade have those even become good enough for use in land-based vehicles. [1] https://en.wikipedia.org/wiki/Control_theory https://en.wikipedia.org/wiki/Control_theory
- perl4ever 8y ago
  "But even basic control theory – what you'd need to actually stabilize a dynamic system as complex as a helicopter – fills a textbook or two, and the field is still evolving." I am not that good at math, but I feel like if I was trying to solve that problem, it should be as simple as finding someone smart and saying "use a Kalman filter"... In one sense I can't say it's "simple" or "basic" because I can't do it, but I'm under the impression that it is a manageable solved problem that can be summarized in four words. Is that wrong?
  - colanderman 8y ago
    Yes, in exactly the same sense that "building a bridge is simple because I can say 'use a truss' to a smart person" is wrong. Things having names doesn't automatically make them simple. There are – like I said – entire textbooks on the subject – and why it is complex – if you are genuinely interested. I mean, heck, there's two textbooks alone about one particular system I randomly found on Wikipedia [1]. However to directly address your misconception, Kalman filters are not control systems. They are related to control theory, and are used to provide inputs to control systems. But control theory is its own thing, and there is no "general" solution to correctly control any given system. [1] https://en.wikipedia.org/wiki/Furuta_pendulum https://en.wikipedia.org/wiki/Furuta_pendulum

by ableal 8y ago

rzzzt 8y ago

Search-and-rescue helicopters are the most likely to have that capability: https://en.wikipedia.org/wiki/AgustaWestland_AW139#Design https://en.wikipedia.org/wiki/AgustaWestland_AW139#Design https://en.wikipedia.org/wiki/Eurocopter_HH-65_Dolphin#Design https://en.wikipedia.org/wiki/Eurocopter_HH-65_Dolphin#Desig...

halbritt 8y ago

Computerized stability control systems exist, are quite prevalent, and have a thriving open source community. Check out: https://github.com/betaflight/betaflight/wiki https://github.com/betaflight/betaflight/wiki Granted, it's for hobby scale drones, but the same principles apply.

dfox 8y ago
The physics involved in multi-rotor drones (or even counter rotating dual rotor toy-level RC helicopters) is significantly different from single rotor+stabilizer helicopters in that there is at least in theory point where the thing is dynamically stable.
- bri3d 8y ago
  For what it's worth many toy RC helis exist with collective pitch tail rotor setups and full feedback stabilization - it started with tailrotor yaw hold automation and now they can stabilize in all axes. Check out something like Blade mCPx
  - gjhan 8y ago
    The dynamics, which are not just the rigid body dynamics plus the aerodynamics of the vehicle but also include many structural modes of the rotorhead and blades, are not scale invariant. An RC helicopter is proportionally stiffer in certain degrees of freedom by an order of magnitude (in rotor angular velocity and thrust:weight ratio e.g.), and so can ignore dynamical factors that dominate the control and structural sizing for manned helicopters.
    - halbritt 8y ago
      Do you think a PID control would would fail to be able to accommodate those dynamic factors?
- namibj 8y ago
  I thought quadrocopters were so unstable that flying one without a PD controller, or better a full PID, would lead to a crash within seconds. A Helicopter, possibly only of the counterrotating coaxial variant, should not be affected by this instability, I think.
  - halbritt 8y ago
    That's correct. The controllers use full PID loops for each axis.
- halbritt 8y ago
  The physics involved in multi-rotor drones and fixed wing aircraft are quite different. However, there is control software that works on either. https://github.com/iNavFlight/inav/wiki https://github.com/iNavFlight/inav/wiki The underlying differences in how they fly aren't that relevant. You have a six-axis IMU, various controls that influence how the craft behaves in those axes and various PID loops that influence the controls. I don't really understand the downvotes. Obviously, having a type certificated auto-pilot for a commercial helicopter is a world apart from the flight control system for a hobby-scale drone, but the gap between them has nothing to do with the technology and everything to do with the process of type certification. My point is that it's pretty impressive what these software projects are able to achieve with sub $100 MCUs. Some of them now have 32khz control loops with 32khz IMU sampling. Anyone that's ever peeked under the hood of a commercial auto-pilot would be amazed by that.

kragen 8y ago

I disagree with most of your comment, but explaining why will require several paragraphs of background. You can read Maxwell's On Governors from 1868 and see that, although it explores their math in some detail, it doesn't attempt to generalize beyond braking or otherwise slowing down steam-engines: https://www.maths.ed.ac.uk/~v1ranick/papers/maxwell1.pdf https://www.maths.ed.ac.uk/~v1ranick/papers/maxwell1.pdf So, although devices with negative feedback had been built for centuries — and many natural or simple artificial objects can be usefully analyzed in terms of feedback — the general idea of negative feedback as such seems to date from the 1920s, and apparently Black discovered its usefulness in 1927. The idea was so outlandish at the time that his patent was initially rejected as describing a useless invention — why would you want to reduce the gain of an amplifier? It wasn't public until 1933. After Black's invention, which upgraded FDM for long-distance telephony (the "carrier system") from three voice channels per wire up to nine, then dramatically more, and permitting long-distance circuits with a sequence of 34 repeaters, rather than, say, three. Bennett's "A history of control engineering, 1930–1955" discusses this history in detail; find it at your favorite library. At the same time, in 1922, Minorsky formalized PID control, which gets into negative feedback from a different angle, that of automatic ship steering. Black's, Minorsky's, and earlier negative-feedback designs were not generalized into a general theory of negative feedback until the work of Nyquist and others in the 1930s — most prominently, Wiener! Nowadays, explicitly designed negative feedback is far more ubiquitous than your comment suggests. By far the most common analog circuit component, for example, is an op-amp, which is almost invariably applied with negative feedback; also, though, negative feedback is central to nearly any kind of homeostasis, optimization, or stable static mechanical equilibrium. So, for example, we have negative feedback in TCP bandwidth usage, in glucemia, and in your toilet tank. Parkinsonian tremors — the opposite of your "very steady hand" — are now being analyzed as oscillations in a negative feedback system involved in human motor control, though the theory is not yet fully mainstream in neuroscience. As for "digital devices", those date back at least to Leibniz; they are far older than analog electronics, which in turn are a couple of decades older than the concept of negative feedback.

ableal 8y ago
I agree with your detailed disagreement ;-) But I get the feeling that a large part of the audience here is not that aware of those possibilities - I was just pointing in that general direction.

colanderman 8y ago

> It's a very simple basic concept I mean, sure, the general concept of feedback is "simple" and applicable to helicopters, in the same way that addition is "simple" and applicable to planetary orbital dynamics. But even basic control theory – what you'd need to actually stabilize a dynamic system as complex as a helicopter – fills a textbook or two, and the field is still evolving. [1] Not to mention that the whole concept of "hovering" is damn difficult for a machine to quantify. You'd need LIDAR and/or visual tracking systems to accurately hold your place in all six dimensions well enough to keep a worker next to a powerline, and only in the past decade have those even become good enough for use in land-based vehicles. [1] https://en.wikipedia.org/wiki/Control_theory https://en.wikipedia.org/wiki/Control_theory

perl4ever 8y ago
"But even basic control theory – what you'd need to actually stabilize a dynamic system as complex as a helicopter – fills a textbook or two, and the field is still evolving." I am not that good at math, but I feel like if I was trying to solve that problem, it should be as simple as finding someone smart and saying "use a Kalman filter"... In one sense I can't say it's "simple" or "basic" because I can't do it, but I'm under the impression that it is a manageable solved problem that can be summarized in four words. Is that wrong?
- colanderman 8y ago
  Yes, in exactly the same sense that "building a bridge is simple because I can say 'use a truss' to a smart person" is wrong. Things having names doesn't automatically make them simple. There are – like I said – entire textbooks on the subject – and why it is complex – if you are genuinely interested. I mean, heck, there's two textbooks alone about one particular system I randomly found on Wikipedia [1]. However to directly address your misconception, Kalman filters are not control systems. They are related to control theory, and are used to provide inputs to control systems. But control theory is its own thing, and there is no "general" solution to correctly control any given system. [1] https://en.wikipedia.org/wiki/Furuta_pendulum https://en.wikipedia.org/wiki/Furuta_pendulum