Printing experience: settings, suggestions, etc.

[romualdas]

Ok, I guess it is time for some follow up.
I have mostly followed updated design by Paddy.

Some details on configuration as it is right now:
- printed wheels with outer rim from PETG
- 3 ultrasonic sensors
- 2 mower motors
- bumper based on outer shell using micro switches (as designed by Paddy)
- GPS (moved to the back, into the motor compartment)
- outer shell attached using custom made dampers (as suggested by Paddy)

At the moment I see couple of serious problems:

- If the mower comes to perimeter at a very shallow angle near the fence, after reaching it, it may try to turn direction towards the fence and hit it. The force may be not enough to trigger the bumper and the mower starts to dig the earth (because of the solid outer shell and quite large spikes on wheels)
- because of the bottom part of the nose in sloped inwards, there might be cases, where the mower comes to a tree or small fence (like for flowers) and manages to go nose up without triggering bumper. This acrobatic eventually broke one (at least) damper

At this point, I'm thinking to fix the damper (build new one) and then increase sensitivity of the bumper by bending switch pins just a bit forward.

There are some other things that annoy me a bit:
- mower disk covers are indeed quite low and while normally it is not an issue, but in some places it just touches ground enough to change direction and/or stop. i had to even out some spots on the perimeter for that. I guess the option is of course to cut a centimeter or so from the covers so the have higher clearance.
- Odometry is somehow not working properly. Need to figure out proper PID parameters for this setup. Can live without it, but movement is somewhat unpredictable in such case
- Maintenance is somewhat complicated due to the wiring that is necessary to have between PCB box and shell.

So far, had 3 runs - none successful, one that would run automatically from start to finish without any assistance.

P.S.
I think, I have missed Pfod related changes in the last commit related to two mowers. The necessary change is here .

 

[paddy]

Hi,

it covers some issues I also found.

- digging into ground at perimeter (never seen this with old chassis and always wonder what happens here). I found that this occurs if state is POUTFORW. I didn't check code right now but I thing mower ignores bumper in this state (there are lot of other states where bumper gets ignored)
- never seen climbing up on trees, bushes. My bumper always triggers, maybe you need to adjust your bumper sentivity.
- no broken Damper so far. It is important to avoid any air inside silicone during mold.
- disk cover is way too close to ground. Mine triggers also often. I used a sharp blade which I heated up to cut 1cm off.
- in general, disc cover needs rework, it is not possible to keep grass higher than 5cm.
- Odometry works good for me with Azurit default PID settings

In total, my mower has approx. 20h mowing time. Sometimes it runs several hours without issue, sometimes it gets stuck every 5 min (PERI OUT Timeout, Stuck). But this was also true with my old chassis.

 

[romualdas]

Well, in couple of cases mine hit the bumper while tracking and it mowed from the perimeter to the right, so I'd say that it uses the bumper while tracking.
On the other hand, when the odometry is ON, sometimes it behaves strangely on the perimeter, e.g. doesn't drive back enough (or at all), etc.
Also, last year I had quite a few issues regarding registering perimeter. It appears that the spool is easy to damage (unroll couple of strands in the bottom) and it is very sensitive to any damage.

 

[paddy]

I'm not talking about tracking. During tracking, it reacts on bumper. But if mower leaves perimeter wire, it starts to reverse and roll. If mower don't reach inner perimeter area, sometimes it tries to move forward (State POUTFORW). Then it ignores bumper. I'm not sure if it always ignore bumper when left perimeter area, but POUTFORW ignores it.

I'm curious about how often my new mower gets off lawn and stuck in perimeter timeout. Yesterday I found it approx 2m off lawn on my terasse. This never happend with old mower (where I never get odometry to work). This may cause dangerous situations like driving on road or fall into water. I need to dig deeper to clarify if it is caused by different odometry behavior and I wonder why I wasn't aware of this before.

 

[romualdas]

Ok, I see your point now regarding POUTFORW. Indeed, that would explain some behaviour.
I am wondering if smaller but heavier (in my case) wheels have significant impact on Odometry too.

 

[bernard]

Hi Paddy.

I'm curious about how often my new mower gets off lawn and stuck in perimeter timeout. Yesterday I found it approx 2m off lawn on my terasse. This never happend with old mower

Maybe you can check the noise on the coil, With the new chassis it's possible that it's Something new that you never see before.
I have big difference between the Denna and Paranello chassis.

By

 

[romualdas]

In my case, signal quality is actually better on 3d printed chassis than experimental one (it never had lost signal in the middle of the lawn, which was happening with the previous one).
Also, I've noted that after mower crosses the perimeter *sometimes* it doesn't drive back, but then turns and moves forward.
This does not happen, if I turn off the odometry. Mower then always tries to pull back, then turns and so one.
It has never stuck on the perimeter with odometry off. On one hand, it had only few hours here and there with odometry off, but not many more with odometry - on (and it stuck several times).

 

[romualdas]

Here are some details on dampers.

First a bit on the dampers that are sitting on the motor block inside these wells. Even though they are below the cover and the gap between cover and the motor block is not big, grass clippings not only get there, they get into the bottom part somehow.

Note that in the image is shown the bottom part. Grass got even there ... hmy:

The dampers that broke in my case, did so at the place where the plastic insert part sits. I don't know if it due to not sticking to PETG, the air bubles (I tried to get rid of them as much as I can) or the fact that the plastic part basically reaches (is visible from) outside. Basically, one can feel that if you bend it, two sides start to tear appart in that place.

Also, the damper that is ok, originally had faults from mold (material got too hard to fill mold completely), so I covered with fresh layers by hand from outside. In that case, the plastic insert is not visible from outside anymore and the part feels a bit sturdier.

At the moment, I'm in progress of making new dampers. And even they look ok, I'll still cover them with extra layers of rubber manually. Will see how it goes.
Attachment: https://forum.ardumower.de/data/media/kunena/attachments/4472/IMG_20190417_001252.jpg/

 

[paddy]

Regarging Perimeter, it is exactly what I discovered too. It seems there is some rework necessary in Azurit so it also pulls back when odometry is on. I wonder why this happens only sometimes. In even most cases, it first pull back and rotate after.

I'm really surprised that you find gras under the damper. Never expected it here :-D
Silicone doesn't stick on PLA or PETG or any other plastic. That is the reason you're able to unmold it afterwards And thats the reason why the grey parts have this holes around. Which kind of silicone did you use? It is also important to keep mixing rate of components as described on package.

The shape of dampers might not be the best. It seems also to work if you design in totally round (complete cylinder shape). I was worried about screwing the dampers without be able to hold inner nuts in place. Thats the reason for this shape and for the grey part touching the outher shell. This way, you can hold the inner nuts in place while screwing everything together.
But if you don't overtight them, they should stay on place even when you can't hold them during screw together.

I hope you understand what I mean.

 

[romualdas]

I have used the "TFC Troll Factory Silikon Kautschuk Typ 3" (search on ebay). Mixing ratio is 1:1.
I've made new dampers and applied some silicone on top by hand. Damper feel a bit stronger.
I guess, if the plastic parts wouln't touch the sides of the mold and would have some vertical walls from center to side to prevent from rotating inside - it would be better, no?
It could be a bit harder to insert into the mold, but otherwise - might work a bit better.

In addition to that, I've cut ~1cm of the blade protection to have higher clearance from the ground.

I don't have much grass here yet, so I plan do to some more test runs during the weekend.

 

[romualdas]

Some updates here.
I think I've dealt with most of the issues regarding mower sticking on the perimeter and/or exiting perimeter. To do that I did the following:
- completely clear user settings (via USB cable)
- decrease the US sensor range to 18cm as in my case they either see the obstacle too late anyway or ghosts are triggering drive back. So, in order to reduce the latter, I've decided to reduce distance for the US sensors below 20 cm.
- increased "perimeterTriggerTimeout" (12 sec now). I've noted that if the mower gets stuck while making turn outside the perimeter, it may stay as the timeout gets reached.
So, in terms of handling perimeter, I'm more or less fine.

There are 3 outstanding issues in my case (may not be an issue in your case):
- because of a nose angle, the mower tends to climb (without triggering any bumper sensors) any obstacles if the obstacle angle is at a shallower angle than straight up. This applies to the bigger trees, small decorative border fences, etc. In most cases it just ends up hanging there
- It looks like that even after cutting 1-2 cm from blade protector, with 3 wheel config it can still get stuck if the terrain is somewhat uneven

Finally, not related to this particular build (I had similar issue with previous build that used "experimental" chassis) - somehow tension element (the one that is used to attach wheel to the shaft) scrapes of the shaft over the time the wheel falls off. I don't know if I do something wrong there, but in both cases it happened on one wheel.
This time I found mower like this:

It drove over the little fence, hung there, but somehow wheel got dismounted. Looking at the motor shaft, it looks like material was scraped off by the tension element.

After that fitting the wheel to the shaft is nearly impossible as the wheel is loose even after tensioning. I've tried drill the hole through the shaft and tension element and put M3/M4 screw there, but looks like the torque is enough to cut the bolt.
Don't know if there's a reasonable way to fix this or why this happens at all. Another wheel is fine so far.
As for now, I'm waiting for replacement motors.
Any thoughts and ideas on this are more than welcome as this becomes a bit too expensive
Attachment: https://forum.ardumower.de/data/media/kunena/attachments/4472/IMG_20190521_22062601.49.34.jpg/

 

[paddy]

Hi,

I also had to increase Perimeter Timeout. I observed strange behavior when using odometry. I've found my mower twice on the road, off the gras with Perimeter Timeout. This only occurs if odometry is enabled. Without odometry, it never happens. I thing this is caused by different behavior in code if odometry is activated or not.
For US, I never had issues but I just slow down when ostacles are detected and touch them with bumper. This works really well.

Climbing was not an issue so far but I added a micro switch to front wheel. This way I can detect of mower gets lifted or if it climbs up. I just added this switch to the bumper connection.

For me one big issue is noise. The wheel motors are really loud and noisy. This comes from gearing, do you know if it is possible to add some grease here? Also Mow motor is loud too. I changed the motor bracket and added some rubber dampers. Also I used a rasp / file to get the disk perfeclty round. But as soon as I attach blades, it makes noise.

Also, mow motor often switches off. Overload counter doesn't count up so this might be an heat issue with motor driver. I will replace heat sinks by larger ones to see.

I never discovered the issue with tension element. This looks awful. It seems that the motor shaft has an D- shape? Can't remeber but if it has a D-shape, I would try to print an machting eleemt with an D-shaped hole. This way it can't slip but you need a small screw or an hole throught the shaft to avoid loosing the wheel.

But to be onest, I think it is a bad idea to attach the wheel to motor the way we did. Someone talked about Raindancer shassis at getting lots of wear in motor gearbox. It would be even better to use a separate shaft with bearing blocks and attach motors either by couplers or by belt/gears. I plan to build new motor boxes for next season

 

[roland]

The tension element has to be put on the shaft and shifted until the motor. And then you have to tighten the 4 screws in clockwise in many rounds until the are fixed. I mean tighten the upper screw. Then the Richt one, then the lower one then the right one and begin with the upper screw again.

Paddy :I think like you. This kind could be a better solution but in my opinion you need metal gears.
https://www.ardumower.de/index.php/de/forum/chassis-technische-loesungen/1908-neukonstruktion-2-raeder-kompakt#17411

 

[jussip]

Tension elements: I had to use filling under the "ring". Soda/beer can strip works fine for that. Wedged parts has slits, make sure those are not towards each other nor towards shaft d-slot.
Overload: I measured motor currents with multimeter on all motors when adjusting current sense factors. I noticed that readings are not reliable on whole range without changing code. (we need at least two point calibration..)
I adjusted current factors so that reading is reliable near over current point. For drive motors my factors are 3.15 and 3.45. Mowing motors 1.87

 

[romualdas]

Interesting, I don't use the rings at all and I think somewhere I saw post that they are not necessary (as they are too big).
On the other hand, the tension element hold everywhere, but that one wheel. I guess the part alignment is more important than I thought.

I'll have to do the repair (change the motor) soon, so will let you know.

B.t.w., one drawback of the 3d printed chassis is that maintenance is something from the hell
E.g. to replace the motor, I'll need to:
- unscrew the box that holds emergency button
- remove the emergency button
- unscrew and lift the top shell (will still be held by US sensor wiring attached to the main chassis)
- unscrew the motor box
- ... change motor (some more screws there)
- assemble everything back