S9 review

The very latest news and updates for the iRobot Roomba robotic vacuum, the Scooba Robotic Washer and the Dirt Dog workshop sweeper. All discussion and troubleshooting questions go here.

Re: S9 review

Postby bytemaster0 » June 17th, 2019, 2:01 pm

Thought I'd provide an update after almost a week, for those who are interested. Mostly, these are observations that arise from round vs. D-shape evolution.

1. There is simply no superior shape than a round shape for navigational purposes. Plain and simple, the round bots lasted as long as they did because they become stuck less, and provide very reliable maps, with the possible exception of detailed corners.

Both in training and in operational navigation, the i7+ is simply more efficient. It's faster, and just works better. It's a lot less frustrating to watch, in general. The S9+ seeks to add a ton of detail to its maps, and requires fitting into each corner, along with its side-mounted IR sensors, to train properly.

Watching the S9+ train is an exercise in sheer frustration - just leave the area and come back 3 hours later, or you'll be tempted to yell at the machine for how stupid it looks while it's training, especially as it tries to wedge itself into tight corners over, and over, and over. Not to mention the sometimes completely unnecessary violent side-to-side "bull in a china shop" maneuvers it does when it thinks it's stuck. Sometimes these moves are needed, but often they're not, and simply cause the robot to become more stuck in a spot. Round bots can simply spin around on a dime and head back out. The S9+ can't really back up or out of a tight spot because it has no rear-facing sensors. And then the chair legs, oh, the chair legs. Once it has completed two training runs, its behavior is significantly more logical. This training "stupidity" doesn't affect the i7+ nearly as much.

2. There is simply no better shape for cleaning than a D-shape. Corners, closeness of rollers to wall, and not flicking around hard pieces all over hard floors are simply vastly superior in D-shaped designs compared to round ones. The S9+ excels here. Neato clearly knew this ages ago, but their attempts to make a vacuum that didn't get stuck always fell quite short. We mostly stopped using our Botvacs because of how often they'd be stuck when we came back hours later. However, their cleaning power was really hard to argue against. Prior Roombas were always inferior in suction and cleaning coverage. The S9+ clearly is a solution.

3. I was hoping that the brain power at iRobot would take a Neato strength and totally solve navigational issues. However, it doesn't quite. The S9+ is clearly superior to any Neato product we've ever had. Not to mention the build quality issues that plague Neato - I've had to 3D print parts for my D80s just to be able to use them straight out of the box without it complaining about the bin being removed. That being said, they've clearly learned a lot of lessons from Neato's issues to significantly improve the bot's ability to navigate without being stuck.

4. It's sort of frustrating that vSLAM isn't complemented by LIDAR. Clearly, cost is an issue, but for $1300, iRobot could have easily put one on. One would think that this could alleviate some of the small-scale issues that still plague its much-improved D-shaped design. However, I'll cede the point that LIDAR is unreliable from a longevity standpoint, and that it suffers significantly from artifact issues. Furthermore, it would probably prevent the bot from aggressively pushing under bedsheets to see if it can clean under a bed, or other obstacle with soft borders.

5. The firmware versions on i7+ and S9+ are different (2.x vs 1.x). Not sure how this affects its navigational intelligence or other features. The i7+'s recent upgrade to 2.x made significant strides in eliminating navigational issues where the robot would frequently become "lost" when navigating perimeters or table legs. Hopefully the S9+ will catch up, as it doesn't seem quite as capable yet in some certain situations.

6. The rollers on the S9+ don't tangle up nearly as much hair. Having long rollers with the suction point in the middle seems to largely reduce this issue, unlike previous bots. This is a very impressive improvement.

7. The S9+, all in all, is an impressive attempt to combine points 1 and 2 - the best shape for navigation with the best shape for cleaning. The wheel placement, caster placement, brush placement, width and aspect ratio are clearly very carefully chosen. It's quite evident when the S9+ is able to maneuver around objects or obstacles that would have completely ensnared a Neato Botvac. And, with the higher suction power, it's a formula for success. It'd be nice if they could do something about the price, but it is very solidly made, and feels like it will run for a long time and be able to deal with a large variety of operational scenarios.

Here's to hoping for its continued improvement through firmware updates!
bytemaster0
 
Posts: 43
Joined: October 15th, 2018, 3:48 pm

Re: S9 review

Postby jdong » June 17th, 2019, 3:02 pm

Thanks for the updates! After spending 5 years with Neatos I actually really love the round shape for Roombas too. The D shape is great for cleaning corners but it adds a ton of issues for navigation since it changes its footprint when rotating. I can imagine it suffering a lot more with getting stuck under a dining table where the legs are too thin to efficiently plan a route through.

Given how iRobot didn’t even get one navigational update released for the i7 (Other than begging for 2.0.x), I would rather wait and see.
jdong
Robot Addict
 
Posts: 329
Joined: April 10th, 2015, 11:24 pm

Re: S9 review

Postby BoxerBoi76 » June 17th, 2019, 3:59 pm

Excellent follow up review! Hoping the s9 is just having teething issues and iRobot will rollout the equivalent of what 2.0.9 is to the i7 with more fixes!
BoxerBoi76
 
Posts: 38
Joined: June 9th, 2019, 5:06 pm

Re: S9 review

Postby Julian » June 18th, 2019, 4:37 am

Superb. Thanks for the update.

On the navigation issue and the ability of a round shape to spin on its axis, that is good for getting out of tight spots quickly but I'm actually quite looking forward to getting a robot that doesn't have the ability to spin on its access because I have "crop circles" in the corners of all my rooms where over the years my 560s have spun on their axis tens of thousands of times in the same spots (and I mean that number literally, I just did a quick mental calculation and I put it at somewhere between 45,000 and 90,000 times in each spot!). After all this time I've mostly stopped noticing them but I still do sometimes so it would be quite nice to not have something that leaves those persistent patterns in the corners.
Julian
Robot Addict
 
Posts: 131
Joined: May 29th, 2009, 7:23 pm

Re: S9 review

Postby RTC » June 18th, 2019, 5:56 am

Julian wrote:I have "crop circles" in the corners of all my rooms where over the years my 560s have spun on their axis tens of thousands of times in the same spots (and I mean that number literally, I just did a quick mental calculation and I put it at somewhere between 45,000 and 90,000 times in each spot!).

I have never noticed my 980 spinning in a corner, so I really wouldn't expect the newer circular shaped Roombas to do that either. The 980 does a series of linear back and forth runs in perpendicular directions intermixed with wall following "wobble movements" to cover the edges. The only times I have I have ever caught the 980 spinning in place were when it had entered a dark room at night and lost its orientation (this was easily resolved by turning on a light).
Technologist (i.e. one who studies technologies)

From the earliest "sticks and stones" technologies of our prehuman ancestors to the latest "high tech" technologies of today, I like studying it all!
RTC
Robot Addict
 
Posts: 162
Joined: August 23rd, 2018, 7:23 pm
Location: Lost someplace in Montana, USA

Re: S9 review

Postby jdong » June 18th, 2019, 11:51 am

The newer round robots with the camera (eg the 980 and above) don’t do crop circles in corners anymore. The combination of the camera / IMU / optical ground tracker makes it surprisingly precise about remembering how to free itself. I’ve seen it many times stumble under a chair and then spin precisely to the angle required to back out or drive forward and escape the chair. Sometimes it’s off by a couple of degrees.

Totally understand what you are describing with the older Roombas though — the bounce bot era Roombas would frequently do in place spinning in tight spots, drawing useless carpet pizzas :D
jdong
Robot Addict
 
Posts: 329
Joined: April 10th, 2015, 11:24 pm

Re: S9 review

Postby Botty » July 3rd, 2019, 3:49 am

S9 vs D7 comparison

https://www.youtube.com/watch?v=OW2s93lU1_8

I'd love an S9, but i could buy three D7s here for the price of one S9, so pretty hard to justify the marginal difference.
Botty
Robot Addict
 
Posts: 260
Joined: August 27th, 2018, 9:53 am

Re: S9 review

Postby Foster6565 » July 3rd, 2019, 4:15 pm

I've been and avid neato user and regardless of what that review says the s9/i7+ are I'd say more then worth the cost. The roombas basically never get stuck or lost can empty themselves and cleaning the brush rolls is much much easier. I cant see how the neatos rank that high after owning both, maybe my house just isnt lidar friendly.
Foster6565
 
Posts: 9
Joined: June 11th, 2019, 5:27 pm

Re: S9 review

Postby jdong » July 3rd, 2019, 9:23 pm

Foster6565 wrote:I've been and avid neato user and regardless of what that review says the s9/i7+ are I'd say more then worth the cost. The roombas basically never get stuck or lost can empty themselves and cleaning the brush rolls is much much easier. I cant see how the neatos rank that high after owning both, maybe my house just isnt lidar friendly.



I 100% agree with you. Loved what Neato has done to advance the tech in this field but hands down recommend the Roomba flagship products if money is not an issue. They have fewer navigation issues but much more importantly rarely get stuck and can free themselves from sucking up cords far more often. I’ve had the D7 utterly destroy fragile cables under my home theater cabinet. With no modifications to the setup Roombas since the 980 do just fine.
jdong
Robot Addict
 
Posts: 329
Joined: April 10th, 2015, 11:24 pm

Re: S9 review

Postby Foster6565 » July 4th, 2019, 6:12 pm

jdong wrote:
Foster6565 wrote:I've been and avid neato user and regardless of what that review says the s9/i7+ are I'd say more then worth the cost. The roombas basically never get stuck or lost can empty themselves and cleaning the brush rolls is much much easier. I cant see how the neatos rank that high after owning both, maybe my house just isnt lidar friendly.



I 100% agree with you. Loved what Neato has done to advance the tech in this field but hands down recommend the Roomba flagship products if money is not an issue. They have fewer navigation issues but much more importantly rarely get stuck and can free themselves from sucking up cords far more often. I’ve had the D7 utterly destroy fragile cables under my home theater cabinet. With no modifications to the setup Roombas since the 980 do just fine.


I actually was thinking of posting up a review/comparison of the neato D4 vs my i7+. I purchase the d4 over a week ago to be a permanent helper for my upstairs (all carpet), but it was constantly stuck/lost and completing the training run was just pure pain to be honest. irobot is miles ahead of everyone right now. I returned the d4 needless to say and will just wait a little longer to convince the wife to get the s9+.
Foster6565
 
Posts: 9
Joined: June 11th, 2019, 5:27 pm

Re: S9 review

Postby Julian » July 11th, 2019, 4:33 am

ertooso wrote:The s9 seems to be only a few millimeters higher at most. It's also only the IR sensor on the front that causes the issue too. Shame they couldn't do away with it with this redesign.

Maybe it was to hedge their bets on no-go lines. As I understand it the IR sensor on top is designed to inter-operate with the similar-looking transmitters on top of the non-emptying docks and the virtual walls/lighthouses. With the smart mapping in s9 it seems pretty clear to me that the need to use it to find its dock has gone away (was it ever used for that or was the one on the dock simply a halo to stop older Roombas pushing around their own docks when cleaning?). Probably more significantly though, given that s9 didn't initially ship with no-go lines, meant that if iRobot wanted that sort of functionality out of the box then being able to interact with the virtual wall, particularly the dual mode one in halo mode, presumably meant that retaining the top IR sensor on the s9 was the easiest and safest way to go. That also gave iRobot a get-out if implementing their no-go features in the firmware turned out to be more difficult than expected rather than leaving purchasers of the flagship product with no way to cordon-off dog bowls etc. Shipping s9 without the ability to use existing virtual walls might have been felt too big a risk.

Once Roomba is confident that the no-go box/zone/line/however-it-evolves capabilities in the mapping software are solid I wouldn't be at all surprised to see that top sensor go away and future models rely 100% on software no-go functionality to implement what used to need lighthouses. And I agree, that would be a good thing.
Julian
Robot Addict
 
Posts: 131
Joined: May 29th, 2009, 7:23 pm

Re: S9 review

Postby ps0001 » July 11th, 2019, 11:41 am

Julian wrote: With the smart mapping in s9 it seems pretty clear to me that the need to use it to find its dock has gone away (was it ever used for that or was the one on the dock simply a halo to stop older Roombas pushing around their own docks when cleaning?).

Yes, it's definitely used for that. The older models kept away from the dock signal while cleaning, then sought it afterward. I believe the first mapping models (i.e. the 9xx series) generally know where the dock is, but they still use IR signals to precisely locate the dock and park. Some of the 9xx models had an issue where one of the wires from their RCON sensors would get pinched and snap. They would return to the same area after cleaning, but they couldn't actually dock.

Despite having an i7, I don't know exactly how the i7 and s9 work regarding IR, but I know navigation doesn't really need to be more precise than 5-10 cm to clean a house. It needs to be significantly more accurate for repeatable docking for charging and bin evacuation. I can't say for sure, but I wouldn't be surprised if IR is still used for this purpose.
ps0001
Robot Addict
 
Posts: 202
Joined: December 5th, 2017, 4:25 pm

Re: S9 review

Postby bytemaster0 » July 11th, 2019, 2:05 pm

Definitely still uses IR for docking. The emitters are in the window below the dust bin section of the clean base. You can also tell it uses IR for docking based on its circling behavior when placed too close to the dock.
bytemaster0
 
Posts: 43
Joined: October 15th, 2018, 3:48 pm

Re: S9 review

Postby Julian » July 12th, 2019, 7:15 am

bytemaster0 wrote:Definitely still uses IR for docking. The emitters are in the window below the dust bin section of the clean base. You can also tell it uses IR for docking based on its circling behavior when placed too close to the dock.

Indeed, but is the s9 itself using its top sensor or its sensors in the front bumper? The robot's top sensor is carefully designed to be at the same height as the ones on lighthouses and the dock but on the self-emptying docks all the IR stuff is hidden behind a panel so it's hard to see what height the dock IR transmitter is set at.

I take the point on precision IR navigation to do the final drive-on and line up to the charging contacts but I do wonder, if it is still using the top sensor for the s9 (and I accept that it probably is - that circling behaviour is a bit of a giveaway), whether in a future model that final docking navigation could switch to using the bumper sensors and the top sensor dispensed with since, once software no-go functionality was rock-solid, there would be no requirement for the top sensor to be there to interface with lighthouses.
Julian
Robot Addict
 
Posts: 131
Joined: May 29th, 2009, 7:23 pm

Re: S9 review

Postby bytemaster0 » July 12th, 2019, 2:34 pm

Pretty sure it uses the top sensor. The wiggling back and forth that it does during docking is identical to every iRobot product produced in the last decade. Neato robots don't wiggle as they get close to the base (they do have a tail wagging behavior to ensure good contact with the charging base in the last few cm) as they use the LIDAR receiver for positioning. The top-mounted sensor on the Roombas requires wiggling to sense changes in brightness and maintain its centering. It's cheap and easy to implement, makes sense that they'd still keep it for that purpose as well.
bytemaster0
 
Posts: 43
Joined: October 15th, 2018, 3:48 pm

Re: S9 review

Postby rotorwash » July 17th, 2019, 9:17 am

So after 23,000 sqft of vacuuming I seriously need to replace the extractors. The finned one has multiple rips which now has become a hair collector. The foam on the inside is starting to fall out when I pull the extractors out for cleaning. Also the back extractor is worn till where it is almost smooth in the center.
I guess this just shows how close to the ground these are compared to earlier models. I just wish they were a little cheaper.
rotorwash
Robot Groupie
 
Posts: 92
Joined: September 16th, 2018, 2:10 pm

Previous

Return to iRobot Roomba and Scooba Chat

Who is online

Users browsing this forum: KeldBroe, Per L and 117 guests