[resonanttoe]

Absolutely agree. The lack of LiDAR (or equiv) seems to really hamper iRobot's effectiveness - I had an iRobot i7 and its mapping/pathfinding was ramming in to things at near top speed and ricocheting off the walls.

They also still sell vacuum and mopping* robots as seperate units which just drives up the cost. My old setup of the two robots took somewhere on the order of 5-6 hours to do a ~100m2 apartment.

I moved to a Roborock and then Dreame bot (someone really needs to talk to Xiaomi about all the sub-brands :P) comparitively, they both took around 1.5 to just under 2 hours to do a much better job.

I actually felt huge guilt selling my two iRobot's to someone, wanting to tell them they were complete junk and they were wasting their money.

*There is always mixed opinions on mopping effectiveness, but the iRobot roll-and-spit model of mopping is truly horrendous.

[ixs]

Based on the comments here I did a quick look at Amazon and was surprised to not even see a single vacuum only robot from Evovacs. At least not prominently showcased.

Why are there only mixed models and what are the benefits?

If I look at my place the ground floor is tiled and the top floor is all carpet.

That means I would need two vacuums anyway as they haven’t learned yet to climb stairs.

Why wouldn’t I want a vacuum only model for upstairs? For downstairs the universal model is great obviously. But upstairs?

[resonanttoe]

I know there were a lot of simple vac only bots in the roborock line up. S5 and similar the mopping funciton is a small pad that you manually wet and it drags along behind. It kind of sucks but if you were in the market for just a vacuum, you remove the bracket and it's fine. The robot has no knowledge of the mop as there isn't a tank.

The S6Max that has a tank, you can also remove (easily, two button clips) and it'll not attempt any mopping feature. It also supposedly doesn't squirt water on to the pad when it knows its on carpet, but at some stage you may be dragging a mildly damp cloth on to the carpet. (was never an issue for me, the pad was never that wet that you could reasonably detect it on the carpet)

Overall mopping with drag-behind pads is pretty... opportunistic? Its not great and you'll be in a position where you have to manually mop. Many models are switching to the spinning buffer pads which is a much better solution (that's the Dreame W10) but it refuses to do carpet. So its always a matter of finding a model that fits your use case.

>That means I would need two vacuums anyway as they haven’t learned yet to climb stairs.

Many allow for multi floor mapping on a single bot, but stairs will always be a problem for the Daleks. :P

[KptMarchewa]

S5 works the same like you describe S6Max, maybe you're thinking about some older one.

[nerdjon]

I don't understand this, I had a Neato before I switched to iRobot and the lidar just seemed to be the worst idea for a home environment.

On an almost weekly basis it would fail to start because it would think it was moved if something just moved a little bit nearby or a cat just happened to be sleeping nearby. This generally forced me to remap

Then it would constantly get lost and be confused about what room it was in. Seemingly for the same reasons of things moving around.

A home isn't a static environment and a camera just seems much more efficient. Not that a camera is perfect, but at least a camera can account for things moving around were lidar cannot.

Or was this a case of Neato just not being good? The lidar issues was why I went with iRobot. It was a worse vacuum but if it actually works consistently than I can just have it run more often.

[moepstar]

I've a Xiami / Roborock S5 for almost 5 years now (give or take a few months) and their LIDAR implementation is absolutely genius - no issues whatsoever, no matter what gets moved around or not...

Must've been Neatos implementation i suppose...

On top of the really top-notch LIDAR the S5 is beautifully engineered as well - basically everything is a module (i.e. wheel+motor) which is plugged in and secured with a few screws - really good repairability and also spares are available (needed so far: 1 LIDAR motor, 2,50€; 1 resettable fuse for charging).

[nerdjon]

I have been eyeing the new Roborock S7, but that has both lidar and camera I think.

Interesting that others handle this better. I had heard great things about Neato before so that's why I went it. But it just ended up being more frustrating and not worth it.

Guess I just had a really bad experience.

[KptMarchewa]

I would not complement S5 that much. Mine can never return to the dock alone, which is pretty basic - it knows where it is, but tries to move there from a weird angle, repeats this few times and ultimately loudly fails. Also, it always gets stuck on dryer rack.

The engineering on the modularity is neat though, I agree.

[moepstar]

Hm, that's strange - i've even used mine in 2 apartments now, on 4 floors no less..

No issues whatsoever - on the floor where there is no dock it just goes back to the point you dropped it off and started cleaning...

> Also, it always gets stuck on dryer rack.

Have you tried to put some magnetic tape (barrier) on to the legs?

[fomine3]

For Dreame Z10 (it also has LiDAR and Xiaomi sub brand), I confirmed the same. Moving objects is not a problem.

[jitl]

> Or was this a case of Neato just not being good?

I think this is a software issue with Neato. My previous bot was the cheapest Ecovacs with LiDAR and vacuum base station, and it could adapt to furniture movement, large boxes appearing and disappearing in the living room, etc without too much trouble.

LiDAR (on the Ecovacs bots) does struggle with mirrors and floor-to-ceiling windows. Even the X1 Omni is convinced there's another room "inside" the large mirror I have in the living room. It's okay though since I just draw a no-go zone or no-go lines on the map wherever there's something odd.

Really the best results will always use sensor fusion - LiDAR mapping + camera object avoidance + physical bump sensor seems to work pretty well on the Ecovacs bots. I love the improvement in cable and shoelace avoidance since upgrading to a model with vision.

[swores]

> but at least a camera can account for things moving around were lidar cannot

Why do you say that, am I missing something? Surely both technologies are able to keep "looking" as the hoover moves around, the lidar ones surely aren't doing a single analysis and then not checking again?

[nerdjon]

Others have mentioned that this may just be a problem with the Neato software.

The issue I was having with the neato is that is exactly what it was doing. It would come off its dock, do a scan around it. And if things were too different it would go back to its dock and complain that it has been moved.

So I had to find a place that things never moved, but that proved difficult. Even small things like the trash can shifting or books on a shelf shifting seem to throw it off.

I finally got to the point that I basically tricked it with blocking it from seeing certain things near its dock but that is less than ideal.

Now once it did that initial scan it seemed to be mostly ok, but it did seem too often get lost near my computer where the chair often moves. It would constantly tell me it was lost and I had to pick it up and put it back on the dock. Or my favorite was when it seemed to try to dock on the other side of my home.

My feeling (and has been confirmed by the iRobot) was that at least a camera with decent AI could account for a moving object vs most other things being the same. It is far from perfect, but compared to the Neato the camera has been leagues better.

I am very curious though how other Lidar system handle this since I would assume that there is a lack of data available to identify a change in environment vs a moved object.

[swores]

That does sound shit, sorry you've had that experience!

But I believe lidar should be able to do the same as a camera in terms of real-time updates - from a tech point of view it's just a different type of sensor to a camera, there's no reason (other than difficulty/cost) why lidar robots can't be constantly scanning & understanding the difference between a chair that you've moved since 5 minutes earlier and a cat that keeps moving. Both camera and lidar products can be made dumb enough to fail badly, or smart enough to do a great job. After all, self driving cars have been demo'd using full lidar, it's not like that technology itself takes hours to image a room.

[moepstar]

Of course i can only talk about my Roborock S5, but that one handles each and any changes in the environment flawlessly.

While cleaning, you can watch it on the map (in the cloud/app as well as when rooted on the local webserver) and can see it constantly scanning the surroundings and things appearing when they do.

I once saw an object "appearing" when there should've been none when i was out, when i got back home it was my cat that wanted to roadblock the Roborock, laying on the floor...

[_ea1k]

I have a neato and yeah, I think its neato's software. Its chock full'o'bugs. It routinely loses network connectivity and has to be rebooted to restore it. Of course, support will blame signal, but I've had that issue too and the behavior is different. A dead spot will just lose signal until it moves past it and gracefully recovers. This causes a weird error beep and then it doesn't recover.

Beyond that, I've had it vacuum a floor twice. When it finished, it showed the space twice, because it became disoriented. Basically, it didn't just think the room changed shape, it thought it doubled in size!

I've had it get lost because it thought it was on the opposite side of a rectangular room from where it actually was.

They never update it to fix these things and, IMO, haven't shipped a truly new product in several years. I'm not shocked that a lot of other lidar-based products have surpassed them now.

[nomel]

What you described is completely unrelated to LiDAR, and could just as lazily be implemented with a SLAM based vision stack. LiIDAR and vision based systems make 3d measurements of what the sensors see, whenever they look at it. What's done with those measurements in software isn't related to the ability to collect the real-time sensing of those distances.

[marcinzm]

My Roborock with Lidar has had no issues.

[jsmith45]

I've felt that the iRobot mapping was acceptable.

Once my i7 has generated a proper map, it will will slow down when it approaches known walls, or even places where obstacles have been several times in the past. It will need to bump into the obsticle or wall to confirm its position though.

It does use a mostly predictable back and forth pattern, taking note of areas with walls/obstacles, and after the main sweeping it will combe back to go around the obstacles, and finally (optional) do a sweep of the walls. Then it moves onto the next room.

The system does have limitations. For example, if it needs assistance in the middle of a clean job, after you fix it, there is a reasonable chance that it will do dumb stuff afterwards, like missing a large section of floor. (It seems to even know this, but have problematic coding that prevents it from doing the right thing).

But yeah, I'd not call the mapping good, just acceptable.

There are other annoyances like the i7 battery being so small, especially since they have a giant battery compartment. Like seriously big enough like ni-cad batteries battery of equivalent capacity. But they only offer a slightly larger capacity battery as part of their Costco exclusive model, and try to lock out third party batteries. The net result is that even when brand new, it could not quite do a single floor of my house in one go, needing to go recharge for like 40 minutes before it could do the last 10 minutes of cleaning.

The app is not very good. It does not even do some basic things like prompt the user for regular maintenance periodically based on number of hours used.

The floor maps in the app are weird. For example, on one of my floors, the online map decided that the entrance to one of my rooms was on a completely different wall from where the entrance really was. It thought the actual door was just a wall. However the actually mapping data on the robot was just fine. It knew where the door really was, and never tried to drive though the wall where the online map though the door was.

[thatfrenchguy]

> My old setup of the two robots took somewhere on the order of 5-6 hours to do a ~100m2 apartment

I mean, yes, the first time, but my i7 takes like 2 hours to do my ~140sqm apartment?