Failure is an option


It's raining robots-

That will be the headline all of us will cringe at, when the Media goes manic over the first commercial drone crashes of any significance. The same way they do anytime something new takes a stumble (or two).

The thought of my craft hitting somebody really haunts my mind. I'm sorry to be a downer, this topic is definitely a buzz-kill from the heady intoxication of being in the midst of the next great economic boom, but I can't let it go.

So the topic is failure. Despite what your high school sports coach told you, failure is not just an option, it's a guarantee. Yes- our thrust makers, power systems, flight controls, and firmware will get better quickly. However, even if we could magically achieve 100% reliability with all of our flight critical components, failure will still be there. The most random shit is always going to happen.


 The project-

I'm focusing on multicopters in particular with this project, but I feel that a similar strategy would be effective with fixed wing craft as well, in fact it's used all the time.

The specific type of failure scenario I want to tackle here is a total loss of thrust and/or control. Deployable parachutes are being developed, and seem to be working well. I would like to pursue a more passive system, an aerodynamic design built into the airframe itself. Inspired by watching Marcy fly around and wingsuit pilots, I hope to achieve a passive "auto-rotation" or glide of the entire airframe. If the the craft looses thrust, the actual act of falling makes the conversion into an alternate flight mode. A slow spiraling glide to the ground, with a gear down orientation maintained. Energy absorbing landing gear, maybe even a supplemental air bag would compliment the design. The key in my opinion is making the recovery mode completely non-dependent on any system other than gravity and wind resistance. Time is so critical in a recovery scheme for these types of failures at the altitude multis usually work. No sensors, servos, or firmware required, just a thoughtful arm and frame design. The challenge will be to prevent undesired flight characteristics while in normal powered flight, in fast controlled descents we don't want to be induced into a flat spin! Can the free fall decent rate be slowed enough to usefully lower the risk of damage on the ground and airframe itself?

Step one- Get to the sewing machine and build a Bat suit for one of my quads


To be continued.


E-mail me when people leave their comments –

You need to be a member of diydrones to add comments!

Join diydrones


  • Gary,

    I used auto-rotation with quotes because I don't mean typical prop disc auto like a mono-copter, but rotation of the entire craft working toward large gliding spirals as vertical speed builds. I like your idea of the flexible poles for deploying a traditional parachute!

  • Multicopters in particular suffer from the WileCoyote Syndrome.

    Basically if anything goes wrong they simply fall straight out of the sky.

    They have no real flight speed and for the most part autorotation isn't even an option.

    The only real option for current multicopters is some sort of deployable parachute like device.

    You could build a multicopter with substantial built in airfoils that could permit some sort of rapid autonomous but perhaps non lethal descent but that would add weight and the aerodynamic surfaces would no doubt interfere with normal flight and respond unfavorably to winds and gusts in normal operation.

    Aside form deploying them at the right time and only at the right time, existing parachute systems suffer from a reasonably high likelihood of getting entangled in the multicopters superstructure and thus becoming part of the problem rather than the solution.

    Even spring, air or rocket ejected parachutes don't work to well if the copter happens to be upside down when it is deployed.

    I came up with a design that uses four thin flexible fiberglass poles (fly rod stock is good) extending upward and outward from the copter body.

    The poles would be bent back and down to the center of the copter and a square Chute could be folded and kept in a servo released "pack" when released the poles would snap upward and the chute (the edges of which would be directly attached to the end of the poles) would unfurl. It couldn't get caught in the props or other copter mechanism and should even cause the copter to rapidly right itself if necessary.

    Probably you would size the chute adequately to slow descent to at least a non-lethal rate, maybe not enough to save the copter completely from damage. This would also work much closer to the ground than most "deployable" chutes.

    Just a thought, multicopters are dangerous and eventually we are going to be forced to address this.

  • I should mention that I'm a firm believer in motors mounted under the arms, so with a truss style arm made of flat vertically oriented members it will be easy to attach a fabric gore along the top of each arm, normally tucked in, but a strong relative upward airflow would balloon them out, possibly this action could then deploy other more tightly packed surfaces elsewhere. Ultimately I guess it's still a parachute when it comes down to it. Plus some lifting body shape to the airframe center maybe. There's just so many possibilities here I think. Thanks for the feedback guys, like I said "To be continued", lot's to figure out still. I now have a use for my old APM1s and pile of cheap motors! One quadcopter bat suit free fall video is worth a thousand words, as they say ; )

  • Ah, that makes a lot more sense, more of a glider than a maple samara (helicopter leaf).  Thanks for the clarification.  Makes a lot more sense now.  As you mentioned, it seems like deployment is the major difficulty so that you don't unduly affect the flight dynamics of the multicopter during controlled flight but still have a high enough glide ratio to protect it during a fall.

  • Mmm... you're talking about multicopters... but without thrust, your copter is a brick with (typically non-zero) velocity and a constant force being applied. If you attach lifting surfaces to this craft that can be relied upon during un-powered decent, then these will naturally interfere with normal flight (unless, as you suggest, they are deployed when the failure is detected). If you go for a deployable option, that is mechanically more complex and prone to failure... and they almost always fail when you want to rely on them (i.e., when you try to deploy them). Even parachutes fail at times (just ask the many skydivers who have had to deploy a reserve shoot).

    I'm not convinced you'll get anything better than a parachute, nor anything (electro)mechanically easier or cheaper to deploy. I'm open to being convinced though by novel designs or fanciful prototypes... where would we be without them!? ;)

  • George, I think I confused things when I mentioned flat spins in the original post. I'm talking about glide with a small amount of rudder deflection to prevent traveling anywhere during the decent. Air speed with no net ground speed. A helical path, maybe with spin as well to deploy alternate lift surfaces if that's necessary. 

  • While I think that such a system in interesting, they key will be balancing the rotation force (stability/controlability) with the drag force (actually slowing the copter down as opposed to just trading fall speed for rotation speed).  Like the above commenter said, I'm not sure I understand how this system would slow the multicopter as opposed to just converting momentum.  For this reason, I think that this type of system would be excellent if combined with a parachute as a hybrid system.  The auto-rotation airfoils/surfaces will cause the copter to rotate and stay upright while in the early stages of falling, enabling a top mounted parachute to deploy.  The parachute should obviously be triggered by a passive mechanical system or a fully discrete (from the autopilot and motor battery) electronic and power system.  If that were the case you could dramatically reduce freefall crash events, assuming that both the primary autopilot is reasonably reliable and that the separate parachute system is independent and fairly reliable as well. (And obviously the airfoil rotation/stability system just works.)

  • I don't want to count on sensors, what I'm talking about is flying squirrel technology, or better yet a sail plane with a little bit of rudder deflection locked in. Maybe a 5 or 10 meter radius spiral with as much glide ratio as possible. Those wing suits are achieving upwards of 2.5:1. Obviously, just attaching glider wings to a multi will reduce it's usability in normal operations. I'm thinking of something like centrifugally deployed lift surfaces or fabric segments that are forced out of the arms by the wind resistance of a rapid fall. No worries about entanglements with props, just enough carefully placed drag to create an angled spiraling decent. A damage free  impact is probably too much to expect, but at least slow things down. Maybe just enough to give people on the ground a shot at getting out of the way. 

  • Just a thought... if you got for an "auto rotation" solution, that all you're doing is turning gravitational potential energy into increased angular kinetic energy. Ultimately, parts of the craft will be travelling at a tangential velocity at or close to the same linear velocity they would have if in free-fall. Sure, you could potentially gain some effeciency in energy bleed-off by careful design of rotational form drag, but my intuition is that it wouldn't be nearly enough to save the craft upon final impact. I would think the safest and easiest option (and potentially cheapest, when accounting for craft damage) would be a parachute deployed by a sustained freefall reading on one or more accelerometers. Ultimately this is going to curb your acrobatics display, but if we're talking about saving commerical drones operating over civilian areas, we're probably not talking about outside loops or self-induced flat spins. ;)

This reply was deleted.