DJI Mavic, Air and Mini Drones
Friendly, Helpful & Knowledgeable Community
Join Us Now

Help understanding crash on a balcony

Status
Not open for further replies.
1) you clearly know the correct translation because you know the original Chinese, right?
The insight I gave is simple and obvious for someone who has read DJI manuals, flown DJI drones for years, analysed flight data from hundreds of incidents and helped numerous new flyers.

I really doubt that DJI, with the dominant position it has gained in important markets as the English speaking countries, would put out a product manual not vetted by engineers. For this reason, the fact that the manual does not reflect the actual operations should be way more worrying for the user of the product.

One week ago, you posted flight data that revealed almost no understanding of basic GPS or drone flying and asking for help.
But now no-one can tell you anything. You know more than anyone else.
 
Since the information to put in the manual is provided by engineers and the translation is continuously vetted and corrected, my educated guess is that there is a very high likelihood that any confusion present in the final version of the manual reflects some level of confusion in the firmware rather than creative storytelling by an inspired translator.
Lol.

Those of us following DJI from the start have seen enough cases of the manual being wrong through obvious quick and dirty work, "take the one for the previous model, change a few things, done" where some of the info was wrong because some bit was copy/pasted that didn't apply or stuff was forgotten, where the one thing that had changed and you'd want to check the manual to learn about isn't even mentioned.

There's poor English wording both in manual and app that's been misleading people for 6 years now.
 
The insight I gave is simple and obvious for someone who has read DJI manuals, flown DJI drones for years, analysed flight data from hundreds of incidents and helped numerous new flyers.



One week ago, you posted flight data that revealed almost no understanding of basic GPS or drone flying and asking for help.
But now no-one can tell you anything. You know more than anyone else.
Let's reason for a moment. The manual says something that is inaccurate and does not reflect the actual system operations. This is a FACT.

We can make different hypothesis on why that happened

I expressed my opinion that derives from reading many online sources and the manuals that there is a certain amount of confusion on the subject that reflects a lack of clarity by DJI that might well affect their firmware implementation.

You intervened ridiculing my hypothesis with a strawman argument ("You think that the software was created by the people that translated the manual ?") and presenting as indisputable certainty what is nothing else but your assumption i.e.: the firmware is perfect and the firmware engineers are great, therefore the blunder in the manual is necessarily due to "a subtle issue with translation".

So basically you are countering my doubts on the firmware by making the assumption that it's perfect. Not really the way a counterargument works.

In response I simply stated the obvious, that the only way you can claim with certainty it is a translation error is by knowing that the engineers provided the exact description in the original Chinese. Everything else is based on assumptions
I also pointed out a logical difficulty in your argument: even assuming the translators were given the correct description "when the drone is flying at speed in Sport Mode, the obstacle avoidance system cannot function properly because the forward sensors cannot see far enough forward when the drone is tilted forward at 30°" it is quite difficult that just by virtue of being bad they could come up with a completely different description that involves unrelated subsystems and make perfect sense per se (it just does not describe operations) as " "The Forward, Backward, Lateral Vision Systems and the Upward Infrared Sensing System are disabled in S-mode, which means the aircraft cannot sense obstacles on its route automatically."

At this point neither my suspicion of bugs in the firware nor your certainty of its perfection are more than hypothesis based on assumptions. The only difference is that I appear to know that, you only express certainties.
 
So basically you are countering my doubts on the firmware by making the assumption that it's perfect. Not really the way a counterargument works.
I was telling you how it is, from extensive actual experience.
You don't care about hearing facts from people who know.
You don't have the experience to know whether I'm correct, but assume you know better.
There is no point continuing to communicate with someone so argumentative and with so little knowledge.
Good luck with your flying in future.
 
  • Like
Reactions: Myetkt
Those of us following DJI from the start have seen enough cases of the manual being wrong through obvious quick and dirty work, "take the one for the previous model, change a few things, done" where some of the info was wrong because some bit was copy/pasted that didn't apply or stuff was forgotten, where the one thing that had changed and you'd want to check the manual to learn about isn't even mentioned.

There's poor English wording both in manual and app that's been misleading people for 6 years now.
There is no doubt that the manuals are poorly written and the translations are horrible. I was once again today trying to make sense of DJI's three types of RTH, Smart, Low Battery and Failsafe whose operations are explained by recourse to a mysterious fourth Emergency RTH, (hinting that possibly the manual was written by a fan of Alexandre Dumas and his Three Musketeers) and the weird table that shows multiple entries for actions to perform in reaction to the same warning.
The point I was trying to make is that not all that is wrong can be attributed with such certainty to bad translations and assume that the firmware is perfect. In particular, bad translations lead to incomprehensible paragraphs, but they rarely lead to a perfectly sounding sentence that involve completely different (and wrong) subsystems. Since translators, good or bad, translate what is given to them, they were either given wrong stuff or they botched a copy and paste job.
The same botched copy and paste job that I have seen time over time made with the code. Do you guys think that new code is developed for each product by the same experienced engineers? Nope. The old code, written by the previous engineers that have left the company because after one project they had become experienced enough to command a higher salary and hence too costly (chinese companies like only fresh out of college engineers with entry salaries) gets now taken over by rookies. Old code that is generally uncommented (commenting takes time) and has a few magic numbers here and there, maybe some ad-hoc thresholds for mode switching. And, by the way, who knows where the engineers that developed the control algorithms are now. Probably Whong did it. He left the company on May but he might have left some notes on that feedforward computation...
My money is on DJI not being the exception. Of course, it's just my hunch
 
One week ago, you posted flight data that revealed almost no understanding of basic GPS or drone flying and asking for help.
But now no-one can tell you anything. You know more than anyone else.

was telling you how it is, from extensive actual experience.
You don't care about hearing facts from people who know.
You don't have the experience to know whether I'm correct, but assume you know better.
There is no point continuing to communicate with someone so argumentative and with so little knowledge.
Good luck with your flying in future.
Ok, I got really tired of this. Since I keep getting dragged into this sort of useless discussions by your hissy fits, here are my final words. I really have no interest in any competition on who has the bigger balls and I have already wasted enough time replying to the insults of your cohorts. I am new to the site, I don't give a **** about who feels like the king of the hill. I know what I know and I'm not ashamed of what I don't. However, I find it pretty arrogant for you to judge from a single log that I posted what I know and what I don't. Especially when one of your favorite mantras, "ATTI is P-GPS with no GPS" immediately gives away that, despite your vaunted vast experience and knowledge, you have no clue about the difference between closed-loop and open-loop. Arrogance seems to me the cypher of your interactions: you do not formulate possibilities, you spout edicts: this is what happened and this is totally wrong. I am sure your experience makes you right most of the time. However, in my case some of your edicts were not compatible with my experience of the event. So after your salvo of three consecutive posts (#21, #22, #23) in which, with your trademark grace, you ridiculed a strawman version of my argument (providing such eye-opening revelations as "Losing GPS does not cause a crash. How you handle the drone without GPS could well cause a crash.") and declared the thread over, I shut up and did my research. I discovered CsvView, analyzed the flight log and found out evidence that contradicted your edicts. Your reaction when I posted it? You boasted that you were not going to read it because I am too wordy. Pretty lame excuse, if you ask me. And when I challenged you to provide the evidence to support your conclusions, I heard only crickets. Considering your lack of interest in holding your fortress, your stated dislike of me, and the amount of people in the forum surely in need of your help, I had very little hope to see you again in here. Yet here you are again, with your snarky strawman remarks on a hypothesis I formulated decreeing that all is good but the translators of the manuals. And after I remark the obvious, I get treated to your victimism, and of course, your insults. And please do not pretend you did not come here swinging looking for it
 
Especially when one of your favorite mantras, "ATTI is P-GPS with no GPS" immediately gives away that, despite your vaunted vast experience and knowledge, you have no clue about the difference between closed-loop and open-loop.
Atti is P-GPS without GPS.
If you can't understand the basic concepts, you haven't much chance with more complicated issues.
 
There is no point continuing to communicate with someone so argumentative and with so little knowledge.
Good luck with your flying in future.
Meta4, I understand you have different obligations, but this is precisely why Grandpa left the room.
 
Atti is P-GPS without GPS.
If you can't understand the basic concepts, you haven't much chance with more complicated issues.
Sure, how much variation do you expect in the system poles as GPS is lost? Grandpa can answer as well if he's peeping in
Atti Mode doesn't cause "instability".
The drone will maintain height and is perfectly controllable.
For a basic introduction to the concepts of stability and controllability, I recommend this paper by George Raju.
 
Last edited:
Sure, how much variation do you expect in the system poles as GPS is lost? The question is open as well to those outside of the room, peeping through the keyhole.
Dude. Get a toy drone and learn how to fly it. No barometer, no GPS.
 
you have no clue about the difference between closed-loop and open-loop.
I have no clue about that difference, could you possibly give me a quick explanation? Been wondering about that ever since you mentioned it at the first of this thread. TIA
 
  • Like
Reactions: CodaDiLupo
I have no clue about that difference, could you possibly give me a quick explanation? Been wondering about that ever since you mentioned it at the first of this thread. TIA
An example of an open-loop system would be the drone in atti mode. The aircraft simply responds to stick input from the remote. The drone will not maintain position and will be blown around by the wind.
In a closed-loop system (or a feedback system) the stick commands are modified for a desired result. When both sticks are centered, GPS and/or VPS are used to maintain a stationary position.
 
Last edited:
  • Like
Reactions: CodaDiLupo
An example of an open-loop system would be the drone in atti mode. The aircraft simply responds to stick input from the remote. The drone will not maintain position and will be blown around by the wind.
In a closed-loop system, the stick commands are modified for a desired result. When both sticks are centered, GPS and/or VPS are used to maintain a stationary position.
In either case, the quad is still for the most part a fly by wire operation.
 
  • Like
Reactions: Grandpa
I have no clue about that difference, could you possibly give me a quick explanation? Been wondering about that ever since you mentioned it at the first of this thread. TIA
Grandpa pointed out correctly when open loop and closed loop operations are used in the drone flight modes. And this was the central point of my disagreement with the motto "ATTI is P-GPS with no GPS" because it focuses on the presence/absence of GPS as the difference between the two, while the fundamental difference between the two modes is that in ATTI we fly the drone open loop while in P-GPS we fly it closed-loop. P-GPS is not the only mode in which we fly the drone closed-loop, as we do it in P-OPTI as well, with the only secondary difference that the drone position is computed by using the VPS system instead of the GPS system.

The best way to think about open loop vs open loop is to think about the concept of manual control vs. automatic control. Control means acting on an object (say the drone) to make it achieve a desired goal (say fly from A to B along a certain path). This can be done manually or automatically.
To identify the core difference, it helps to reduce complexity by using for a moment a simpler system, like a thermostat to control temperature in a room,
Old manual thermostats have a lever with different numbered positions (say 0 to 10). Each position represents a certain amount of heat generated by the burner and delivered to the room. If you want the room to be at 70 degrees, you do not have an immediate way to achieve that goal, but you have to play with the lever in a trial and error type of way. Say you start setting the lever at 8, wait and realize the room became too hot, then move the lever down to 5. and wait. If it gets too cold, you bring it up to 6 and let's say it's the correct position that brings the room to 70 degrees. The point is that it took several iterations to get to where you want. With this simple system, the second time that you use the thermostat, you would directly set the lever at 6. However, this is not the end of the issue, because many conditions can change both internal to the system or external. Suppose that all of a sudden the outside temperature plummets and your windows are drafty: you need to start again the process of manual lever adjustments. Or, as an example of internal system change, say your burner loses efficiency and starts having higher heat losses: you need to redetermine the new lever position. Same story if you move that thermostat to a larger room. Basically every time the situation changes you have to re-learn how to adjust the lever to get the temperature you want. This iterative process of adjustments is the essence of manual control: you know the desired goal (room at 70 degrees), you act on the system (move the lever) and measure the result of that action (current temperature in the room). Based on the error between current and desired room temperature, called the tracking error, (too hot? too cold? by how much?) you readjust the lever and repeat until you are satisfied that the residual error is small enough.

In a closed loop system this whole process is completely automated by the use of a feedback controller. When you use a modern thermostat, you don't have to play with lever positions, but you just set the desired temperature to 70 degrees. Upon receiving that command the controller reads the sensors that provides the room temperature, computes the tracking error and, based on its value, starts moving the lever to compensate for the difference.

The advantages of closed loop are many, some immediately obvious, some less so:
1) first of all you do not have to nurse the thermostat. Just set the desired temp and you are done
2) speed and precision: an automatic controller is an electronic device that can read the sensor, compute the tracking error and apply corrective action many times in a second and based on many more factors than those that are easy to read for a human; for example not just the value of the temperature error, but also how fast it is changing, or how to reach the goal by minimizing the amount of energy used.
3) automatic adjustment to changing conditions: since it operates based on the error, an automatic controller will strive to achieve the goal you set indipendently of the cause of the error, be it the result of a suddenly plummeted external temperature and windows draft, decreased burner efficiency or a larger room. As conditions change making the room temp different from the desired value, the controller will act on the system to bring it there.

Let's go back to our drone. In closed-loop operations, by acting on the joystick you "set the desired temperature", to use our analogy. In other words, you tell the control system where you want to fly (by moving the joystick in a certain direction) and how fast (by the entity of the move). The controller reads that command, computes where the drone currently is (either by reading its GPS sensors in P-GPS mode or its VPS sensors in P-OPTI mode and doing the appropriate computations) and determines the amount of current to send to the electric motors to achieve the propellers’ action necessary to do what you commanded. If there is a sudden burst of wind trying to move the drone in another direction, the controller sees an increased tracking error in that direction and generates the appropriate corrective action, i.e., adjusts the current sent to the appropriate motors to keep the drone on the desired trajectory independently of the wind.

If neither GPS or VPS is available, the controller has no way to know the current position and hence cannot compute how much current to send to the motors to get where you commanded. For this reason, no closed loop operations are possible and the drone enters ATTI mode, in which the operations are open loop. Going back to our analogy, by acting on the joystick you are not setting anymore “the desired temperature" but you are setting “the lever position of that old thermostat”.
In other words by moving the joystick in one direction you select which motors you want to act on and by the entity of the joystick move you determine how much current you are sending them. Since knowing how much current the motors need to bring the drone where you want is not very intuitive and depends on many external factors (like, e.g., the wind direction and strength) you have to resort to the trial and error mode we discussed: move the lever in the desired direction, look with your eyes whether that brings the drone too far or not close enough, at which moving speed, and make the appropriate correction. Note that unlike the thermostat case, where one wants to reach a simple static goal (a set room temp) in flying the drone one has to track a dynamic goal, i.e., bring the drone to a continuously updated desired position along a path that reaches the landing point. Clearly one can do this only with the speed and precision that are allowed by his visual ability to estimate the drone position and his reaction time to perform a corrective action. Both can be improved by practise, gaining familiarity with the system and an instinctive feel for how to move the joystick, but it’s highly unlikely that one will ever manage to get close to the level of performance guaranteed by an automatic controller. If there are obstacles closeby, one does not have many margins for error and adjustments so he needs to be very cautious with the joystick, making the drone move super slow. However, if there is a sudden burst of wind, there might not be enough time to correct before the drone hits an obstacle. To further complicate things, the drone in open loop is not a stable system, which means that if you do nothing with the joystick, the drone does not stay put. Every small perturbation such for example a little wind, or a small asymmetry in the propellers, will cause it to start drifting, making active joystick operations necessary even to keep the drone in position by counteracting those external disturbances.
This is the reason why one must be extremely prepared to react to a sudden switch to ATTI mode. My case shows how dangerous being unprepared can be. The drone was hovering a couple of meters from me standing on the balcony as it had on the way out and there was plenty of light, so even without GPS it was operating in stable closed-loop conditions in P-OPTI mode. I was trying to bring the drone over the balcony to land it but I was prevented from doing so by the Obstacle Avoidance (OA) system. Since I had a clear view of all obstacles that were present, I disabled OA and prepared to move the drone over the landing point under stable P-OPTI closed loop. However, a few seconds after disabling OA, something happened that made the VPS unexpectedly (to me) fail and the drone switched to ATTI mode. The fact that I did not expect this coupled with the wind that was present and the closeness to obstacles resulted in a crash that I was unable to avoid. All I could muster was a last disperate attempt to move away, but it was too late.

I apologize for the long reply, but I believe this is a fundamental concept to understand. Since both in closed-loop (P-GPS and P-OPTI) and open loop (ATTI) we are using the same joystick, it’s not immediately obvious that by operating it in the two situations we are actually doing a very different thing. In closed-loop we are just telling what we desire and do not have to worry about how to achieve it. In open-loop we need to tell exactly how to achieve it based on all the external factors that are present and that can vary very rapidly.
 
Last edited:
This is a long-winded and unnecessarily complicated way to explain something that's quite simple.
It comes down to ...
If neither GPS or VPS is available, the controller has no way to know the current position and hence cannot compute how much current to send to the motors to get where you command. For this reason, no closed loop operations are possible and the drone enters ATTI mode,
Which seems to be just a long way of saying that Atti Mode, is what's left from P-GPS mode when the satellites are not available.
 
Last edited:
And this was the central point of my disagreement with the motto "ATTI is P-GPS with no GPS" because it focuses on the presence/absence of GPS as the difference between the two, while the fundamental difference between the two modes is that in ATTI we fly the drone open loop while in P-GPS we fly it closed-loop.
You missed the whole point of the high level approach to the discussion, linking to a paper about control theory is pointless in that regard.

The point of the "ATTI is P-GPS without GPS" was to try to make you understand that from the operator's perspective that is how it is, and explains what's in some log fields. You have selected P-GPS mode, if GPS is unavailable it falls back to OPTI, if that's unavailable as well it falls back to ATTI - but neither of them is manually selectable, so the selected "desired mode" is still P-GPS, which is why the relevant log fields will show P-GPS even when the aircraft has entered a fallback state and the "effective" mode is OPTI or ATTI.

Again understanding of the underlying control theory is irrelevant - there are a few control modes, and the aircraft will be using one of those (the most automated one) based on sensor availability/confidence, it's all that matters, along with the understanding of what each one does, what it requires to be functional and thus when they are more likely to become unavailable, and the awareness that regardless of that the more advanced modes are subject to becoming unavailable at any time in which case the pilot must be able to deal with ATTI.
 
Last edited:
You missed the whole point of the high level approach to the discussion, linking to a paper about control theory is pointless in that regard.
I think you are making some confusion. The link to the control paper was not a response to this question. It was a reply to statements that had been made previously by Meta4 that showed confusion between the concepts of stability and controllability and wrongly claimed that the drone inn ATTI mode is not unstable.

The point of the "ATTI is P-GPS without GPS" was to try to make you understand that from the operator's perspective that is how it is, and explains what's in some log fields. You have selected P-GPS mode, if GPS is unavailable it falls back to OPTI, if that's unavailable as well it falls back to ATTI - but neither of them is manually selectable, so the selected "desired mode" is still P-GPS, which is why the relevant log fields will show P-GPS even when the aircraft has entered a fallback state and the "effective" mode is OPTI or ATTI.
I hope you realize that all you had to add to explain what you mean by that motto provides enough evidence of the fact the that it is per se meaningless. It suggests the wrong idea that without GPS the system switches necessarily to ATTI. It does not even mention P-OPTI. The system is perfectly able to fly in closed loop mode without GPS by using the VPS.
By the way, this is eaxctly what happened in my flight on the way out and partially on the way back and which was denied by you guys only on the basis of the fact that the logs do not report it. If you knew of this issue, then you both provided me a willingly misleading answer when you told me that P-OPTI was never entered. It was a wrong statement easily disprovable by looking at other state signals. However, Meta4 refused to reconsider his wrong analysis as he obviously believes that his experience makes his judgment infallible and above the actual log evidence.
It is funny that you guys keep attacking me for what I write to someone else but refuse to debate the direct questions I aked you. This time you feel you have to poop on a thoughtful response I wrote to someone else using a considerable amount of my free time.
I don't know why is so difficult to realize that pooping on something that is freely offered and not directly to you is not really a great looking behavior. Don't read if you don't like
It makes you look like a better person.
 
I don't know why is so difficult to realize that pooping on something that is freely offered and not directly to you is not really a great looking behavior. Don't read if you don't like
It makes you look like a better person.
And this only applies to others ??
 
Status
Not open for further replies.
Lycus Tech Mavic Air 3 Case

DJI Drone Deals

New Threads

Forum statistics

Threads
131,560
Messages
1,564,252
Members
160,454
Latest member
Rejdmast