Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Bug Software

iRobot Says It'll Be a Few Weeks Until It Can Clean Up Its Latest Roomba Software Update Mess (theverge.com) 58

iRobot, maker of the robotic Roomba vacuums, has confirmed that a software update has been causing issues for some users of its i7 and s9 robots and that it's working on another one to prevent future issues. The catch? It might be a bit before things get sorted out, with iRobot expecting the update to roll out "over the next several weeks." From a report: According to users on Reddit and Twitter, the recent 3.12.8 firmware update has been causing navigation issues. One user described their robot cleaner as acting "drunk" after the update: spinning itself around and bumping into furniture, cleaning in strange patterns, getting stuck in an empty area, and not being able to make it home to the dock. What's more, some other users are reporting that the environment maps their Roombas made were wiped out by the update.
This discussion has been archived. No new comments can be posted.

iRobot Says It'll Be a Few Weeks Until It Can Clean Up Its Latest Roomba Software Update Mess

Comments Filter:
  • by lobiusmoop ( 305328 ) on Thursday February 25, 2021 @10:05AM (#61098760) Homepage

    There's a scene in 'Idiocracy' where a Roomba-like device in a hospital constantly drives into a wall and repeats the phrase "your floor is now clean!"

  • Internet of things (Score:5, Insightful)

    by Tony Isaac ( 1301187 ) on Thursday February 25, 2021 @10:11AM (#61098780) Homepage

    Makes you want to hook up your refrigerator to the internet, doesn't it! At least with Roomba, you just end up with a dirty floor, not food poisoning.

  • A better update would be to reprogram them to chase pets around. YouTube would explode....
  • Why isn't there a lawnmower version of these things yet?

  • We have a constant stream of updates raining down on us as our operating sytems, browsers, internet of things devices all constantly want to update. Then an update goes wrong and mass corrupt lots of devices. One day an update will get someone killed.
    • Like how? A Roomba is going to suck one's toes to death?

      • by Hadlock ( 143607 )

        Roomba wanders into wrong room, pushes ventilator or oxygen machine cable from wall, runs over surge protector turns off medical device etc
         
        My roomba has been acting wonky too, but the worst it can do is accidentally wander into the nursery on the wrong side of the house, and break my baby's fingers

      • e.g.: Some brand of car receives an OTA update to its self-driving capability and subsequently starts steering towards oncoming headlights instead of staying in its own lane. It's inevitable.
  • the bot stops repeatedly running into the table leg.

  • by fennec ( 936844 ) on Thursday February 25, 2021 @10:36AM (#61098864)
    Hard hard is it to publish a the previous version again?
    • by sbjornda ( 199447 ) <sbjornda&hotmail,com> on Thursday February 25, 2021 @11:05AM (#61098926)

      How hard is it to publish a the previous version again?

      I've used an i7+ since late 2019. I've had to call their Customer Support a few times. The front line support staff try so hard to be helpful and they really do their best, but I get the impression that, behind the scenes, this is a company that loves its creative Computer Science people but is not so much in love with I.T. production best practices to provide a reliable tool for their customers. I don't think they take an engineering approach to their software. I doubt their software designers are actually using the tool in their own homes, with chairs that aren't in the same spot every day, with doors that are sometimes closed and sometimes open, with Amazon boxes sometimes piled in a corner, with a big dog that doesn't move until it's been bumped into a few times and then suddenly isn't in the same spot any more, etc. I expect if they tried a software rollback their "clever" toddler-level AI would throw a tantrum. I doubt they architected their solution end-to-end to take that into account; I suspect their "architecture" has become a Frankenstein's monster over time. I have no inside knowledge of the company, but from my (customer) point of view this looks like a company that is trying to be too clever for its own good and is not prioritizing making a tool that is dependable and reliable from a homeowner's point of view.

      --
      .nosig

      • by MrLogic17 ( 233498 ) on Thursday February 25, 2021 @11:29AM (#61099044) Journal

        Ditto.

        Christmas was a real problem. Areas of the house have stuff in them (tree, decorations, stuff moved to make space for tree, etc) - and when we want back to normal, the Roomba just refused to clean the newly open areas. Even picking it up and starting in the neglected spaces didn't help.

        Had to wipe the map and start over. The experience also made me wonder if the engineers actually used the thing in their houses.

        • I've had luck creating virtual zones and having the robot clean those specifically once or twice. I then have a "whole house" zone that encompasses the other zones and it all works good for me.
        • by antdude ( 79039 )

          Companies really need to eat their own cookings on production. My former employer did this wirth their own products and services. At least, they get to see their own issues and can fix inhouse.

      • I've used an i7+ since late 2019. {...} this is a company that loves its creative Computer Science people {...} but from my (customer) point of view this looks like a company that is trying to be too clever for its own good and is not prioritizing making a tool that is dependable and reliable from a homeowner's point of view.

        I've been using a 780 for roughly a decade (and a Scooba 450 for half that time) - these things are build very nicely, and are laughably easy to service and maintain, which is part of the problem.

        And indeed I have largely the impression that they have been putting too much CS (mostly in the field of AI) in their product against their own good.

        To me it looks like the old "random dance" approach to cleaning - i.e.: randomly driving around and bumping into things until eventually they will end up having statis

        • Exercise: If a house is 2000ft large, and a roomba is 1ft large, how many times does the house need to be cleaned with a random walk till every sq ft has been cleaned?

          • Exercise: If a house is 2000ft large, and a roomba is 1ft large, how many times does the house need to be cleaned with a random walk till every sq ft has been cleaned?

            Answer: Who cares; it's a mindless robot and has plenty of time on its hands (wheels?).

          • Exercise: If a house is 2000ft large, and a roomba is 1ft large, how many times does the house need to be cleaned with a random walk till every sq ft has been cleaned?

            Model the house as a circle with 2000 sq ft area. Its radius is thus about 25.23 feet. A 2D random walk with 1 sq ft steps takes about (4/pi)*n^2*(log n)^2 steps to cover the room with radius n ft, or about 8447 steps when n=25.23. So it is about 8447/2000 = 4.2 times less efficient than doing the whole room without repeating a spot. See

            • Oh, cool reference. Thanks!
              I was more thinking of random jumping around, so first square takes 1 step. Then chance to hit non visited on next step is 1999/2000, and once that happens, chance to visit another non-visited 1998/2000 and so on, giving 2000*sum_{i from 1 to 2000) (i/2000), or about 8 times the whole house.

              But isn't the result of the paper that to cover a torus of size n x n would take 4 n^2 log(n)^2, which if n^2=2000 gives log(2000)^2/pi or about 18?

              Strange that my random jumps gives about log(

    • Apparently, yes, it is hard for them. Based on my experience, and that of others with this issue, they're currently only rolling back customers that call customer service. And you can only get in the rollback queue once you get the second-tier tech support involved. After that, it sounds like it's a manual process to revert each robot. From the first call, to the eventual rollback, it's likely to take a week or three. And that doesn't even address the users that got the downgrade... only to push back t
      • > And you can only get in the rollback queue once you get the second-tier tech support involved. After that, it sounds like it's a manual process to revert each robot

        Why on God's green Earth would they inflict this much support cost on themselves when rolling out a revert would ALSO make their customers happy?

        Are they covering up an intrusion/attack?

    • by Nahor ( 41537 )

      Rollbacks are often forbidden for security/privacy reason. If one can rollback to an older version, a hacker could force a rollback (e.g. via a security hole on a remote server) to a less secure version of the firmware (and thus gain access to your home).

      (and sometimes this is used by companies that don't want to support old versions for too long, and PhB who plan to install unpopular "features", to prevent user from reverting to a "better" version)

    • I was thinking this same thing. I work in tech and if something screwed up this big we'd immediately roll back the release and kick it back to engineering...

      Now that all said, this now explains why my map got corrupted on my device and it kept saying my devices map and the map on my app didnt match. And there was no way to resync them. The ONLY option was to factory reset it and then set it about on mapping runs for the next three days to relearn my house. I consider not being able to resync the map

    • They've apparently never heard of version control...
  • if it finds a cigarette with a tiny bottle in it
  • And, why would you expect otherwise when you have an "auto-update" feature turned on? Come on, hasn't this happened regularly over the years? Company pushes out a doodad with the expectation that firmware updates would roll out "as needed" (AKA whenever we need to spy or roger you on something). Eventually, someone at home base abandons the idea of robust testing -- and creates a lot of bricks. I do not understand why anyone would set up a doodad without immediately turning off all updates -- with the exc
  • ... one, they make one that can do stairs

    ... two, they can somehow make one that does not attempt to clean up after sick pets.

    You do not want cat vomit spread all over your carpet, or worse, diarrhea. I've heard stories, and they make me a bit leery of automated vacuums in a household with pets, because you never know when something like that could happen unexpectedly. Ideally when such anomalies are detected, the device stops attempting to clean further, returns to base, and sends a text message to


    • Ok I get your points. first one. I kind of agree, but there's way too many reasons for them to never make a roomba that does stairs.
      • * It could trip a person
      • * It would require multi-floored owners to buy LESS of them
      • * They would soon evolve into Daleks and that would be creepy (and bad)

        Second Point: I have not one, but two dogs, and one is an *almost* fully housebroken puppy (3 month old Vizsla). We don't run the Roomba when home, or sleeping, because it's loud. not jet engine loud, but not something I'd
  • by pavon ( 30274 ) on Thursday February 25, 2021 @01:24PM (#61099502)

    Meanwhile, the dumb Roomba that I've had for 15 years is still working fine (other than having to replace the battery once). Some of the more intelligent mapping/navigation features in the newer models look interesting, but requiring internet connectivity makes them a hard pass for me. Between having them all stop working just because AWS is down, to crummy updates, and all the privacy concerns - what exactly is this connectivity buying us anyway?

  • first world problem.
  • is better at testing updates.

  • What world have we built, where even a vacuum cleaner requires softwares updates. (and fail at it) Is there a counter somewhere evaluating hours / years spent on such updates, adding no function, fixing eventual bugs ... and driving people mad by FEAR of the risks of not being up to date. Don't fix what ain't broken.

"Experience has proved that some people indeed know everything." -- Russell Baker

Working...