I think I have concluded that CCC mode has a very dumb flaw.

[Eslader]

Some might remember the thread I started a couple weeks ago about my H crashing in CCC mode. In that thread, the consensus was that if you go from 1 waypoint to a 2nd, then climb rapidly to a 3rd, it will dip down sharply between 1 and 2 to curve the line on the way up to 3. That conclusion was borne out by testing.

Well, that's not ideal, but it's at least understandable. Further testing today, though, leaves me thinking the CCC logic should have been better thought out. I decided to see what would happen if I started at one altitude at waypoint 1, then went straight up to 2, then farther up to 3. So unlike the last time, there would not be several waypoints at the low altitude before transitioning to the higher one. The transition would happen immediately from the start.

So the way (cinematically) this should ideally work is that the drone moves into position for 1, and then when you push the left stick forward, head straight toward 2, and then 3.

What actually happened is that it curved the vertical again, even though there is no "curved cable" reason to do so because the starting point of the cable is also the start of the climb. For a smooth "curve," it should start at 1 and just go straight up. Instead, it dropped 8 feet very quickly. I was starting to flick the abort switch when it went back up. But you can see in the video that, having started at 18 feet (over a ditch), it got close enough to the ground to blow some leaves around.

The really dumb part is that the "curve" part wasn't even smooth, eliminating the point of putting the curve in at all.

Anyway, I just thought I'd post this in case anyone else was planning on doing altitude changes in CCC mode. If you do, be very careful, especially if you're going for the crane shot.

 

[Grumpy]

Some might remember the thread I started a couple weeks ago about my H crashing in CCC mode. In that thread, the consensus was that if you go from 1 waypoint to a 2nd, then climb rapidly to a 3rd, it will dip down sharply between 1 and 2 to curve the line on the way up to 3. That conclusion was borne out by testing.

Well, that's not ideal, but it's at least understandable. Further testing today, though, leaves me thinking the CCC logic should have been better thought out. I decided to see what would happen if I started at one altitude at waypoint 1, then went straight up to 2, then farther up to 3. So unlike the last time, there would not be several waypoints at the low altitude before transitioning to the higher one. The transition would happen immediately from the start.

So the way (cinematically) this should ideally work is that the drone moves into position for 1, and then when you push the left stick forward, head straight toward 2, and then 3.

What actually happened is that it curved the vertical again, even though there is no "curved cable" reason to do so because the starting point of the cable is also the start of the climb. For a smooth "curve," it should start at 1 and just go straight up. Instead, it dropped 8 feet very quickly. I was starting to flick the abort switch when it went back up. But you can see in the video that, having started at 18 feet (over a ditch), it got close enough to the ground to blow some leaves around.

The really dumb part is that the "curve" part wasn't even smooth, eliminating the point of putting the curve in at all.

Anyway, I just thought I'd post this in case anyone else was planning on doing altitude changes in CCC mode. If you do, be very careful, especially if you're going for the crane shot.

Maybe I'm missing something but, it looks smooth to me. I didn't see where the aircraft dropped any. Has anyone else had this anomaly?

 

[Phaedrus]

It looked to me like it dropped a bit right at the start and then climbed out smoothly.

 

[AeroJ]

Quite separate to your point, that's a very attractive lake, and should come out beautifully in the grade...

 

[Eslader]

Yeah, and best of all, it's nowhere near any airport. Even the tax-deduction ones!

It looked to me like it dropped a bit right at the start and then climbed out smoothly.

Exactly. I started the recording after it got to wpt 1, then waited a second for preroll and began the run. That dip at the beginning is when it's going from 1 to 2.

 

[AeroJ]

That dip at the beginning is when it's going from 1 to 2.

As a 3D artist it is part of my job to have a very good understanding of how bezier curves work, and I suspect the CCC function of the H behaves very much like a (cubically interpolated) bezier spline would, given that input of start, straight up, and straight up again. In 3D land, to avoid 'curve-back' problems we would simply make the first 2 points of that hard or zero interpolation, but as that is not an option with CCC, we get bezier-like behaviour even if we input non-bezier style moves like a staged linear rise. And that movement would manifest as a drop before the climbs if that theory is correct...

Is that a flaw in the concept or the programming ? Difficult to say - presumably it would add a new level of complexity if you were able to choose between eased (curved) and linear movement, so I guess it's understandable why they just went for the 'curve everything' approach...

 

[Ty Pilot]

I think Aero J has a good point but there are some other considerations that I think also impact the flight of an H during a CCC Mission.

First, it has been known for quite a while that running a CCC over and over or even days apart that the craft can lose altitude at each waypoint. There was a thread about about a year and a half ago if my memory is correct.

Secondly, the accuracy of the way points being recorded and then repeatedly being in the same spot is questionable. The accuracy of our GPS's is good, but not that good.

And lastly, the flying accuracy of the H to hit that point precisely - while moving? Unlike something like a 3 axis router that can repeatedly run a precise path, the H is moving through a 3D space, are we to believe it is accurately hitting all the marks OR, did they design the software in a way that allows an approximation of the mark? I say it is the latter.

In all of my testing of CCC on my Typhoons I have seen plenty of variation but it is especially true when the craft encounters a sharp direction change whether in the vertical or horizontal plane. Does it deviate heavily before the mark or after or does it calculate the perfect spot to deviate before the turn so as to even out the bezier curve? My guess is - it 'wings' it.

In that video above, the dip at the beginning I don't see as a big deal though I would not have set a point that low. In the accent, I could see very little deviation but I was not there watching the aircraft climb.

While I am not disagreeing with @Eslader totally, I would just say CCC does not necessarily have a flaw but; it does have limits in the overall capability of CCC to accurately execute a perfect path. We as operators just need to understand these limitations and plan to use it accordingly.

 

[NorWiscPilot]

I would submit the title of this post should not read “flaw,” but rather edit to read more like a “notice to airmen” or an “equipment operational alert.”

“CCC Operational Alert: When initiating Curved Cable Cam missions, pilots should be aware of a possible drop in altitude between horizontal waypoints at the transition point to higher altitude waypoints. Planning must allow sufficient altitude for the pre-transition waypoints in anticipation of this potential altitude drop as the aircraft prepares to ‘curve’ toward the next waypoint.

This drop is similar to the curve induced into horizontally related waypoints.”

——

Just saying the above restatement might tend to be more readily observed rather than the original heading that could be construed as “complaining” and thus ignored by many who are not interested in reading rants, perceived or otherwise.

Respectfully,

Jeff

 

[Eslader]

I respectfully stand by what I said. Any time unexpected behavior results from a system performing as designed, I don't think it's inappropriate to view it as a flaw.

In that other thread, we went around for several days before the idea that it was curving the vertical surfaced. Almost everyone on here has more experience with the H, and with sUAS in general, than me. If this behavior were expected, logical, and predictable, then the answer should have been considerably more immediate. That it wasn't isn't due to lack of expertise here, but due to lack of consideration in the way the system is set up.

I don't think anything I wrote can be construed as a "rant." I still like the thing, a lot. But it is not immediately obvious that the aircraft would dive 8 feet before climbing when a straight climb is programmed, and things that aren't immediately obvious cause crashes.

I do agree that expecting the system to perfectly hit the waypoints time after time like a CNC machine would be silly. But expecting the system not to suddenly dart 8 feet in the completely opposite direction of the next waypoint is not quite as absurd. That doesn't mean the whole thing sucks - it just means that operators need to be aware that it will do this dumb thing and plan accordingly.

 

[Phaedrus]

This reminds of the guy not long ago that thought that the fact that if you hit auto land it lands right where it is at was a flaw. He was quite intent at making that point on Facebook. He saw his lack of understanding of how the programming worked as a serious flaw in that programming.

If there is a flaw here it is that Yuneec did not make the situation clear on how the points get connected during CCC.

 

[NorWiscPilot]

I respectfully stand by what I said. Any time unexpected behavior results from a system performing as designed, I don't think it's inappropriate to view it as a flaw.

In that other thread, we went around for several days before the idea that it was curving the vertical surfaced. Almost everyone on here has more experience with the H, and with sUAS in general, than me. If this behavior were expected, logical, and predictable, then the answer should have been considerably more immediate. That it wasn't isn't due to lack of expertise here, but due to lack of consideration in the way the system is set up.

I don't think anything I wrote can be construed as a "rant." I still like the thing, a lot. But it is not immediately obvious that the aircraft would dive 8 feet before climbing when a straight climb is programmed, and things that aren't immediately obvious cause crashes.

I do agree that expecting the system to perfectly hit the waypoints time after time like a CNC machine would be silly. But expecting the system not to suddenly dart 8 feet in the completely opposite direction of the next waypoint is not quite as absurd. That doesn't mean the whole thing sucks - it just means that operators need to be aware that it will do this dumb thing and plan accordingly.

Sorry, @Eslader,

My point was not intended to describe your post as a rant.

My post was trying to point out a possible alternative title, to perhaps increase the size of your audience as to what your warning is about.

Apologies to anyone perceiving my “alternative” as a slam against the original poster and the impetus of that post. This was not the intent.

Jeff

 

[Eslader]

If there is a flaw here it is that Yuneec did not make the situation clear on how the points get connected during CCC.

I don't think we're necessarily disagreeing here. If Yuneec is going to program a system that will make the aircraft dart 8 feet in the opposite direction of the next waypoint before proceeding to it, then that feature should be very clearly documented.

Some forethought should also go into what a user will expect the system to do when designing what the system does. I have my doubts that anyone would expect it to dive toward the ground when you have programmed a climb.

But let me ask a clarifying question - I haven't tested this yet but as soon as weather permits I will. Program 2 waypoints, both at the same altitude, but with waypoint 2 150 feet south of waypoint 1.

Would the aircraft go straight from 1 to 2, follow a curved path between 1 and 2, or would it shoot 8 feet north and then follow a straight line to waypoint 2?

If the answer is, as I suspect, that it would follow a curved line between the two waypoints, or that it would follow a straight line from 1 to 2, then this diving at the ground thing is *really* unexpected behavior in that the curve path behaves differently in the vertical than it does in the horizontal.

If that's the case, this may indeed be a flaw in the curve logic itself as well as a flaw in the documentation.

Sorry, @Eslader,

My point was not intended to describe your post as a rant.

I didn't take any offense. No worries.

 

[NorWiscPilot]

But let me ask a clarifying question - I haven't tested this yet but as soon as weather permits I will. Program 2 waypoints, both at the same altitude, but with waypoint 2 150 feet south of waypoint 1.

Would the aircraft go straight from 1 to 2, follow a curved path between 1 and 2, or would it shoot 8 feet north and then follow a straight line to waypoint 2?

.

In my experience as well as understanding Curved Cable Cam, the aircraft will take a straight path from one point to another when a mission consists of only two waypoints. The absence of a third waypoint prevents the flight controller from knowing in what direction it should “curve” to set up for, and to proceed to that next waypoint. No reason to curve, in other words.

As for vertical, I suspect the amount of drop in a mission flight path is dependent on both the altitude climb necessary to reach the next waypoint as well as the horizontal distance between waypoints. The greater the horizontal distance, the less severe the curve.

Perhaps one reason this issue has not surfaced before (or gained much attention) can be attributed to no one programming a mission with such tight parameters, let alone at as low starting altitude as being discussed in this thread.

All good discussion, nonetheless. Getting the word out is not a bad thing.

Jeff

P.S. I should also mention I have never witnessed this “drop” phenomenon in any of my pre-planned autonomous missions. I will definitely have to pay closer attention, as well as do some testing of my own when time permits. My curiosity had been piqued.

 

[RPR]

s for vertical, I suspect the amount of drop in a mission flight path is dependent on both the altitude climb necessary to reach the next waypoint as well as the horizontal distance between waypoints.

I am with Jeff on this one..

Hi @Eslader are you talking about the drop of altitude in 0:01 to 0:03 on your video? The discrepancy depends on who wrote that source code, and how good he or she is. When we set the first POI the system inputs this data as absolute, Lat and Alt can be performed repeatedly, with accuracy. Can you replicate the CCC and maybe give it a higher altitude, let drone settle for a few seconds before creating the next POI? I am running CCC points for progressive reporting, I cannot say that I am having problems with my points todate, but I've experienced something like this in the past, with another drone using a 3rd party app, where the drone abruptly changed altitude. But you have raised a great topic, and whoever is writing the source code, should look into... (OT: The accuracy of RTH on Yuneec drones is retarded)

Creating a waypoint or CCC mission is not perfect right off the bat, as we would expect, like our checklist, we go over its accuracy, and make a determination before finalizing our autonomous mission. I would retest your CCC again, and refine it.

 

[Eslader]

Hi @Eslader are you talking about the drop of altitude in 0:01 to 0:03 on your video?

Yes.

Can you replicate the CCC and maybe give it a higher altitude, let drone settle for a few seconds before creating the next POI?

Yup. I'll be doing that and more testing as soon as weather and time permit, and will report back.

 

[RPR]

Yes.



Yup. I'll be doing that and more testing as soon as weather and time permit, and will report back.

Since that you brought this up, I will also be doing my test, and thank you for sharing your field report.

 

[ChadCloses]

I wish you could activate obstacle avoidance when using CCC. I think that safety measure could save our drones from unexpected results from CCC actions that we can’t predict.

 

[Rubik]

My 2-bits on this post. How the CCC path is rendered by the TH between points 1 and 2 depends largly on points 3 and 4. If you were to draw your waypoints on paper in the vertical plane, you would see how come points 3 and 4 play such a large role in deciding how to make a smooth curve between points 1 and 2. I think it's pretty cool that the TH can figure that all out in 3D before executing the flight.

 

[Crazy Crow]

Some might remember the thread I started a couple weeks ago about my H crashing in CCC mode. In that thread, the consensus was that if you go from 1 waypoint to a 2nd, then climb rapidly to a 3rd, it will dip down sharply between 1 and 2 to curve the line on the way up to 3. That conclusion was borne out by testing.

Well, that's not ideal, but it's at least understandable. Further testing today, though, leaves me thinking the CCC logic should have been better thought out. I decided to see what would happen if I started at one altitude at waypoint 1, then went straight up to 2, then farther up to 3. So unlike the last time, there would not be several waypoints at the low altitude before transitioning to the higher one. The transition would happen immediately from the start.

So the way (cinematically) this should ideally work is that the drone moves into position for 1, and then when you push the left stick forward, head straight toward 2, and then 3.

What actually happened is that it curved the vertical again, even though there is no "curved cable" reason to do so because the starting point of the cable is also the start of the climb. For a smooth "curve," it should start at 1 and just go straight up. Instead, it dropped 8 feet very quickly. I was starting to flick the abort switch when it went back up. But you can see in the video that, having started at 18 feet (over a ditch), it got close enough to the ground to blow some leaves around.

The really dumb part is that the "curve" part wasn't even smooth, eliminating the point of putting the curve in at all.

Anyway, I just thought I'd post this in case anyone else was planning on doing altitude changes in CCC mode. If you do, be very careful, especially if you're going for the crane shot.

Indeed CCC has quite a large margin when it comes to altitude changes. I had to do a major damage repair after my CCC crash.

The crash itself happens at 1:00 min

 

[Ty Pilot]

This^ If one considers the w

My 2-bits on this post. How the CCC path is rendered by the TH between points 1 and 2 depends largly on points 3 and 4. If you were to draw your waypoints on paper in the vertical plane, you would see how come points 3 and 4 play such a large role in deciding how to make a smooth curve between points 1 and 2. I think it's pretty cool that the TH can figure that all out in 3D before executing the flight.

Yes; and the other component that has a large impact on the accuracy of CCC is the speed at which you are running the route. The faster it runs, the less accurate it is especially when direction and/or altitude changes are greater.