Help understanding ActiveTrack parallel mode

[SpudGT]

I am new to drones and still learning how to fly them … liking the Mavic Air 2 so far.

Posting my recent experience with ActiveTrack in an attempt to understand what went amiss, hopefully this will also help out other new folks like me.

I talked my wife into riding her bike up and down a trail next to a field covered in tall grass (the grass will be important later).

If you look at the trail you can see that it climbs up a small hill with no obstacles taller than about 5 feet.:

What I did:

• Got the drone up in the air at the desired altitude (11-12ft)
• Verified I was in ‘Normal’ mode, recording using 4k at 24fps
• Selected my wife as the subject, started ActiveTrack (parallel), and off she rode with the drone in tow
• I started out somewhat behind her but wanted to move around to just in front
• Held the roll stick hard over to circle the subject (kept it held)
• The drone moved around, but never reached the desired position
• Noticed that as the trail climbed the drone did not climb or stop, decided to see what would happen
• Eventually the drone met the grass, no harm done

What did I expect:

• I expected the drone to either climb or simply stop based on its downward obstacle sensors. Clearly my expectations were wrong, I need to understand this to avoid making the same mistakes again.

The manual says ‘The aircraft cannot sense obstacles in Parallel’. Makes some sense to me as the drone is mostly flying sideways in parallel mode and we have no side obstacle sensors. I guess ‘cannot sense obstacles’ really means can’t sense obstacles!

Interesting that the manual also says the following for trace mode: ‘The aircraft cannot avoid obstacles when there are roll and throttle sticks movements’. Wonder if this might also somehow apply to parallel mode as I had a constant roll input.

Looking at the mobile flight log the altitude remained constant at 11.5 feet (even after it met the grass). I guess this means altitude is relative to the home position and the drone has no concept of ‘above ground level’ altitude.

Questions:

• Should the drone have adjusted it’s altitude or stopped to avoid hitting the grass?
• Why couldn’t the drone circle my wife based on my roll input? At the time of the crash it was going 13mph, I thought parallel in Normal mode was capable of 26mph.

Any help understanding this would be greatly appreciated!

 

[Greekislandlover]

I've not had a lot of practice with active track, but I have had a couple of times playing with it recently.

OK, so you have the Mavic Air 2, and I have the Mavic 2 Zoom. They run different versions of active track, so the following may differ / not be of help. I've not tried adjusting the sticks when doing active track, as it's me trying to film me on my bike. Hard enough cycling!

Firstly, the automated modes do odd things at times, so it's good to test out in a safe environment such as a big open space - no trees!!! Me and them things have drone history Although the sensors are very good, as are the programmed modes you can't rely on the sensors, and if you see something happening always hit cancel and go up.

If you want to circle a cyclist you are probably better trying the trace mode with the slider set to circle. I've found the parallel works OK, but has a habit of deciding not to follow sideways, but go round behind me and follow.

I'm afraid is a process of trial and error!

My go with active track:-

 

[ShannonRawls]

• APAS or not, I think the drone should have climbed altitude. In fact, it looked like it actually descended into the grass.
  
  
• I can’t speak on this one. Not sure how that works. Sorry.

 

[SpudGT]

• In fact, it looked like it actually descended into the grass.

I think that may be an optical illusion ... it appeared to me that the drone maintained its original altitude and the ground came up to meet the drone.

Looking at the mobile flight log, the drone altitude remained between 11.5 and 12 ft, here is the snippet at the end (it is on its side at the end, still commanding roll):

 

[boblui]

I can only try to answer the first question.

If the drone was moving VERY SLOWLY, it should be able to ascend to avoid hitting grass. You can try that by flying the drone uphill close to the ground, again, VERY SLOWLY. The problem is when the drone moves fast, the image captured by the 3D camera at the bottom will be too blur for estimating distance. There is the IR sensor but I think it works on relatively flat surfaces only. The grass field may not be able to reflect IR light very well.

If you can share the flight log file, more insight can be gained.

 

[SpudGT]

If you can share the flight log file, more insight can be gained.

Here is the mobile txt and dat log files. Please let me know if I grabbed the correct DAT file, still new at this!

 

[DJake_2000]

As to following a cyclist with Activetrack on parallel mode, I've noticed that you get a better chance of making it work, if you angle the drone to track the object slightly from the front. If you try to track directly from the side, the drone seems to quite readily fall back to basically chasing the object from behind.

 

[Dave Maine]

Shooting while flying sideways may also make the jello effect from the camera shutter produce bothersome artifacts. It does on both my MA, and P4adv.

 

[SpudGT]

I haven't noticed the jello effect in the several 'sideways' test flights I have done so far with my MA2, but I haven't had many flights yet.

I was shooting at 4k 24fps, with a manual shutter speed of 1/50 and an ND16 filter attached (forget the ISO).

 

[boblui]

OK, I was wrong about the belly sensors. They were able to detect the ground clearance throughout your flight. The drone maintained a constant height ( red line below ) relative to the takeoff point and the ground clearance ( green line ) gradually reduced as the drone flew uphill until it hit the grass and roll over at 0.3 meter above the ground.



Is this normal ? I think so. This is the behavior of my M2P when flying up the ramp in a carpark :



I have not moved the throttle stick but the drone automatically climbed and maintained a ground clearance of 0.3 ~ 0.4 meter as it flew up the ramp. In your case, the drone should have been able to do the same but the grass was a bit too long hence the crash.

 

[Martinus]

is there any other software that can control the height automatically?

 

[SpudGT]

OK, I was wrong about the belly sensors. They were able to detect the ground clearance throughout your flight. The drone maintained a constant height ( red line below ) relative to the takeoff point and the ground clearance ( green line ) gradually reduced as the drone flew uphill until it hit the grass and roll over at 0.3 meter above the ground.

Thank you very much for analyzing my flight data, that was very good to see (at some point I need to get setup to do that myself).

So the belly sensors saw the ground coming up, but the drone didn't take action because it hadn't reached the minimum ground clearance limit. So it was likely sensing the ground, not the grass. I suppose the grass could have been a foot tall around the area of the crash.

Based on the manual I was concerned it ignored the belly sensors for other reasons, for example:

• It was in ActiveTrack parallel mode, so no obstacle detection
• I had a consistent roll command (which was listed as a limitation for track mode)

Did you have ActiveTrack on in your carpark test?

 

[fourbeer]

I miss the Boise foothills.