Failsafe

[zenguerilla]

I think what the person is mentioning here is that they would prefer if CF Failsafe added 1 or 2 additional throttle step downs with additional _delay duration values.

I don't know what already happens about ACC engagement, depending on what mode is being flown in.

But, it's still a crapshoot with any option.

 

[ZoomNBoom]

Thats still silly if you have a working barometer (or Z axis acc). Why wouldnt you use that, just because your RF link went down? There is no step down configuration with fixed throttle positions that will work acceptably from both <10 and >200 meter and that will work with any AUW, prop size or 3/4S battery configuration.

 

[zenguerilla]

Thats still silly if you have a working barometer (or Z axis acc). Why wouldnt you use that, just because your RF link went down? There is no step down configuration with fixed throttle positions that will work acceptably from both <10 and >200 meter and that will work with any AUW, prop size or 3/4S battery configuration.

If I was that concerned I would consider $50+ and 50g+ for GPS and Compass with a full Naze and shoot for accurate RTH.

I believe barometer alone will be much like throttle_value /guess landing at a certain point of approach, where it is assuming +/- 10-20ft. where the ground is.

Beyond that I think some type of sonar probably is, and likely will be, the thing.

There is plenty of stuff out there, but I also came across this the other day from the Hydrogen quad link which looks like it uses some kind of sonar.

In a couple or few years, like most things, sonar-type stuff could probably be smaller, lighter, more durable, accurate, etc.

Short of adequately accurate RTH or similarly potentially highly complicated Sonar guidance, I think it's still a crapshoot.

 

[ZoomNBoom]

I believe barometer alone will be much like throttle_value /guess landing at a certain point of approach, where it is assuming +/- 10-20ft. where the ground is.

Not at all. It doesnt need to know how high it is, or the ground is, it only needs to know its vertical speed, even if only roughly. And it does know that, in fact it knows it pretty darn accurately with a baro (check your telemetry logs if you have them), its accurate to ~0.1 m/s. Thats ~10x more accurate than you'd need for a failsafe.

And even without baro it can make a very decent guess using the accelerometer. And it does so already, just try it: hover it before you without using baro/alt hold, and push it down or up. It will resist it, because it senses the acceleration. Ie, gravity. It can sense if gravity is > or < than 1G and also fairly accurately and while that doesnt give you velocity, it should be good enough to make an emergency crash landing and far better than a fixed throttle position. At least you know it will come down, and not go up because you decided to fly light or use other battery/props.

Beyond that I think some type of sonar probably is, and likely will be, the thing.

Sonar is great. At altitudes below ~3-5 meter. Then its really accurate, but thats totally not needed for a safety feature.
BTW, you can buy sonar's for like $20 and hook them up to your flight controller. At least on multiwii, not sure if CF supports its already. Again though, its only really useful for a nice, soft, autoland or maintaining very low altitude.

 

[zenguerilla]

Sounds like you have it all figured out.

I'm aware of some of the current sonar stuff. It's just probably not as applicable as it may be in a couple of years +/-.

 

[FinalGlideAus]

A lot of the time simplicity is the key. Understanding your radio limitations and not over stepping their bounds as well as setting up the gear for reliable operation is a point that many over look but instead rely on solutions that while look great on paper don't tend to work that well in multiple environments in reality.

I can only suggest you take your ideas and put them to practice. If it works then the rest of us will embrace it with open arms. I'm always happy to be proven wrong but for the type of quads I fly at least there is little benefit yet many pitfalls in doing such things in my mind.

 

[ZoomNBoom]

I wish I had the C++ skills to implement it on the naze, but unfortunately, I dont. But that the concept works, and isnt new is proven by plenty of other flight controllers, including NHA's forks of multiwii and the APM/Pixhawk which will autoland in situ just fine if you configure them for it, even though most will configure it for RTH.

I dont know if they included all the sanity checks mentioned earlier for their RTH algorithm, but at least I see it reverts to autoland in the event of GPS loss and Ive yet to hear of an APM flyaway, so Im gonna assume they did at least better than DJI.

 

[nilsen]

I have used the RTH on my APM many times, not because of an issue but because I have the shots and then I flip the switch for it to come home, only because it's fun and I love to see it coming home to daddy...

That being said, in my hex the position it lands is always accurate to about 2 meters from where I took off and the descent is very smooth and slow. Even the auto-land works quite well, I know they did lots of work on it in the latest release to detect a landing, the motors stop pretty quickly.

I mostly land manually once the copter is close to me, and I never use RTH when I'm flying around people or buildings, I just wouldn't trust it, but when I'm in the open I love to use the mission planner, make some autonomous routes, set my camera angles and direction of camera, slap on my goggles and hit start and get taken on a brilliant aerial journey.

As for the Cleanflight failsafe, I think it could be smarter but in this situation I am just glad it was there, a flyaway would be heartbreaking!

 

[ZoomNBoom]

Sounds like you have it all figured out.

I'm aware of some of the current sonar stuff. It's just probably not as applicable as it may be in a couple of years +/-.

Sonar is inherently very limited in range. There is reason its not used for anything more than a few meter at most (in air), even if power is hardly a concern - think cars, sonar is okay for parking help, but if we need to look further, for instance for adaptive cruise control, we need optical systems, radar or laser. Sonar just cant.

Of course, its quite possible, even likely, one day on our quads we will have "speed of light" 3D camera's like on a Kinect, or a laser scanner like you find on autonomous cars from Google etc, and it will map its environment in 3D and never crash in to anything.

But until then, a baro is perfectly useful to control a descend in case of RF link breakage

 

[nilsen]

Sense and avoid... of buildings and other objects. Sign me up.

 

[jipp]

id think id want the sonar to navigate over trees etc.. seems like that is where that sensor would be useful.. sure you say its only good for a certain length but it seems to me that is more than enough distance so it stays above the trees on its way home or whatever, yeah.

which would come in handy for AP.

chris.

 

[jipp]

I have used the RTH on my APM many times, not because of an issue but because I have the shots and then I flip the switch for it to come home, only because it's fun and I love to see it coming home to daddy...

That being said, in my hex the position it lands is always accurate to about 2 meters from where I took off and the descent is very smooth and slow. Even the auto-land works quite well, I know they did lots of work on it in the latest release to detect a landing, the motors stop pretty quickly.

I mostly land manually once the copter is close to me, and I never use RTH when I'm flying around people or buildings, I just wouldn't trust it, but when I'm in the open I love to use the mission planner, make some autonomous routes, set my camera angles and direction of camera, slap on my goggles and hit start and get taken on a brilliant aerial journey.

As for the Cleanflight failsafe, I think it could be smarter but in this situation I am just glad it was there, a flyaway would be heartbreaking!

i have to say that is what im interested in too. i think it be fun to go for a ride.. but also have the ability to over ride the controls just in case you need to take over.

i look forward to playing with AP and robot stuff down the road i think it will be quite fun.
chris.

 

[ZoomNBoom]

id think id want the sonar to navigate over trees etc.. seems like that is where that sensor would be useful.. sure you say its only good for a certain length but it seems to me that is more than enough distance so it stays above the trees on its way home or whatever, yeah.

Wouldnt count on it. ~5m is the maximum practical range limit when the target is solid ground. It wont get nearly that far detecting branches (if it detects them at all, I wouldnt count on it). Avoiding rooftops or rocks, ie, terrain following, that should be doable if you tilt the sensor forward enough and you fly slow enough. I imagine you'd need at least 2 seconds to detect and avoid, if your detection range is 5 meter, that puts you at ~2.5 m/s. So you're not gonna be breaking speed records.

But by all means, give it a try. I think APM supports it pretty well. Keep that ladder handy though

 

[zenguerilla]

Sonar is inherently very limited in range. There is reason its not used for anything more than a few meter at most (in air), even if power is hardly a concern - think cars, sonar is okay for parking help, but if we need to look further, for instance for adaptive cruise control, we need optical systems, radar or laser. Sonar just cant.

Of course, its quite possible, even likely, one day on our quads we will have "speed of light" 3D camera's like on a Kinect, or a laser scanner like you find on autonomous cars from Google etc, and it will map its environment in 3D and never crash in to anything.

But until then, a baro is perfectly useful to control a descend in case of RF link breakage

The term sonar was intended very generally and loosely, meaning what ever it might be for the application of such a thing, optical, radar, laser, etc.

Regarding a controlled landing for a failsafe measure, I think the additional concern would be that a controlled landing in a river full of water, a road full of traffic, or an area full of people is still not a very good thing.

Not for the sake of argument, but this would seem to make an accurate and effective RTH function of much greater benefit in comparison.

 

[ZoomNBoom]

The term sonar was intended very generally and loosely, meaning what ever it might be for the application of such a thing, optical, radar, laser, etc.

Then call it "ranging" or situational awareness or something. Sonar is a specific term, it stands for sound navigation and ranging.

Regarding a controlled landing for a failsafe measure, I think the additional concern would be that a controlled landing in a river full of water, a road full of traffic, or an area full of people is still not a very good thing.

We're talking about a loss of RF link. You shouldnt be flying above crowds or intersections with a working RF link. As for water, well, soft landing may not matter there unless you mounted floats, but if you're flying over water, you're kinda taking that risk anyway

Not for the sake of argument, but this would seem to make an accurate and effective RTH function of much greater benefit in comparison.

No question RTH is better, if it works. But it also requires additional hardware (gps, good compass,..) not found on a bare naze. It requires a solid GPS fix which isnt guaranteed, so you still need a backup plan for that. And its also vastly more complicated and prone to bugs and errors. I wouldnt be surprised if more crafts where lost due to RTH failure than crafts would have been lost due to unfortunate semi-soft baro landings instead. So this is where I agree with the other posters in this thread saying simpler is better. Baro descend is simple, RTH is not (and cutting the engines at altitude on a working multirotor with working baro seems stupid and potentially dangerous).

 

[zenguerilla]

Then call it "ranging" or situational awareness or something. Sonar is a specific term, it stands for sound navigation and ranging.



We're talking about a loss of RF link. You shouldnt be flying above crowds or intersections with a working RF link. As for water, well, soft landing may not matter there unless you mounted floats, but if you're flying over water, you're kinda taking that risk anyway



No question RTH is better, if it works. But it also requires additional hardware (gps, good compass,..) not found on a bare naze. It requires a solid GPS fix which isnt guaranteed, so you still need a backup plan for that. And its also vastly more complicated and prone to bugs and errors. I wouldnt be surprised if more crafts where lost due to RTH failure than crafts would have been lost due to unfortunate semi-soft baro landings instead. So this is where I agree with the other posters in this thread saying simpler is better. Baro descend is simple, RTH is not (and cutting the engines at altitude on a working multirotor with working baro seems stupid and potentially dangerous).

I doubt the difference between reflected light waves, sound waves, etc for the sake of obstacle determination and avoidance was really a critical issue for what was initially being mentioned.

But, I would say quoting sonar as generally having only ~5 meter detection capability was probably more of trying to win an argument rather than find a solution of benefit and improvement.

And, as has been mentioned, a controlled landing as a failsafe measure is probably not going to offer control over where it is going to land, or manage to be a straight vertical descent, which would probably easily include any of those examples of water, traffic, or people.

As for situational awareness, it was also already mentioned that an adequate amount of planning and that with a motor cut-off failsafe is probably going to be one of the simpler and more effective methods of responsible flying and signal loss preparation as well as prevention.