powerbot turbo 9350 (quick review)

[noslenwerdna]

As a quick review, the robot itself is pretty awesome and that's coming from a Roomba owner who has had four different roombas starting back when they were really lame all the way up to my last 650 model. I have had a ton of fun with the roombas but I decided that I wanted a robot that did a better job at some kind of cleaning pattern. I decided to go with the 9350 because of the "select and go" feature and have been sorely disappointed in how poorly it performs but it's really just that the app is garbage and should be turned over to the open source community as we'd crush it. Specific changes that should be made in the mapping is that the robot should only run the perimeter of each room and not vacuum the house as it maps. The second thing that needs to happen is that they need to incorporate user input into the mapping so that rooms can be broken up especially with open floorplan houses. This could also be accomplished by allowing the user to have an edit feature. They should have a manual mapping mode that lets the user drive the thing around with the remote to map out areas that they want to get into or to avoid. One of the features that really lacs on this robot is its inability to determine if it is sensing a cliff (or stair) or if it's just going over a dark pattern on a floor. This robot hates running across my area rug in my living room because it thinks that the dark patterns are cliffs and it tries to get around them, trapping itself like a cat in a tape square (try it, it works). And speaking of cliffs, it does not handle them well at all. Sure, it finds them and stops but then it has the tendency to turn around and back up off of them if they are in a door jamb where it might run into the jamb as it avoids the cliff. Then it backs off the cliff to avoid the door jamb... seriously? WTF? All of these problems are simple to fix and it seems that the real problem with this little robot is that they were trying too hard to get it out and the firmware is still in beta 0.0.53 after my last OTA update. I am sure that if they ever release the 1.0.0 firmware, it will be a rocking little robot but their customer communication is horrendous. I would not recommend that anybody buy a samsung robot with the way they go about things. However, if you are a hacker, this might be the exact little robot for you. I am keeping it since the only problems that I can find with it are in the app and firmware, it won't be long before I have my own system running on it and it will at that point out perform all of the other robot vacuums that I have had. Samsung, you neeeeeeed to put the 9350 on github and let us fix it for you. Then you'd have, by far, the best robot vacuum of them all but I doubt your ability to deliver this on your own.

[glnc222]

To operate on dark colored floor all brands of robots need to have the cliff sensors disabled because the IR Proximity detectors used do not work on some floor material. Covering the sensors with light colored paper or tape creates a false floor to disable them. A boundary marker either optical virtual walls, magnetic tape or physical barrier has to be used for any stairway cliffs etc. Users have had success with this sort of fix for the problem however inconvenient.

I am experimenting with a different optical method of using IR beams employing angles or parallax instead of intensity of the reflection off the floor, independent of intensity, which would work with very dim reflections if enough sensitivity can be obtained with simple parts. A lot of black materials still reflect enough in IR for the extant robots to work, so it is hard to get the particular materials which have been found to be a problem. Some rugs at IKEA have been mentioned, but the nearest store is hours away.

Some users of all the brands of robots have reported some problem with dark floors similar to that mentioned with the Powerbot.

I am just testing the principle, not looking at making replacement sensors to modify particular robots. Parts arriving today, 20 degree IR emitters and receivers 20 and 50 degrees. Mount on a vertically mounted sensor circuit board at 45 degree angle an inch apart, with space between available for a mag sensor possibly sharing components, combining into a single barrier signal for economy.

[noslenwerdna]

See, that's an awesome solution to a common problem. The thing that I think should be done with the powerbot is to have "safe zones" that could be mapped where the cliff sensors are turned off. This should be easy to do, but they'd have to incorporate a way to edit the floorplan map on their app (which is junk). Once again, that function should be part of the mapping of the "select and go" feature. The other way that they could easily do it is to assume that the interior of a room at any distance away from its detected perimeter should be safe. Either of these would work with the powerbot since it uses a camera to map the room but they are still in beta with their firmware so the robot can't do it (yet) and their app is lacking in every possible way but especially where the floorplan is concerned.

But that being said, your idea for a parallax sensor is a far better solution for cliff detectors and something that should have been better thought through by all of the manufacturers. Even a sonar system would seem to be a better way of doing a cliff sensor than an infra-red detector that can't see certain, especially dark, colors.

Cheers!

[glnc222]

Sonar would be best but I expect too expensive or also high powered, take away run time from the limited battery capacity, constrained by space available. I don't have detailed data.

I have long imagined some "virtual" boundaries could be marked on a graphics screen to set up robots, but it has not happened. A lot of robots now show the travel path on smartphone screens, so I wonder why they cannot make some use of it. Accuracy and reliability could be an issue (critical when the product could be damaged falling); these are not professional surveying instruments. There are reports of the ambitious "select a room" feature not always working well -- more reports needed. Or it is just the slow dribbling out of improvements for marketing.

Also automatic, setup free cliff sensing is needed instead of relying on a user to setup properly -- product liability or returns.

I recall some Asian bot, maybe LG I forget, with a feature where you would manually guide it around the house and mark avoidance areas with the remote, and then it would conform to those instructions cleaning. Problem with some of those bots is they have little cleaning power despite a lot of expensive sensors. First things first...