Problems with upgrade to botvac from xv

[crimsondr]

Hi all,

I'm having trouble after replacing my xv with a botvac.

I have bar stools that have a circular base. The botvac always climbs the base and then errors out with a clear my path error. My previous xv had no such issues.

Another issue is I have a cliff that hits the wall at about a 30 degree angle. The botvac follows the cliff to the wall and then tries to backup and turn out of the corner and fails because of the cliff behind it. Again, previous xv unit had no problem with this scenario.

I tried to use magnetic strips on the hardwood but the side brush always pushes the strip away.

Any suggestions on how to deal with these problems?

Thanks.

[glnc222]

A very unusual floor. The mag strips have always been vulnerable to being pushed around in some cases even by XV models. It is necessary to stick them to the floor in some cases. They can be taped over, double-sided tape under, and even tacked through. In one case I tape them over a thin wood molding strip easily placed, removed, and stowed.
An alternative in some cases is molding sufficient to engage the bumper. Photographs of this unusual floor would be needed to advise more specifically.

While the XV navigated those bar stool bases there are plenty of cases where even the XV does not navigate properly, and can be trapped, by obstacles too low to engage the bumper. As shown in the mods list top of main page, it is easy to extend the XV bumper lower. The BotVac bumper could also be extended but might require screw holes voiding the warranty (the XV has lots of screws already). There could be some possibility of attaching extensions with double-stick tape or glue, I am not sure.
What you might try is simple cardboard which can bend around the sides and be fastened with tape; run it high up the bumper to provide leverage against contacts at the bottom. The BotVac is better in this regard having the wall sensor completely behind the bumper, making for a simple wrap around fitting. More detail can be provided if interested. [edit] completely cover the bumper area with permanent grade double stick tape, and regular tape over the top edge. The problem is getting the right material stiff enough yet can bend around the sides; vinyl plastic sheet is the thing but not very available. Plastics vary in their properties, acrylic for example being very brittle, but optically good for picture frames. Corrugated cardboard might work, but no telling until tried. Insure the corrugations are vertical.

[edit] When raising the bar stool only 1/4" would engage the BotVac bumper, it is also practical to insert 1/4" wood risers under the bar stool bases. I treated one furniture piece this way rather than modify the bumper. But when there are several pieces...

[crimsondr]

Here are images of the cliff and stool base.

[glnc222]

Thanks for posting these photos. If I understand the top picture correctly, not being sure what that dark area upper left is -- an opening to a lower floor? -- it looks like you have a standard stairway with just a couple steps lower right. If it is not stopping by cliff sensor at the steps a mag strip could be placed across there, a few inches back, and fastened down some way. I cannot tell what the edge of stair is, the carpet edge or is there wood beyond that? Contrast in the picture... With the carpet some thin brads pushed through the strip into the padding under the carpet might hold it in place -- it is shoved sidways, not picked up...

Why the cliff sensors would not detect a step down stairs is not clear. Usually that would be a hardware failure of the sensors. Maybe the BotVac is defective. There are tests for those sensors over the USB connection, as well as other cliffs in the house, whether those get a proper response.

The stool base looks unusual in being mirror finish, so it is hard to tell exactly but it looks like it slopes upward? I have something wood like that under a swivel recliner which the XV climbs, even with a lowered bumper. May I presume both the BotVac and the XV climb that base, but the BotVac gets confused? Does it appear stuck and unable to move, or just confused? There is a little different size and stance with the BotVac so that a borderline trapping condition which does not affect the XV might affect the BotVac. The taper of that base could cause the bot climbing not to contact the bumper on the center pedestal, if the front were lifted up at a steep angle.

It might be necessary to raise the bottom edge of the base 1/4", maybe with a black or silver painted edge to a disk of wood, combined with lowered bumper on the BotVac to engage the bumper instead of climbing the base.
More exact measurements would be needed to tell precisely, given that mirror finish.

A curved polished surface like the pedestal might not return reflections to the Lidar, reflecting the beam off to the side, similar to problems with mirrors, glass doors, and polished stone facings.

So the exact mode of sensor failure would determine what, if anything could be done. How does the XV succeed where the BotVac fails? I would have to see a video of each or a detailed description. At the minimum it might be necessary to surround the bases with mag strip material. Or put the bar stools up on the bar when cleaning.

It could be someone revised the program for handling obstacles not seen by the lidar, bumper contacts, with behavior that does not work as well as before. I may find out if a BotVac ordered for drop shipment gets delivered soon. I had to modify the XV to get it to handle some furniture here, so I am not keeping the BotVac unless I can get it to work right. They make some improvements, but it is all something of a mixed bag with these machines.

These devices do not handle every sort of furniture and flooring. If the BotVac will not work on that stuff, you might want to stick with the XV.

[glnc222]

Cliff Sensor testing without USB -- use spot clean arranged near a cliff, ready to catch in event of failure. Note how the unit moves when starting, to position usefully. I think it turns to the right immediately and goes off that way. So you put it parallel to the cliff a couple feet away and it will turn and go towards the cliff, and presumably be deflected by it if working properly. The cliff sensor windows must be clean or the dirt is sensed as a nearby floor.

[glnc222]

Possible bumper extension material;

2.25" clear plastic corner guard Lowe's, 4ft around $3.25: http://www.lowes.com/ProductDisplay?par ... &cId=PDIO1

The XV takes a corner guard for the right angle. Here cut off a straight 2.25" strip which may bend around the sides -- have not tested it, depends on type of plastic. Some are polypropylene I am not familiar with. The angle type used for the XV bent well though. You extend the bumper down only 1/4" or so and use the longer width to provide a large fastening area to the bumper surface with double stick tape. The side brush area requires study.

[crimsondr]

Thanks for all the comments so far. Below you can see a picture of the floor with the cliff. Does it help make it any clearer? The previous picture is a close up of the top left of the picture below but from the opposite direction. The botvac travels along the right wall with the door, up towards the cliff. It then travels along the cliff until it reaches the left wall. At this point it tries to backup and turn out of the corner but it fails.




The stool does slope upwards. Both the XV and Botvac climb the base and hit the center pedestal. However, the XV will navigate around it like a regular chair leg. The Botvac on the other hand will error out IMMEDIATELY after hitting the center pedestal once. I don't know why it does this. I have tried placing the magnetic strip around the base since it does stick but the Botvac will eventually push through the strip. I've tried placing the strips to make a square on the base midway between the center and outer rim, again the Botvac will eventually go over the strip. I think it has something to do with the angle since the Botvac head is raised and perhaps not picking up on the strip. I'm pretty sure there is nothing wrong with the sensor since I use strips elsewhere and it isn't a problem. I've also tried to wrap the center pedestal with magnetic strip and usually it will detect the strip and not hit it. However, again sometimes it will and as soon as it does it errors out. I may have to modify the bumper for the Botvac if I can't find any other solution.

[glnc222]

I have to wonder from this whether Neato messed up the behavior programming in the latest v3.4 release, presumably based on changes made for BotVac. Has this XV been updated to that, and still handles the problem areas?

The stairway is an old problem seen before, a basic defect in the behavior. Interesting if the XV handles it. The system is not integrating the cliff sensor into a memory or responding to that data when hitting the wall. At the wall it backs up, which it should not do near a cliff, were it programmed effectively. Neato has never been good about such details of close quarter maneuvering. You would have to leave that corner piece uncleaned with a mag strip cutting off that part so there is room to back up away from the wall. Unless the XV does it right. And don't upgrade it to v3.4 until that is verified to have good behavior. Neato Robotics has not provided well for undoing an update, except for temporary reversion to the previous version. Scattered discussion of that around this site after previous behavior changes. Customer service may have arranged something for some customers, but the details were never clear.

Oddly Neato customer service has taken to discouraging updating altogether unless there is an internal error reported by the system, very peculiar and inconsistent practice. I cannot trust them to even know what technical abilities there are in updating, reverting and so forth, with insufficient programmer contacts, out-sourcing and what not.

The behavior at the cliff and wall intersection may have been introduced with the so-called "corner cleaver" more intense maneuver near corners, in v3.1. The response hitting a wall was made more elaborate, and it is conceivable they did not take account of a case where a cliff was next to wall.

The BotVac behavior when striking the center post of the stool is weird. That part of the behavior should not have changed. If all BotVac's do that, there is a serious problem in the behavior. I have to wonder whether it will fail with any obstacle below the lidar striking the bumper, assuming the mirror finish post is invisible to Lidar. Lidar visibility does not always affect such bumper contacts, as the short range mapping has never seemed to be developed. At close quarters it seems to rely entirely on the bumper and temporary obstacle strategies. That could be checked over USB positioning the bot in front and reading out the Lidar scan. You have to wonder why it works anywhere in the premises with this particular failure. I would want to check another BotVac in the situation.

Very interesting report revealing possible software design problems in the BotVac and possibly in the latest release for XV's. The older XV's with old cpu's cannot be updated beyond v3.1.

[glnc222]

The reference to the BotVac failing to respond to the mag strips around the stool base suggests that BotVac does not function properly at all. I have to wonder how it manages to do any of the premises under those circumstances.

[crimsondr]

My old XV was on 3.1 (I don't have it anymore) and never had a problem with that cliff. It would turn, back up a bit and then turn again where the right side of the bumper would git the wall again and then continue out of the corner. The botvac does the same maneuver but errors out when it bumps the wall the second time.

The botvac seems to respond in all other situations properly to the bumper and mag strips. So I assumed it erroring out when hitting the stool center bar was related to it running up the base and being on an angle during the impact. Also, regarding the mag strips, it does see the mag strip and then backs up a bit to turn but as it wiggles around, one corner passes the mag strip and then it continues up the base of the stool. I thought that this may have something to do with the mag strips sticking to the base. Could it diffuse the power of the mag strip and confuse the botvac?

[glnc222]

Using mag strips around such small items as those stool bases is just a technical matter for analyzing BotVac problems; I would think it not very appealing decor wise etc. anyway. The strips have to be off the base, on the floor. I am not sure what effect the steel might have on the field, it could draw it all into the iron. But the sensors are on the sides of bot so if the front climbs up the base the sensors will get lifted off the floor and be ineffective -- the base is too small.

There has always been some weakness in the bumper contact and mag strip response, in that the bot is very aggressive and probes for any opportunity to slip past. Well, that goes with effectively bypassing isolated obstacles, so quite a puzzle to the programmers. When extending the bumper lower, the sides must be done because it tries to side swipe over things it cannot climb head on, very clever -- or too clever...

You could check whether the bumper gets actuated hitting the stool post by manually shoving the bot up on the base and checking the audible click on the bumper tact switches (audible on XV's anyway); the data can also be read over USB. It is not clear that is the problem. The thinner BotVac bumper and mechanism with plastic tabs instead of springs on the tact switches, plus maybe different clearance with side brush added, might make a borderline case between the XV and the BotVac. The worst case is the BotVac software does not handle contacts well. Unless you have just the right obstacles not seen by lidar, such as thin black metal chair legs, or something too low, this might be the only instance in that environment. I arrange little tests for such things with sticks and such, spot cleaning. BotVac arriving in a week; stock was adequate.

If the bumper is not actuating on the center post lowering the bumper edge might help. This can be tested with temporary fitting of a ruler (vertical, sticking up, a little extension down) or something just for manually testing the geometry as above.

The only solution might be preventing it climbing the bases by getting bumper contact around the edge.
Square blocks (without fancy woodworking to cut disks) can be fastened underneath raising the edges slightly and the BotVac bumber can be lowered -- not always necessary, depending on the exact dimensions involved, aesthetic considerations, etc.

[glnc222]

On the stair you might need to block off with a mag strip only a small corner against the wall, just to adjust for the different backing motion of the XV and BotVac, get just enough room to accommodate the motions. May need to be tacked down. The cliff sensors may not be tracking the edge close to the wall, if it is rolling along straight into it. The software then has no data on the entire configuration. It has to swerve over the edge to see it. All a borderline effect of the different BotVac dimensions and parameters for motion. It is sheer luck the XV handles it.

[crimsondr]

Last night I put a square box in the corner flush with the cliff. So it makes a 90 degree wall against the cliff. Tested a couple times with spot clean and it actually errored out with a clear my path error once.

[glnc222]

I think "clear my path" comes from obstacles preventing motion which are not detected by Lidar or contact sensors, locking the wheels -- suggesting intermittent lidar problems. Especially if the wheels are slipping, so it does not detect obstacles that way (the only way for obstacles in the rear lacking a bumper). With no sensing it cannot tell which way is out. It is usually persistent and tries numerous times at different angles and what not to overcome obstacles. Is it trying a proper "corner cleaver", backing up etc., at the box/wall intersection? That should be no different than any other corner on the floor.
[edit] "corner cleaver" engages only when the walls are detected by lidar, not with mag strips or barriers below the lidar -- on older XV's anyway. One thing to see if improved.

[crimsondr]

Isn't clever corner for outside corners? In this case it's an inner corner.

I may try another Botvac and see if I encounter the same issues.

[glnc222]

Both inside and outside corners were modified by "corner cleaver" firmware addition, though the impetus was the inside corner, going against the wall and backing up to turn, taking more time to do it right.

Try the same box obstacle against another wall. Is the box high enough and visible to lidar and the wall sensor?

Getting "clear my path" with no visible obstacle can be considered a malfunction justifying replacement of unit, if it is under warranty. Nothing like an LCD msg to certify a defect. It has been associated with intermittent lidar failures, the drum spinning motor stopping, worn commutator brushes or what not. The USB GetLifeStatLog might show something about what is happening at the corner.

[glnc222]

Simple cardboard solution to lowering the BotVac bumper is now shown http://www.robotreviews.com/chat/viewto ... 68#p127168
[edit] Plastic version also from Rubbermaid tray material.
in case of any use with those upward curving stool bases -- if replacing the unit did not improve.

[crimsondr]

Finally got a new BotVac. The new unit no longer errors out with a clear my path error when cleaning the corner/cliff scenario.

Here are videos of the old and new units navigating the cliff/corner scenario.

Old unit: http://youtu.be/DNgCCO4SXrE
New unit: http://youtu.be/JBU9KdtEdAM

[glnc222]

Thanks for reporting back. Any luck with the bar stools?

[crimsondr]

Problem with the bar stools is also resolved. It still tries to climb the stool base maybe a dozen times in a back and forth motion as if it's stuck in an infinite loop but it does stop eventually and continue.