Merkwürdiges Wendeverhalten am Bahnende

Ich glaube nochmal vorsichtig Bescheid geben zu können das ich eine funktionierende Lösung gefunden hab.
Hab in der .317 nochmal die Zeile rausgenommen weil das ja auch bei der .318 schon geholfen hatte.
- get_turn_direction_preference -> auskommentieren

Jetzt dreht er wieder wie erwartet eigentlich immer in die passende Richtung, keine einzige fehlerhafte Drehung gesehen UND kein Geruckel bis jetzt!
Zu mindestens in den ersten 2h, länger konnte ich noch nicht testen, aber das sieht schon mal deutlich besser als in der .318 aus mit den fixes aus!

Hab mir eben auch mal die Stats angeschaut.
Hier der Mähvorgang vom Vortag über die komplette Fläche mit der .317 vom ohne den Workaround.

Grün - gefahrene Meter /100 damit es in den Graph passt, also 50 = 5000m
Rot - Gesamte Obstacles
Gelb - No Motion Obstacle Triggers durchs GPS

Wodurch der Versatz zwischen Gesamt und GPS Obstacles kommt kann ich nicht feststellen, die anderen Obstacle Triggers sind alle leer, wird also vom MQTT leider nicht erfasst.

2024-04-27-20-08-59-Desktop-Dashboards-Grafana.png



Hier dann nochmal das gleiche rangezoomt auf die ersten 2h von dem ganzen Tag:
2024-04-27-20-09-18-Desktop-Dashboards-Grafana.png

Ab 12:20 Uhr wirkt es schon fast als wäre da ein Rhythmus dahinter...

Und hier die 2h vom Test Heute mit der einen Zeile auskommentiert:
2024-04-27-20-08-33-Desktop-Dashboards-Grafana.png

Der eine Obstacle Trigger kam von einer Baumwurzel wo er sich festgefahren hatte, mich wundert aber dass das nicht auch als GPS no Motion Trigger kommt.
Kann eigentlich kein Overload gewesen sein, die Räder drehen auf den Wurzeln einfach durch und dann kommt der keinen Meter vom Fleck.

Aber jedenfalls dezent zu sehen dass er jetzt passend fährt und nicht andauernd sich tot dreht am Bahnende bis ein Obstacle kommt.
Gefahrene Strecke über die Zeit ist nicht vergleichbar weil der Mähplan anders war, bei großen Flächen macht der natürlich deutlich mehr Meter als bei kurzen.
 
Sorry, stupid question. I recognized myself that I'm able to look into your code.

But I recognized you are using only a map.isInsidePerimeter function instead of the map.isInsidePerimeterOutsideExclusions
Maybe the 318 works in your way because you keep it simple.

I'll try it, too. My Muh crashed down 1,5m into a cellar shaft with 318 1m outside of the perimeter where it never had problems. And I don't know, why it was in this region.
 
They are the same function but with different names, I just disabled one of them.

I dont know if its the same, but sometimes my ardumower drives away until it goes into kidnapped mode, its rare and I have no idea why it happens.
Sunray works in mysterious ways.
 
I’m just getting my Ardumower ready for it’s fourth season and I am wondering if anyone tested v319 that Alexander patched? I presently have v308 which worked Ok except for the occasional wander which I now see was likely caused by the spike detector which I can comment out. There seems to be much confusion at the moment regarding what works and what doesn’t so I would appreciate any comment regarding loading v319. Rgds
 
if anyone tested v319 that Alexander patched?
I did a short test and for me this is still unusable.
The rotation at the end of the mowing path looks way better but there is still the problem that the roboter gets from time to time a stroke and just dancing around. I also noticed that the new version triggered random obstacle while driving straight, not sure whats going on here.

But was just a short test, reverted back to 317 with the fixes and everything worked fine again.
 
I dont know if its the same, but sometimes my ardumower drives away until it goes into kidnapped mode, its rare and I have no idea why it happens.
Sunray works in mysterious ways.
Yes, it seems very mystic. I observed in my app that there were generated some obstacle detections in an area where noting is that can be detected as obstacles. Therefore I decided to go to the mower. As I came the mower already crashed - but at another side.
In the app I had the obstacles fresh generating at the left side of the perimeter. But it crashed round about 10m away at the lower perimeter side.

I think the mower thought it would be at the left side (where it had to drive round about 2m to lower side) but it was at the lower side round about 10m right from this point and tried to drive to lower side where an obstacle is.

I can imagine that sometimes the mower has a fix but operates with wrong positioning coordinates.
 
I did a short test and for me this is still unusable.
The rotation at the end of the mowing path looks way better but there is still the problem that the roboter gets from time to time a stroke and just dancing around. I also noticed that the new version triggered random obstacle while driving straight, not sure whats going on here.

But was just a short test, reverted back to 317 with the fixes and everything worked fine again.

Thank you! I have already loaded v319 but have not yet been able to test it myself due to bad weather once the weather permits I will give it a short test but if it misbehaves I will use v317. Could you summarise the patches you have applied? I’m finding it difficult reading the thread back to ascertain exactly what you have patched in v317 to make it usable.
 
- get_turn_direction_preference -> comment line 171
- USE_LINEAR_SPEED_RAMP -> disable in config.h
- SMOOTH_CURVES -> disable in config.h
- SUPER_SPIKE_ELIMINATOR -> comment line 24 in AmRobotDriver.cpp to disable this function
- EscapeReverseOp -> I would recomment to disable this, so the mower is not driving randomly backwards on obstacles
I have rain again this morning 😟 but I did manage to run for a while last night. Using v319 I mowed about 400m which has at least 60 turns and all worked without any errors or dancing this included dock & undock. So far so good!

My Ardumower is a v1.3pcb with origional gear brushed motors the GPS system uses Sparkfun RTK. Latest ESP32 code and Cassandra via HTTP.

Looking now at your modifications against the v319 code...
- get_turn_direction_preference -> comment line 171 - I think this was deleted by Alexander along with a big block of code abt two weeks ago
- USE_LINEAR_SPEED_RAMP -> disable in config.h - Turned off by default in v319 and I believe earlier
- SMOOTH_CURVES -> disable in config.h - Turned off by default in v319 and earlier
- SUPER_SPIKE_ELIMINATOR -> comment line 24 in AmRobotDriver.cpp to disable this function - Disabled by default by Alexander on 5th April revision
- EscapeReverseOp -> I would recomment to disable this, so the mower is not driving randomly backwards on obstacles - Not tested this yet

As soon as I can I will get the mower out again, test further and report back.

Regards
 
Zuletzt bearbeitet:
Oben