Does RTH work with Compass Error?

[ianwood]

1. The Phantom already reports compass error when it detects bad data, (out of range or rapidly changing values.)

That is the system guessing and it's often wrong. DJI has made it more likely to not error out even when the data is bad but even still it sometimes falses positive. This is when the compass error pops up yet it still flies normally. We've all seen it.

2. You still have GPS to hold position. You correct for wind as you go. The trip may be a curve or even a spiral but your can get there unless the wind is faster than the bird can fly in.

No. You don't have any GPS hold without the compass. The drone gets pushed by the wind. It calculates a heading and distance to correct. How does it fly on that heading when it doesn't know where that heading is relative to its orientation? It can hunt for it. This would look like TBE. Big circles.

3.You never even need to yaw the craft. Fly by pitch and roll only as if the craft was in Homelock mode. If there is yaw drift, it will be very small and you can correct for it as you go.

How does it know which way home is relative to its orientation? If there is no wind or a perfectly stable wind (unlikely at low altitudes), it could figure it out. But more likely, the wind will change just enough for it to need to correct so it will be constantly hunting which would result in circles.

Without the compass, GPS is mostly useless. It's not like using GPS on the ground where there is no drift.

 

[HWCM]

Car navigation systems do have compass.

Yes, mine even has a section in the manual on how to calibrate it. Driving in circles in a parking lot.lol

 

[Ed209]

I posted a video earlier in this thread of my compass error that I fought through last week. I have analyzed the data with datcon and dashware and I can say these things for certain:
GPS was partially useful. The position on the map was accurate.

Heading was wrong. The heading logged and indicated in DJI Go was not the true heading.

I did not have gps lock. I flipped to ATTI to fly within 2 sec of the compass error message. However those first 2 seconds of the error, my P3 began to fly off course. It's clearly visible in the flight track.

If you guys are really interested I could post the datcon telemetry data tomorrow. But here's the video again:

 

[Mark The Droner]

... Bottom line, anyone who thinks it's possible to go home on GPS but no compass, you've got a few big issues:

1. How do you know when the compass data is bad?
2. How do you factor wind into the difference between planned course and actual course?
3. How do you correct for gyro drift to determine your heading?

Here are some hints:

1. There's no absolute way to know. It's a guess.
   
2. You can't. You can correct for wind but as soon as it changes (which it does all the time), your correction is bad.
3. You can't. Your margin of error will grow every second.

If you want to navigate without a compass, you will need a much better IMU and an RTK ground reference signal. ...

It won't start spinning but it will experience yaw drift since it only has the gyro to rely on for angular rate.

...
2. You still have GPS to hold position. ....
....

This is what frustrates me about internet discussions. The comment above completely ignores everything that was already thoughtfully posted above it. It's as if ianwood never posted. So now he has to re-write the same ideas in a different format - all because the responder either didn't bother to read the post or chose to ignore the post.

That's why I usually just walk away from threads like this.

 

[Mimoid]

The DJI tech guy I spoke to some days ago, was clear that none of the smart navigation functions will work reliably if there is a compass error. Compass errors are extremely serious malfunctions and should not be taken lightly. You should immediately switch to Atti mode and enjoy the bumpy ride back home, flying by the seat of your pants (and the FPV image on the screen). Once you are near your home point and you can actually hear and see the drone, you can try switching to P mode again and, if the screen turns green, land normally. Yay!

However, if you initiate RTH in a compass error situation or if the drone loses the RC signal and self-engages the failsafe RTH, it will not try to use any kind of hot/cold algorithm - it will simply assume that it knows its correct position and heading, and will fly accordingly. You may - or you may not - end up where you intended. If this is poor programming or not, I cannot tell.

For me, this explanation will do.

//Tom

 

[BudWalker]

........
However, if you initiate RTH in a compass error situation or if the drone loses the RC signal and self-engages the failsafe RTH, it will not try to use any kind of hot/cold algorithm - it will simply assume that it knows its correct position and heading, and will fly accordingly. You may - or you may not - end up where you intended. If this is poor programming or not, I cannot tell.

For me, this explanation will do.

//Tom

This is exactly what happened in this flight Standard - Insight on my Fly Away? - HealthyDrone & YT Link The irony is that the compass was initially in error close to the ground and then became correct just a few feet off the ground. But, the P3 was still using the incorrect computed yaw when RTH was initiated because the RC signal was lost.

 

[Shammyh]

Fwiw.... The brand new A3 flight controller can fly without its compass, but only when using RTK GPS.

It's worth mentioning that the Phantom can always tell how much it's turning without the compass. It has both an accelerometer and a gyroscope after all. i.e. it knows its movement in all three dimensions (and time too). However, it only knows relative position to the gravity vector (aka, the ground) without the compass. Or put another way, it always knows its pitch, roll, and altitude relative to the ground. Unfortunately... It doesn't know it's absolute, as opposed to relative heading without the compass.

That said, it's totally possible for it to make an attempt to fly home without the compass (but with good GPS), DJI simply hasn't programmed it to do such.

 

[Erised]

Mmmm. If you put the bird in homelock and pull the stick back it dosn't matter witch way the bird points, it's coming back home. Where's the compass in that. Walk in the bush with your gps when you stop moving it goes where ever and as soon as you begin to move it points you in the right direction even if you hold your gps backwards. I don't know if phantom is programmed that way but location is just numbers, go to these numbers, as mentioned before the bird just need's to start to move and simple calculations will indicate if it's getting closer or not. A few more calculations and it can come straight home. I can understand the need of a compass to know witch way your camera is pointing (bird) but for navigating the gps is better. Point a compass north facing in front of you, then start side stepping to your right, the compass stays pointing to the north but your moving east, you cannot navigate very well with just a compass that's why ships use the stars as reference, we use satelites. The bird can't come home with just a compass but sure can with just a gps. Hope my P4 is programmed that way!

 

[ianwood]

Fwiw.... The brand new A3 flight controller can fly without its compass, but only when using RTK GPS.

It's worth mentioning that the Phantom can always tell how much it's turning without the compass. It has both an accelerometer and a gyroscope after all. i.e. it knows its movement in all three dimensions (and time too). However, it only knows relative position to the gravity vector (aka, the ground) without the compass. Or put another way, it always knows its pitch, roll, and altitude relative to the ground. Unfortunately... It doesn't know it's absolute, as opposed to relative heading without the compass.

That said, it's totally possible for it to make an attempt to fly home without the compass (but with good GPS), DJI simply hasn't programmed it to do such.

No. As has been said many times before, the gyro drifts. Every second that passes, the margin of error increases. You would need precision IMUs sampling at very high rates to get more than 50ft in the right direction.

Also, RTK is an entirely different system using much better IMUs that are corrected once per second by a fixed ground reference point.

Mmmm. If you put the bird in homelock and pull the stick back it dosn't matter witch way the bird points, it's coming back home. Where's the compass in that.

The compass is the part that makes that possible. End of story.

Walk in the bush with your gps when you stop moving it goes where ever and as soon as you begin to move it points you in the right direction even if you hold your gps backwards.

Do you drift with the wind when you walk in the bush?

 

[Erised]

Take a bearing say 1/2 mile away and walk with your head down only looking at your compass and see where you end up. For some reason I bet you will end up to the right of your end point. Every body I know goes to the right.(you dont need wind to drift). You need a known start point and a final point to calculate a compass bearing. Gps gives you these points and only then can a bearing be calculated just like using a map. Our birds probably do this. Aircraft flying in the arctic have to depend on a gps because compasses will go wacky the closer they get to the pole, then the needle just points down then as they cross the pole then they point 180° and it's still north.
Check out drones in the arctic on google, they worked hard on getting a strong satellite signal because compasses are useless. They fly with gps. Dji should work on gps only redundancy.

 

[syotr]

This is what frustrates me about internet discussions. The comment above completely ignores everything that was already thoughtfully posted above it. It's as if ianwood never posted. So now he has to re-write the same ideas in a different format - all because the responder either didn't bother to read the post or chose to ignore the post.

That's why I usually just walk away from threads like this.

1. How do you know when the compass data is bad?This is a given in the problem as presented. The Phantom has reported a compass error so we ignore the compass.
2. How do you factor wind into the difference between planned course and actual course? If you really want to measure wind speed and direction, just drift for a second in Atti mode. You have you start and ending positions from the GPS. It is a trivial calculation but you don't even really need to know it. Getting a heading is also easy. You command the aircraft to fly forward for a few seconds at high speed. You know your start and end positions from the GPS now you can calculate your orientation accurately enough to start flying home. You make additional corrections as you go.
3. How do you correct for gyro drift to determine your heading? As stated above, you can recalculate heading with each move.

 

[syotr]

The DJI tech guy I spoke to some days ago, was clear that none of the smart navigation functions will work reliably if there is a compass error. Compass errors are extremely serious malfunctions and should not be taken lightly. You should immediately switch to Atti mode and enjoy the bumpy ride back home, flying by the seat of your pants (and the FPV image on the screen). Once you are near your home point and you can actually hear and see the drone, you can try switching to P mode again and, if the screen turns green, land normally. Yay!

However, if you initiate RTH in a compass error situation or if the drone loses the RC signal and self-engages the failsafe RTH, it will not try to use any kind of hot/cold algorithm - it will simply assume that it knows its correct position and heading, and will fly accordingly. You may - or you may not - end up where you intended. If this is poor programming or not, I cannot tell.

For me, this explanation will do.

//Tom

If you can do this, so can a computer.

 

[syotr]

There is a very simple skill that every pilot should learn for flying multirotors. When your aircraft is at a distance where you cannot tell its orientation and you need to fly it back home, without the use of a screen to show you what is happening, there is an easy way to do it.
First you move the stick to roll right for a few seconds. If the craft moves to the right, it is facing away from you. If it moves left, it is facing towards you.
If it does not appear to move, you move the stick to pitch forward for a few seconds. if the craft moves to the right, it is facing right. If it moves left, it is facing left.
Once you know the approximate orientation, you move the stick in the proper direction to head home.
As it is coming home, you will likely see it drift left or right. This can be due to wind or the fact that you orientation is not precise. It does not matter which. You simply add more roll or pitch as need to stop the drift.
If you have never done this, try it first with GPS then as you get better at it, try in Atti mode.
You will find that it is entirely possible to return to home without GPS, compass, FPV, or a screen showing your location and even with a bit of wind.
Many have become too dependent on GPS, compass and apps to fly their aircraft.
If you have already practiced this skill, then you must realize that a computer can also do it, only faster and more precisely, especially when it knows its position at any instance as well as speed and can determine its heading with much better accuracy that you can.
If you have not tried this, you should.

 

[Shammyh]

No. As has been said many times before, the gyro drifts. Every second that passes, the margin of error increases. You would need precision IMUs sampling at very high rates to get more than 50ft in the right direction.

Also, RTK is an entirely different system using much better IMUs that are corrected once per second by a fixed ground reference point.



The compass is the part that makes that possible. End of story.



Do you drift with the wind when you walk in the bush?

Ah, I see I was a bit unclear. Not saying you follow the gyro home, as I said, it only provides relative not absolute rotation. And sure, it'll drift, but by how much, how quickly, and how much of that can you predict?

Either way, the drift is largely irrelevant... You pick a bearing, any bearing, and attempt to follow it in a straight line. You check gps, and see how you did. Then you adjust and try again. For sure, it would require some tricky algorithms, but I'm certain it's *possible* to get some sort of last ditch attempt to fly home without the compass. Well dji do such? Probably not. ;-)

And as to RTK, well aware of how it works. My point was that with sufficiently accurate location (cm accuracy with RTK) you don't need a compass at all.

 

[ianwood]

This is getting dull. If you want your Phantom to swirl around all over the place in the off chance it can eventually get home, go for it. The success rate will be low. There's a reason why every AHRS based navigation system that is in the air and on the water uses a compass.

 

[Mimoid]

If you can do this, so can a computer.

No doubt - but if the Phantom had a visual guidance system capable of analyzing the actual image and performing the correct maneuvers to navigate, it would cost millions.

//Tom

 

[syotr]

No doubt - but if the Phantom had a visual guidance system capable of analyzing the actual image and performing the correct maneuvers to navigate, it would cost millions.

//Tom

The phantom has a GPS system which knows its exact location at every instance which is much more accurate that your ability to do it visually. No need for a visual guidance system.

 

[Mark The Droner]

 

[BudWalker]

There have been several posts recently with claims that depend on the P3's actual gyro drift rate.So I measured it on my P3 after the IMU was calibrated, which I hadn't done in 9 months*. Depending on the test conditions the gyroZ drift rate is .01 to .08 degrees per second. Suppose the P3 is moving at 20 MPH for 1000 feet. Then for these drift rates the error will be 3 feet (for the .01 Deg/S drift rate) and 24 feet for the .08 Deg/S drift rate). If the P3 moved 50 feet at 20 MPH these errors would be 1.8 inches and 14.4 inches, respectively.

To obtain these measurements I added a column "totalGyroZ" to DatCon that integrates and then sums the gyroZ value. I'll release a beta later today or tomorrow that will have this new column. I then did several test flights where I compared totalGyroZ to yaw. Ideally, given two points on the time axis the difference in totalGyroZ and the difference in yaw will be the same. Any discrepancy is the drift rate. Here's an example

The discrepancy in this example is 2 degrees which is a .0333 drift rate. This test was to see if yawing back and forth caused a higher drift rate.

The first test I did yielded a drift rate of .4 Deg/S. I then did an IMU calibration* and that drift rate is .now 03 to .08 depending on the test condition. All the actual flying tests yield rates in the .03 range. The higher rate occurs when the P3 sits on my workbench and doesn't move. I looked at a few .DAT file that I have here that I've accumulated. Generally, they have around .05 Deg/S drift rates. One has a .4 Deg/S drift rate.

I encourage others to do their own tests. DatCon 2.2.11 will be released later today or tomorrow. Go here.

*IMHO IMU calibrations aren't necessary unless there is an indication of a problem. After seeing a 13 fold improvement in this drift rate maybe I'll change that opinion and do IMU calibrations more often. I'm supposing the advice to do an IMU calibration at every firmware update is really about doing IMU calibrations periodically; i.e., it's not related to the firmware update.

 

[BudWalker]

About that yaw drift. Sorry for the tilted camera. I had the camera mounted on my cap. I guess I am not very level headed or I've been walking around these mountains so long, I have one leg shorter than the other.

I love this!! How can I do a Like times 10?