VPS Optical Sensor vs GPS position hold?
1145 7 2023-8-2
Uploading and Loding Picture ...(0/1)
o(^-^)o
Zbip57
lvl.4

Canada
Offline

The downward VPS sensor system on the Mavic Mini is composed of two separate functions.  The infrared sensors, when within operating range, are for measuring height above ground. I'm not concerned about those in this post.  The optical sensor, when within operating range and sufficient lighting etc, is used for horizontal position hold when the drone is hovering with the control sticks centred.

The User Manual for the Mavic Mini says,


"The Downward Vision System and Infrared Sensing System help the aircraft maintain its current position, hover in place
more precisely, and to fly indoors or in other environments where GPS is unavailable."
[...]
"When GPS is unavailable, the Downward Vision System is enabled if the surface has a discernable surface and sufficient light."

I have experienced that when flying over a river deep down in a narrow gorge.  The GPS signal was blocked by the canyon sides.  So the Mini relied on its optical sensor to hold position whenever in hover.  That worked fine as long as the sensor fixed on the non-moving rocks at the river's edge.  But while hovering over the moving water, the Mini tracked the foam floating past on the water's surface and started drifting downstream.  That was unexpected at the time, but easily corrected with control stick input.  In hindsight it all makes perfect sense.


But, from what it says in the User Manual, I have always assumed the optical sensor is only enabled when GPS is unavailable.  When hovering over a lake out in the open, with an unobstructed view of a sufficient number of GPS satellites, the Mini should rely on its solid GPS fix to hold a steady position while hovering, and not be distracted by the optical sensor tracking rippling wave patterns or leaves drifting by on the water.  

If both GPS and the optical sensor are active, which one takes priority?  I have always been convinced that GPS should, and does, take priority.  However, I was surprised to discover that's not true at all, as demonstrated in this video experiment.  

I tried these three scenarios:
  • No GPS, optical sensor only
  • No GPS, optical sensor disabled, ATTI-mode
  • GPS and optical sensor both enabled.
See the surprising results...





2023-8-2
Use props
Bashy
Captain
Flight distance : 2354357 ft
  • >>>
United Kingdom
Offline

The sensors kick in when youre within the sensor range, also, the sensors do not like water so you must be on the ball and make sure youre watching the drone at all times when your below say 9m and over water or other glassy surface, acutally, this is in the manual but a bit more in depth. See page 16
2023-8-2
Use props
120ccpm
Second Officer
Flight distance : 1396755 ft
United States
Offline

Thanks for doing the experiment, although I have always assumed that the Downward Position System was always on, as I doubt GPS (even with strong signal) is accurate enough to hold the aircraft in position with the level of precision shown by the Mini.
As for the manual, in some places it seems to indicate that visual positioning is always active, for example here:



Regarding the quotes you mentioned, the first sentence could be interpreted this way: visual positioning (ALWAYS) helps the aircraft maintain its position, and it (ALSO) allows the aircraft to fly when GPS is unavailable. The second sentence instead seems incorrect, because visual positioning requires good surface and light regardless of GPS status.
2023-8-3
Use props
Zbip57
lvl.4

Canada
Offline

I'll admit my previous assumption was incorrect.  I could blame that on a misunderstanding of what the User Manual says, as the translation from Chinese to English (Chinglish) is not always accurate enough.  But it's more a case of me incorrectly assuming how the system should operate, rather than how it actually does operate.

It makes sense to me that, in the absence of GPS, the optical sensor is used for position hold (when within range, when it has sufficient lighting, when it can sufficiently distinguish patterns, etc, etc, etc).  And, in hindsight, it makes sense to me that the optical sensor is always enabled and in use, even with sufficient GPS, because as you pointed out the Mini really does have extraordinarily precise position holding capability when hovering.

And it makes sense that the Mini would follow the motion of any detected moving pattern, when in the absence of sufficient GPS signal, because I've actually experienced that.  What didn't, and still doesn't, make sense to me is that the optical sensor on the Mini will continue to cause it to follow a moving pattern despite the GPS data showing it is no longer holding a stationary position.

Even though GPS is less precise than the optical sensor, eventually the drone will have moved far enough while tracking a floating leaf on the river's surface that the GPS data must show conclusively that the drone is no longer hovering in the same spot that it should be.

As you can see by my reaction in the video above, I was stunned to realize that optical sensor actually does take priority, regardless of what the GPS data is showing.

I made a video nearly ten years ago with my original Phantom-1, which even way back then had sufficiently accurate position hold capability using GPS data alone.  (Note the altitude hold was less accurate only because the throttle stick on the original transmitter wasn't spring-centred.  If you didn't manually set the throttle exactly centred the drone would slowly sink or rise.)  In that video I showed how, when the Phantom was hovering in place, if you grabbed a hold of its leg and dragged it away from that position, when released, it would fly back to resume its original position.


So why shouldn't the Mini also be capable of noticing and preventing itself from drifting away from its intended hover location when it has a solid GPS fix?  Instead, it lets itself be distracted into wandering away following a floating leaf, or tracking a carpet being pulled away from underneath it when the optical sensor takes priority.  That makes no sense at all to me.



2023-8-3
Use props
120ccpm
Second Officer
Flight distance : 1396755 ft
United States
Offline

So why shouldn't the Mini also be capable of noticing and preventing itself from drifting away from its intended hover location when it has a solid GPS fix?  Instead, it lets itself be distracted into wandering away following a floating leaf, or tracking a carpet being pulled away from underneath it when the optical sensor takes priority.  That makes no sense at all to me.

I guess DJI assumed that "carpets being pulled away from underneath" is not a common occurrence... :-)
In all seriousness though, it's possible they implemented something a bit more sophisticated, where the aircraft constantly compares the quality of the GPS and vision lock, and prioritizes accordingly. It's one of those things you can only know for sure by looking at the firmware code.

2023-8-3
Use props
Zbip57
lvl.4

Canada
Offline

"you can only know for sure by looking at the firmware code."

That would certainly be interesting.  But, since I'm not smart enough to do that, I'll need to come up with better experiments to test the system.

The first question should be, how far will the Mini drift away from its intended hover location while tracking a floating leaf (or pulled carpet) before the GPS data eventually, if at all, convinces it to return to its original location?

It only makes any sense to me when it has insufficient GPS fix.  But what if it does have a good GPS fix, while hovering over a clean white floor, or field of snow, with no discernible pattern features?  It should stay hovering in place quite nicely with the GPS alone.  But if a mouse wanders across that white floor, within view of the optical sensor, would the drone move all by itself following the mouse while ignoring the GPS data signalling the drone's displacement?   Apparently, yes.

2023-8-3
Use props
120ccpm
Second Officer
Flight distance : 1396755 ft
United States
Offline

Yes, you can do some tests to see how far it moves when you slide the carpet and also how close the carpet has to be for the aircraft to start following it. I doubt a "mouse wandering across a white floor" is going to be enough to trick the vision system. At the end of the day it's just for fun and science, because it's not something we can change. And its impact is limited to very specific situations (like the one you encountered in the canyon).
2023-8-3
Use props
Zbip57
lvl.4

Canada
Offline

"its impact is limited to very specific situations"

Exactly.  If you're just flying around with active control inputs the situation will never arise at all.  It's only when hovering with sticks centred that this could possibly happen.

It's important though to realize that it could happen, and to be prepared to take control if you see the drone wandering off unexpectedly.  

If you leave the drone hovering expecting it to stay put while you divert your attention to something else, either make sure it's high enough that the optical sensor is well out of range, or if low enough make sure the optical sensor sees enough solid stuff that's not moving.


2023-8-3
Use props
Advanced
You need to log in before you can reply Login | Register now

Credit Rules