RTL during LANDING on Quadplane Triggered by Low Battery

ovolkmann · June 23, 2019, 10:36am

Hi Guys,

I thought the issue with VTOL platforms going to RTL on battery failsafe was fixed in the latest firmware version but sadly, it doesn’t seem to be and almost cost me the platform. I am flying a very modified Nimbus setup with the CUBE and latest stable firmware (3.9.8). I was flying an automatic mission (from start to finish with VTOL_LAND as the last command), the battery was at 22.3V right before the transition to copter mode and then due to the high amp draw, got sucked down to below 21.7V (where I have my Low Battery Failsafe set). At approximately 4 meters from the ground during the landing phase, the platform decided to initiate RTL and proceeded to transition to Fixed Wing and fly home (actually, I have RTL_AUTOLAND setup so that it repeats the landing approach). I switched out of Auto and back into it to over-ride the RTL and it proceeded to return to the landing zone where I took over manual control and landed in Q_LOITER.

This is VERY DANGEROUS!!! I was a few seconds away from watching the platform fly into a power-line and tree when I grabbed the remote and took over control. Please please please can we do something about this? The logs for this flight can be found here: https://we.tl/t-2JTR46H80Z

The Telemetry Log can be found here: https://we.tl/t-iUdAvnvGGI

If someone can get back to me on this asap that would be fantastic. This issue causes major concerns for operations where you do not want to land where you took off from (ie, home) and removes the viability of low battery failsafe as with it enabled, this issue happens, and with is disabled, well, you dont have that security.

iampete · June 23, 2019, 12:09pm

looks like it is fixed on master

github.com

ArduPilot/ardupilot/blob/ad2d7dac3ea892a38a9e1c10ac72b8197f4a5ab7/ArduPlane/events.cpp#L164


        case Failsafe_Action_Land:
            if (flight_stage != AP_Vehicle::FixedWing::FLIGHT_LAND && control_mode != &mode_qland) {
                // never stop a landing if we were already committed
                if (plane.mission.jump_to_landing_sequence()) {
                    plane.set_mode(mode_auto, MODE_REASON_BATTERY_FAILSAFE);
                    break;
                }
            }
            FALLTHROUGH;
        case Failsafe_Action_RTL:
            if (flight_stage != AP_Vehicle::FixedWing::FLIGHT_LAND && control_mode != &mode_qland ) {
                // never stop a landing if we were already committed
                set_mode(mode_rtl, MODE_REASON_BATTERY_FAILSAFE);
                aparm.throttle_cruise.load();
            }
            break;


        case Failsafe_Action_Terminate:
#if ADVANCED_FAILSAFE == ENABLED
            char battery_type_str[17];
            snprintf(battery_type_str, 17, "%s battery", type_str);

although I think if it was a land mission item in auto rather than being in land mode it would be missed

ovolkmann · June 23, 2019, 2:10pm

Hi Peter, the last command was “VTOL_LAND” in the automatic waypoint mission. Why would that be excluded from the protocol you have listed above?

ovolkmann · June 23, 2019, 2:11pm

Do my logs show anything different?

iampete · June 23, 2019, 3:06pm

VTOL_LAND is a mission command within auto mode, so it is in auto mode rather than land mode. It looks like we would need a few changes to the flight_stage code so that a auto mode landing in vtol would stop a RTL

ovolkmann · June 23, 2019, 3:55pm

That would be ideal. So at this point in time, there is no way to safely include a VTOL landing in an automatic waypoint mission is there?

pompecukor · June 23, 2019, 5:15pm

I reported this and opened an issue about it back in April

github.com/ArduPilot/ardupilot

Hidden (Major) dangers in using Do_land_start with any Quadplane

opened 06:13PM - 11 Apr 19 UTC

pompecukor

VTOL-Plane

### Bug report **Issue details** We realized the following undesired and v…ery risky situation with current code during a random training. We are (I am guessing the few, as no one seemed to have noticed this simple problem) who prefer full (preset) control of of VTOL landing, especially with regards to landing approach and so on. So we never rely exclusively on QRTL and/or RTL land. Rather we use VTOL land in every mission with approach and all. Then for RTL we have a Do_land_start somewhere, a few waypoints from the VTOL land. For this always have RTL_AUTOLAND set to: 2 (which makes the mission pointer jump to whatever comes right after the Do_land_start "marker". This is a (best) practice that we carried over from Traditional fixed wing. So in case of a fail-safe and the resulting RTL. The Unit would fly to the landing approach that we set with the marker and proceed from there. I know there is a wonderful new feature in the current master that lets landing be almost automated in case of RTL with the auto-loiter to alt then transition into the wind and vtol land at HOME point. However we found this approach too risky for our clients, as it require above average situational awareness. A simple Mobile phone tower or taller building 90m behind the home possition could cause a crash, due to loitering over home position. Same goes for when flying from a point with elevation at the back, like the side of a hill or next to a very tall building. We fund it safer and more straightforward, to always just have a predefined approach and landing. Yes the automated one is much simpler, but there is a risk as described and only needs one time to forget that the plane would make a big circle at low AGL over the home point before landing. The only place I would advise to use that feature is, if the user is always flying in more or less flat terrain and never next to any object taller than 35m. Anyway back to the topic. The above approach works well as usually. Here is a simple mission example of the safest way that I know to operate a VTOL in the current code (till we get some smart RTL features in the mix). ![image](https://user-images.githubusercontent.com/31744367/55977527-53f2a380-5c8f-11e9-9a4d-edc2d64ccb63.png) As you can see in this way, you can go as far and as high as you want (with for example Guided mode), then proceeding with auto or an RTL/failsafe at anytime will return you to WP6 at a safe altitude (which you have set in RTL_ALT parameter) and then proceed to loiter to a more suitable altitude for transition on WP7 (loiter to alt), Then a predefined approach vector set with WP8, with also initiates transition and finally VTOL land on the set point. So again, this is fine and dandy. However today we realized that there was a huge risk to this with Quadplanes. If at anytime during the mission, a fail-safe occurs. Then the UAV would go directly to WP6. For traditional fixed wing this is not really and issue. However with Quadplane this should NOT happened anytime during VTOL mode. Unfortunately it does. We were out with a client practicing a bit of Qloiter after some missions as we still had some energy left on the last pack. Suddenly, the UAV started to climb on its own. Immediately I realized what was happening and snatched the RC from the client and took over the FS, but if they were alone it would have been another story. What happened was the (over safe) Low voltage Fail-safe kicked in and the VTOL wanted to climb back to transition altitude and transition to Fixed-wing, then fly to WP 6 and redo everything from that. We would have crashed as the battery might have not been enough for a full up and down transition. LONG STORY SHORT. Anyone is free to test this in SITL sim. RTL and the commanded Do_land-start gets called even during VTOL phases, expecially scary is VTOL_LAND, VTOL_TAKEOFF, Qstab and Qloiter. I did not test Qland. **Fails safe should never initiate in those modes with Do_land_start.** I don't think it needs a big imagination, but I will give an example: having any sort of Low voltage FS set to a very low level. Imagine the UAV returning from a long mission with a mission like the above. And very low battery, but still safe. Then just during the VTOL land phase the low voltage FS kicks in, **that plane is doomed!**. **And it gets worse!** I tested in SITL. The biggest danger is not even a depleted battery, but rather any sort of failsafe. Say for example a radio failsafe for any reason. As long as RTL is activated. Try this in the sim. Hit auto in the above mission, and as soon as the VTOL starts to climb, lets say at 3meters. Hit RTL. See what happens? **It starts to transition at that 3 meters crashing into any and everything in its way.** Same goes for Landing phase, only in that case a low battery only adds to the "excitement". The simple reason being that all it sees after a failsafe is engaged, is a fixed wing waypoint somewhere else. So it just tries to transition immediately, to get there. I hope this makes sense. **Version** Lastest master and obvious all of the version before will be the same. **Platform** [ ] All [ ] AntennaTracker [ ] Copter [X] Plane [ ] Rover [ ] Submarine **Airframe type** Quadplane

For now DO NOT use RTL_Autoland = 2. Just set to zero.

xfacta · June 23, 2019, 9:00pm

I’m not sure, but would setting a rally point at the desired Land position mean any RTL during the last phases of flight just send the craft to the rally point anyway?

pompecukor · June 24, 2019, 8:25am

@iampete

Dear Peter,

I have just made and update with a very simple solution in the issue that I reported on github (see my last post in that topic). Can you please take a look and see what you think? If this was implemented, then all but the most extreme cases will be able to operate safely.

Cheers
Alex

ovolkmann · June 24, 2019, 8:30am

Please developers, this issue is really important. We urgently need a solution that allows for VTOL_LAND during an autoland to ignore RTL on failsafe event (no rc, battery etc). This is a safety issue

Shane_R · June 28, 2019, 5:25pm

Last fall we were testing BVLOS landings in the Black Rock Desert.

Lost radio connection as the plane touched down. I think there is a 10 second wait before the failsafe initiates the RTL so the the plane was on the ground but had not disarmed it’s self.

As I walked up it leaped up and immediately initiated a transition.

This is a 3 meter quadplane that weighs around 35 lbs…

I’m reasonably quick on my feet so I got out of the way in time and we were in the middle of a salt flat so there wasn’t anything to hit.

Regardless I can attest it’s a very dangerous behavior.

-Shane

ton999 · June 29, 2019, 5:50am

Hi Guys, I am worried about this Fail safe problem, but fortunately I never got this problem during my Vtol flight auto mission. So, before we have permanent solution, what do you think “temporary” solution to avoid this FS problem??
In think I will not set low batt. FS (since this issue may happen anytime). Also I wish can set the delay disarm (after landing) to Zeo (So ardupilot will Disarm immediately after landing), but unfortunately can not do that now. Based on current firmware, the minimum value is 1 second. I agree that this FS problem is very very Serious issue to be solved as soon as possible. I suggest The logic as follows:
1). FS (and RTL) can not be triggered during landing process.
2). Landing mode can override RTL or FS mode.
3). Disarm delay value (after landing) can be set to zero (autopilot disarmed immediately after landing).
I hope developers consider Safety as number 1 priority to be solved.
Regards

a75g · May 5, 2020, 4:35am

Was this issue ever resolved in newer Arduplane?

vishvesh · January 6, 2022, 10:42am

I recently encountered the same problem. Was this fixed? seems to be a dangerous situation to be in.

tkoivuro · January 12, 2022, 1:37pm

It would be absolutely amazing if there was an option (bitmask) to define flight modes where the battery failsafe is active.

We are using large vtol planes and have experienced very dangerous low battery failsafe rtl events during manual landing in qstabilize mode.

In our case we need battery failsafes only in AUTO mode.

amilcarlucas · January 12, 2022, 1:46pm

Take a look at the third bit of FS_OPTIONS parameter.

tkoivuro · January 12, 2022, 6:21pm

Arduplane documentation doesn’t seem to contain FS_OPTIONS

amilcarlucas · January 12, 2022, 6:26pm

Opps, sorry for that. It seams to be a copter only parameter

vishvesh · April 25, 2022, 12:05pm

Was there any solution to this issue?

tkoivuro · May 7, 2022, 11:38am

We “solved” the problem by using companion computer on board to disable and enable battery failsafes depending on flight mode.

We need failsafes only in AUTO and RTL mode.