Overhead pinch trap

[glnc222]

Finished version of this mod is on youtube under "Neato modification" at
http://youtu.be/Vq3RXH6kzxc
or follow thread to end.
Edit: other mods for a variety of cleaning issues at http://www.robotreviews.com/chat/viewto ... 20&t=16475

The problem of travel stopped by overhead pressure from upholstery needs a firmware change, because when this happens the robot does not shut down, though maybe the wheels do, can't tell -- in a hurry to get it out before damaged. If it get's trapped it ought to be simple to detect no travel is being accomplished after several seconds and declare a navigation difficulty. Potential for unit damage would seem to give it some priority, though I guess not too many complaints. Too persistent for it's own good.

[edit] I have since revised construction to reduce possible lidar interference by using thinner 1/16" stainless steel wire, painted black with auto metal paint instead of black heat shrink cover, and a couple layers of 1/16" thin heat shrink (probably one thick type) over the insertion into the 3/32" holes into the bumper. It worked originally but after a couple years seemed to develop some interference, not clear whether the surface of the heat shrink or the thickness. Easier to work the thinner stainless (needed for stiffness compared to regular nail or screw steel).

[aubie]

If it gets stuck, it backs itself out.
ver 2.6

[glnc222]

cannot move at all once jammed in this way, no backing out.

[holmes4]

I've also had this happen where it gets wedged in and cannot back out. Doesn't happen often, though. For my old Roomba, I had applied a silicone rubber "horn" bought from an online shop that specialized in Roomba accessories. (I don't remember the name offhand). It would hit low overhangs and let the Roomba back out, since it would not get wedged. One might do the same thing on top of the Neato turret.

[glnc222]

My issue is damaging the unit, not avoiding the trap.

Roomba shuts down when trapped this way, and Neato seems like it will burn out a motor. If they can't protect the machine, how well have they done anything else?

It may not be practical to avoid this trap because you want the vacuum to work under hard surface overheads like wooden furniture with minimal clearance. They may not have any furniture pieces of the kind causing this trap and have no way to test it. That's acceptable; damaging the unit is not.

To actually avoid such traps the kind of sensor to use might be a flat keyboard switch on top of the turret, an only slightly raised button or entire top surface acting as a vertical bumper; has to be thin, is the design problem. Fitting under low furniture is an attractive feature not to be abandoned. In some cases the lower Neato front could be pinched before even getting to the higher turret. So it's hard to avoid.

But there is no justification for damaging the motors. Roomba detects excessive loads on the motors and that info might already exist in the Neato, check the command interface -- but it looks like it is not processed in the firmware. That data might already be used for some other aspect and they just didn't carry it through to this problem.

[mfortuna]

Neato probably relies on its navigation to avoid obstacles. Once stuck, if the wheels continue to spin it may not trip a load sensor. The roomba's front wheel tach will detect a case like that since it is not powered.

[glnc222]

Thanks; always wondered what that funny castor wheel sensor is for on the Roomba.

Modification prospects: the turret cover is a separate part of the top cover screwed on, tight atop a separate laser unit cover. Cheapest to relocate the red power button atop the turret, no additional parts. Replacement top covers are available for experiments.
There are definitely ultra-thin switches usable with existing casing, nano-tact-switches for cell phones and accessories, merely 3/100" thick; cheapest is robbing a throw-away brandless bluetooth thing.
Can't tie to the front bumper switch as triggers aggressive behavior compared to a shut-down needed. A different behavior program needed for top bumps, and not clear what action to take -- after pulling out, where should it go, back into the trap? So use the power button or select defaulted to Pause when running. At least locate traps.

Videos show Neato getting stuck and backing off high obstructions with persistence and delays, so it just isn't giving up when held down from the top, using the same response. Needs a timed shut off at some point and not clear there is any.

[glnc222]

David (aka Blind Guy) wrote under "Epic VR100 Video" http://www.robotreviews.com/chat/viewto ... 42#p106842

what if you fastened 2 vertical rods on each corner of the bumper the precise height of the turret?

Easily dismissed as blocking the laser, but noted a 1/8 inch thick rod blocks only 2 tenths of one per cent of the 360 degree circle at the front corners 11 inches from the turret center, amounting to features 2 inches wide at the 15 foot sensing reach. Needs paint absorbing the IR laser wavelength, or maybe the smoke plastic on the charging base front, transparent to it. Used on TV trim over IR remote receivers.

Tested Successfully With 4-40 Nylon Screws
A spot clean with 4-40 nylon screws 3/32" wide on the front corners showed no interference with the laser. Could be a useful mod here.

This method has been marketed for Roomba lacking any laser problem:
http://www.botjunkie.com/2009/12/03/rob ... es-almost/
and I believe the high-end Karcher robot incorporates a built-in vertical bumper (with a possible extra role in mating with the automatic bin cleaner accessing from the top).

Thanks for raising this possibility again; would not have tested it otherwise.

[glnc222]

Full Method Test Works

(possibility re-examined at suggestion of David (aka Blind Guy))

A Scandinavian Eurosleep sofa 60" wide, opening to a queen size bed, previously trapped Neato underneath by progressive soft pinch of bulging upholstery. The front wood panel is still well above the Neato's top.

A temporary attachment of #4 nylon screws on corners of the front bumper with taped wood stops the Neato before trapping, allowing a Spot Clean test to complete. The screw tops imbed well into the upholstery but are too small in area for friction to prevent Neato pulling out, unlike the full turret surface. These screw tops are too small for a finished version because they scrape along the upholstery and might eventually damage it.

Screws into the front bumper plastic may have to be flat head countersunk on the inside pointing up, to not interfere with the bumper edge sliding across the top cover when deflected. Has not been opened for examination of the best installation method.

[voicegy]

One could always raise the furniture.

[JEfromCanada]

Wouldn't such a mod possibly interfere with the ability of the Neato to follow the front edge of a kitchen cabinet underneath the typical overhang? Despite our overhang being lower than the turret, enough of the Neato's side gets under the overhang to be able to effectively clean to the edge of the cabinet.

[vic7767]

Trial & error could help determine the optimum location of vertical avoidance hardware.

[glnc222]

One could always raise the furniture.

I actually prefer to raise it to clean underneath with small blocks under the front edge temporarily for cleaning while out -- too high for permanent.

But annoying to discover these spots with the Neato stuck in hard to reach places refusing to stop motors.
The probes might retract into the bumper and twist locked up for temporary use, or taken out when unused.

Wouldn't such a mod possibly interfere with the ability of the Neato to follow the front edge of a kitchen cabinet

Exactly, a turret pressure sensor best, switches on supporting posts or top surface. Hard to add with unknown LCD board circuits and hard to get parts, easier for factory. Investigated in thread http://www.robotreviews.com/chat/viewto ... et#p106199

[voicegy]

Obviously the thrust of the issue is what behavior should the Neato engage in when stuck in such a position - hopefully our Neato representative can comment. As soon as the device notices its in a no-win situation, it should just stop and cry out for help. But I've not had a scenario come up like that to comment from personal experience.

I think mods to furniture itself would be more useful, such as seen in a video posted by another user elsewhere in the forum. What I want from any robotic device is as little need for hands-on as possible - under my couch, for example, early on I discovered that Neato likes to spend a LOT of time under there exploring, somewhat unnecessarily, thus vastly increasing overall house cleaning time. Now I'm just in the habit of folding up and slipping under the couch both TV trays before bedtime - Neato bumps into these and goes about its merry way. Naturally, another avoidance option is to use the handy magnetic tape for this, and/or other trouble spots. I can always move the couch for a monthly cleaning underneath with the traditional upright vacuum.

[glnc222]

electronics problem

Need to know whether button switches, LCD and bumpers, are contact switches or capacitance sensors, whether can be wired in parallel to add more. Might be able to craft thin contact switches under the turret post footings, above the cover -- washers separated by foam, conductive paint etc. No point without electronics requirements.

Stopping vs Deflecting

Front post advantage: keeps moving so can be followed to locate problem areas.
Turret sensor maybe only good for stopping, having to repeat runs. Have to test if can guide out paralleled to bumper switch, as triggered worse situation.

[glnc222]

Post Installation Convenient

[revised 4/18/12 after disassembly]
The bumper bottom cover screws are into posts to the top edge, but have narrow holes for which drills will not reach the top. But there is plenty of room between them for gluing additional spacers to serve as sockets for posts drilled through the top. Quick insertion removal good, use for reconnaissance runs locating problems.
Partially threaded hex bolt has a smooth shaft portion removable and head can be rounded. Cut head off for plugs. #6 size interesting if length available.

[Blind Guy]

Though I'm glad my idea has garnered much speculation, I think the idea of a pressure switch located on top of the turret the best solution.

David (aka Blind Guy)

[glnc222]

Turret Contact Switch Design

[corrected 4/18/12 after disassembly]
Maybe not so hard after all.
The bumper switches are contact switches, not capacitance sensors, and can be put in parallel with additional turret triggers. Bumpers are on cables to a system board header mounted horizontally at the rear, behind a longer header for other stuff. The smaller Neato case screw is used with a 3/16" wide head.
Some promising ready made switches at http://www.mpja.com/18983-SW/productinfo/18983+SW/

Small elastometric pushbutton switch is only 7.8mm Sq. X 5mm tall, excluding leads. Rated: 12VDC @ 5mA, 500 Ohm contact resistance max. 120gm activation force [4 oz]. P.C. Mount. Used in audio systems

Many similar, even smaller, on Ebay. They could be mounted with #4 solder terminals bent into brackets. Screw holes through the top cover appear would be covered by the turret posts.
Radio Shack SPST 12VDC/50mA 5.0mm High Tact Switch
Model: 275-002 | Catalog #: 275-002 $3.99
(10 cents each direct from Asia on Ebay)

Some difference in room around the rear turret screws, which are inside a cup or tray in which the drum spins, surrounding the belt, while the front screws are outside this. The top cover engages this cup with a lip, making a 1/2" deep container.

[glnc222]

Recommendation for Modification

For reconnaissance runs done only once to identify pinch problems no permanent mod is necessary; just taping on a thin wood post mount shown in the post photo above suffices.

Supplying properly shaped and sized posts to use (maybe #6 hex head screws properly rounded), or something fancier to clip these onto the front bumper ready made could be an interesting accessory. Maybe bumper cover screws and grips over the top edge would work, unknown (there is a sharp straight edge indented into the bumper top in front of the dust bin, which a very inflexible clip might hold -- but can't go around underneath where the bumper slides over the cover). Since only needed for initial runs in premises, could even be rented out! (This is one of Donald Rumsfeld's "known unknowns".)

Some users might have premises with pinch problems and not want to make any furniture adjustments even barrier strips, preferring the vacuum avoid such problems automatically. The front post method is a practical installation compared to custom crafted switches more for electronics hobbyists. Aside from that, permanent mods were analyzed to demonstrate this feature is practical without expensive electronics and suitable to request from suppliers, especially for upscale products. How hard can holes to stick in posts be, or pop up posts?

[details of the permanent mods construction posts above were edited 4/18/12 after disassembly, revising some details]

[glnc222]

cabinet overhangs

ability of the Neato to follow the front edge of a kitchen cabinet underneath the typical overhang

So

determine the optimum location

Indent the posts in line with the turret edges or put only one post in the center. Plenty room for multiple sockets for personal adjustment. Better than nothing.