Imprecise GPS waypoints are causing me to CRASH!

[Dirty Bird]

I set my waypoints based off Google maps with a Nexus 7" tablet. The waypoints aren't perfect but they are pretty damned close. Expecting accuracy within inches is a level of precision which probably isn't realistic. Also I believe the altitude is obtained from the Phantom's barometer unless it is relatively close to the ground where VPS comes into play. Not sure how accurate either is but the barometer could be affected simply by relative barometric pressure (though this is unlikely to change drastically during a 15-20 min flight).

The attached building spiral was plotted with the Nexus. I had zoomed way in on the map for precision & didn't realize until she returned & I watched the video just how close it was. This was 1.5 minutes out of a 16 minute flight. I was so close to the building it failed to provide the effect I was seeking, but there was a chance of catching a secretary in a compromising position! I am still tweaking this spiral until I get it to my liking.

Watch this @ 1/2 or 1/4 speed to get the full effect!

Guys,

I realize that GPS waypoints are not +/- six inches, but boy is their imprecision causing me problems. I fly my drone carefully in Litchi Waypoint mode, dropping waypoints along my route. Then, when I run the mission, there are all kinds of problems:

1. Wrong track (west): It'll be too close to the building (see 19s into this video: https://www.skypixel.com/videos/5c73e9df-28fe-48b0-ba32-fb89a4eee715, if you watch closely, you'll see the video frame actually moves, because the feet of the P4 scraped across the top edge of the building -- can't get any closer to a wreck than that!).
2. Wrong height: the waypoints I set VERY carefully so that the P4 was EXACTLY at roof level, not any higher, because when it's higher, you can see all the ugly gravel on the roof, and I didn't want that to show, and not any lower, because that would cut off the pyramid. Then, when I run it, it's two feet higher than the roof, and all that ugly gravel is visible.
   
3. Wrong track (east): twice when I ran this mission, I crashed into the tree near the building (see near miss at 4s here: SkyPixel - Connecting Creativity). I was not within five feet of it when I dropped the waypoint. UGH!

So, I guess my question is: how accurate are those waypoints? Any way to make them more accurate?

Very frustrating for those of us trying to be precise with our flight paths.

BHM

 

[SpaceDan]

Litchi waypoints are not accurate enough. They seem to be plus or minus 20 feet. Useless at closer distance. Might work for filming a waterfall at 100' away but if the subject closer up (under 40' away waypoints and heading and poi are not going to cut it.

are you talking about placing a waypoint using google satellite imagery?

 

[Dirty Bird]

I agree. I've flown my current project over a dozen times & it's been pretty doggone consistent flight-after-flight.

Strange some of the stuff in this thread, my Litchi waypoint missions seem to be reapeatedly accurate to a foot or 2 on all axis, on all dimensions.

 

[Ibmftw]

how accurate is it using google satellite images to place your waypoints? thx

I couldnt tell you exact ranges only that I have been close to the phantom as its passed me by on a mission I have repeated many times, and reviewed the video, I am talking really accurate. In light of this thread I guess it comes down to individual phantoms as apposed to Litchi.

 

[ussvertigo]

I have plotted missions using conservative altitude and wp placement, then based on the video adjust both to get in closer if need be, no an exact science. Its the altitude I worry about the most, barometer seems pretty accurate, but flying in Lichi I rarely have the drone pointed in the direction its heading, always pointed at a POI, and obstacle avoidance won't help there.

Dirty Bird: I'm sure your aware of this, an alternative would be to use DJI Go and place your POI right on top of the building, and back up to desired distance, during its circling you can change distance and altitude. A little backed off would have looked nice. Where is that?

 

[Wibble]

I set my waypoints based off Google maps with a Nexus 7" tablet. The waypoints aren't perfect but they are pretty damned close. Expecting accuracy within inches is a level of precision which probably isn't realistic. Also I believe the altitude is obtained from the Phantom's barometer unless it is relatively close to the ground where VPS comes into play. Not sure how accurate either is but the barometer could be affected simply by relative barometric pressure (though this is unlikely to change drastically during a 15-20 min flight).

The attached building spiral was plotted with the Nexus. I had zoomed way in on the map for precision & didn't realize until she returned & I watched the video just how close it was. This was 1.5 minutes out of a 16 minute flight. I was so close to the building it failed to provide the effect I was seeking, but there was a chance of catching a secretary in a compromising position! I am still tweaking this spiral until I get it to my liking.

Watch this @ 1/2 or 1/4 speed to get the full effect!

Awesome video!!!
Where was that - Baltimore?
I think maybe you set the spiral without realising the diameter of the building!

 

[Toyz]

are you talking about placing a waypoint using google satellite imagery?

The waypoints I am referring to were created using the birds position and the custom button, not satellite position . The google satellite (and map) info in my area are no less than 15 years old and very low resolution, taken with the sun extremely low and deep into winter.

 

[Emil Cost]

Haha.. I think the issue will come down to the barometer.. There are too many variables that will cause variances. Variances which are far greater than 24 inches.


Sent from my iPad using PhantomPilots mobile app

First of all, when I read the initial post, the thoughts expressed in these comments were like reading my mind. Pertaining to your great comment, is the fact that raw data coming from the barometer is noisy, and the variances even after filtering translate into overall precision and accuracy of the numbers used by the flight controller. I have done some work with these devices and let me just say that you really have to know your stuff in signal processing to get them to work. I am totally amazed at the job the programmers, engineers, and no doubt the scientists have done in producing the Phantom. I completely agree that these flying machines are expensive toys, but they are on the cusp of being serious instruments.

 

[Sim597]

Wow, I guess how we look at things of course depends on our knowledge, or ignorance of certain task, and our environment which we assume things.
I personally would never think my P4 could fly at the tolerances you choose, it just something I know, from being around these types of hobbies and just general knowledge of technical abilities in consumer grade goods, at this time.
I've said before, ignorance isn't a bad word, if defines something we aren't "learned" at, or taught, or just know.
The videos didn't work for me, but to drag the gear! Holy cow! That's a heck of a video I'd like to see it.
Others have already chimed in so no need to beat a horse, but haven't you seen all the post about P4s and the inconsistency in altitude adjustment/control? Obviously you've learned a lot, and even though I already know about flight envelopes and accuracy, you flying these Litchi missions puts you light years ahead of me in that respect as I have yet to USE my autopilot and really want to, I have a couple missions I'd love to do using LSMC (lost signal mission completion) but just to nervous about sketchy issues, even with AutoFlightLogic most excellent algorithms and reliability.
Good job on not totaling your bird man, good to have a little luck,

Misjudging flight tolerances in litchi and ruining your bird=$1400
Scraping gear across the top of a building in autonomous flight mode= Priceless!


Sent from my iPhone using PhantomPilots mobile app

 

[John Decker]

Flyboy was right - the gps system itself is at fault rather than Litchi. If you check (I had an app from the Korean space agency which would show it) you'll see that gps accuracy is usually 1 - 2 metres (3' - 6') out at any one time.

So if you set a waypoint it might be accurate or it might be 6' out. As someone who needs to fly with close tolerances I can sympathise with the OP - I have my own compilation of close calls, some due to poor piloting but a lot due to gps failings.

The DJI app has the advantage of eliminating the inaccuracies by pre-flying the route, but I think we all know the DJI vs Litchi arguments.
(Fwiw, even flying accurately one day is no guarantee of doing so the next day.)

One solution would be to set a course either on the Hub or at the time and then send the bird to the start point. You could at that point see how far out it is and adjust the mission accordingly.

I've been on to Litchi but they say grabbing the whole mission and moving it x feet in any direction is not an option due to the limitations of Android. Maybe they can find a workaround but in the meantime I adjust the waypoints individually.

 

[m0j0]

The phantom is fantastic no doubt. I'm amazed by just how great it is..


Sent from my iPad using PhantomPilots mobile app

 

[GeoffJ]

I'm a geocacher so use GPS technology to find things all the time. What you need to consider is:
- GPS accuracy can be affected by clouds, obstacles such as trees, buildings, solar activity, a whole host of things. You can generally expect accuracy to within a few metres with a handheld GPS or phone, I'd expect a Phantom to be the same. RTH accuracy is always within about 2 m for me.
- The variation from one day to the next can be 4 to 8 or so metres because of these factors.
- We find Google Maps to be a bit inconsistent. In some areas they agree with a GPS to a metre or two, in other areas they can be 15 metres out.

The presence of large buildings and metal objects can have a big affect, so setting waypoints near buildings and expecting high accuracy is fraught with danger.

You have to take a cautious approach where errors of a few metres can have a disastrous impact.

There are sometimes variations between devices too. If you put a couple of smartphones or GPS receivers together you'll see some level of variance in the readings.

 

[pe9999]

In North America, civilisn gps accuracy can be off by 7.8 meters.
When used with augmentation system, the accuracy can improve to cm level

I can sum this up. Is GPS this accurate? No, its not. You can see that here:

Hover Accuracy Vertical:
+/- 0.1 m (when Vision Positioning is active) or +/-0.5 m
Horizontal:
+/- 0.3 m (when Vision Positioning is active) or +/-1.5 m

DJI Phantom 4 – Specs, FAQ, Tutorials and Downloads

 

[Thomas.]

I've been on to Litchi but they say grabbing the whole mission and moving it x feet in any direction is not an option due to the limitations of Android. Maybe they can find a workaround but in the meantime I adjust the waypoints individually.

But you could do this yourself on your PC using Litchi's Mission Hub with the provided move button. An alternative would be to download the mission as a CSV file, modify the parameters with a small script and re-upload the file.

 

[KevMo Photog]

I set my waypoints based off Google maps with a Nexus 7" tablet. The waypoints aren't perfect but they are pretty damned close. Expecting accuracy within inches is a level of precision which probably isn't realistic. Also I believe the altitude is obtained from the Phantom's barometer unless it is relatively close to the ground where VPS comes into play. Not sure how accurate either is but the barometer could be affected simply by relative barometric pressure (though this is unlikely to change drastically during a 15-20 min flight).

The attached building spiral was plotted with the Nexus. I had zoomed way in on the map for precision & didn't realize until she returned & I watched the video just how close it was. This was 1.5 minutes out of a 16 minute flight. I was so close to the building it failed to provide the effect I was seeking, but there was a chance of catching a secretary in a compromising position! I am still tweaking this spiral until I get it to my liking.

Watch this @ 1/2 or 1/4 speed to get the full effect!

**** that was one wild flight man! How did you create all these waypoints that are almost on top of each other as you spiral down? Were you shitting when you were watching it on screen go around that close? Cool place and subject for sure. Where the secretary btw? LOL

 

[dmark1867]

I had a collision that cost me a drone, now I am so careful that my drone always stays intact but I am not getting the dramatic footage I was before. They are amazing machines no doubt, but with that said, there can still be variances enough that at least for me caused a collision that destroyed my drone. I give myself at least 30 feet in all directions now. Might be overkill but I have not lost another drone. . . .

 

[John Decker]

But you could do this yourself on your PC using Litchi's Mission Hub with the provided move button. An alternative would be to download the mission as a CSV file, modify the parameters with a small script and re-upload the file.

It's not possible with a smart phone sadly (the buttons obscure each other). And I wouldn't know how to do it through a csv file.

 

[Fat City]

In some cases the problem is not with GPS accuracy, but with how well the map or satellite photo aligns with the GPS. Before you rely on GPS to set waypoints, visit the location and compare what your map shows with the "ground truth".

 

[Thomas.]

It's not possible with a smart phone sadly (the buttons obscure each other). And I wouldn't know how to do it through a csv file.

The calculation itself would not be a problem for me.

Well, right now I do not know how to download and upload the CSV file under iOS. I even am not sure if this is possible with Android. I will try to find that out and post more here if you are interested.


Sent from my iPad using PhantomPilots mobile app (living near Cologne, Germany)

 

[r62ewa]

Having worked with Google on a project I know that they purchase their imagery from other vendors as well. This imagery has its georeferencing completed by the vendor so their could be variations of the exact same spot in subsequent Google Maps updates if the satellite images were georeferenced differently.

I know for my location I can load one image in Google and then turn on a more recent image and you can see the shift in the imagery. So using mapping sources that are possibly introducing variances and then variations within the Phantom itself and there runs the possibility of crashes. Due diligence is key.