The new JetBot series and three new models of Samsung vacuum cleaners:
VR50T95935W VR30T85513W VR30T80313W
The vacuum cleaners are equipped with a new LIDAR-based navigation system (laser) for this brand, connectivity with the application in which house maps are to be created with the determination of specific rooms for cleaning, virtual barriers. They use the latest self-cleaning soft brush. The VR30T85513W and VR30T80313W models are equipped with a self-emptying docking station where the dust is collected in a removable bag. The VR30T80313W model also has a container with a cyclone module in the robot itself, a rectangular front with a brush and a monitoring camera, as well as a system for recognizing small obstacles, e.g. animal bowls, socks, toys or other objects left on the floor.
It seems these JetBot models shown some months ago are now coming to market.
I prefer the old Powerbot line with camera navigation instead of lidar with mechanical wear, so many repair reports in the Neato products here. They are selling the Powerbots at very low prices suggesting that line is being discontinued, usually the case when new models come out. The larger size 9000 series stopped being listed on their website some months ago. I also don't like the design with side brushes in the these new JetBot models same as old Roomba's.
Curious that the website now shows a lot of negative reports from customers, with company replies, usually not seen at company advertising sites. Never had a problem with Powerbot after years of use. To preserve it most now keep it off and the dock unplugged between cleaning, save wear on electronics.
These models abandon Samsung's previous camera guidance for lidar scanning. At the same time they add virtual "No Go", boundaries previously not supplied, though available from Neato (not needing physical barriers, mag strips etc.). Perhaps the camera is not as precise as the lidar for implementing this feature.
Possible explanation for switch to lidar mapping from camera mapping in the 2022 JetBot introduction: the diagram below shows how location is sensed with lidar or camera, so-called SLAM simultaneous location and mapping. Both lidar scans and camera images can identify a landmark point such as a corner of the room, from pattern recognition -- and track this landmark as the robot moves to different positions. Measuring the angle to the landmark from two positions a known short distance apart from travel gives distance to the landmark by triangulation, locating the robot within the space. With lidar the location relative to the landmark is determined without travel once the room has already been mapped, because the lidar reads the distance, not available to the camera without moving a short distance. This might make the difference for locating relative to virtual boundaries or "no-go" lines drawn on a screen instead of physical boundaries, mag strips etc. The camera may not be as precise as the lidar in this application. The company does not explain their reasoning, so just a possibility.
[edit] Another factor is that lidar sees furniture obstacles low down compared to upward facing cameras mainly seeing the ceiling (though new models now may have forward facing cameras with small obstacle detection -- as well as lidars).
[edit] Note the camera robots have only one camera instead of two for binocular vision with depth perception like a person -- where baseline distance between the cameras would be precise; relying on movement of the robot could be less precise.
So I've recently bought one of these (the Jet Bot AI+) in the UK and can't say I'm very impressed.
The AI element appears to be very flakey and it's constantly reporting objects that aren't there.
For example, my large rug us identified as a sock/towel and the bot just won't go on it, despite the fact the much older and cheaper Neato D700 mounts the rug no issues. Changing the size of the objects to detect does allow it to go on the rug but its not perfect, there are still areas it won't touch because it flag objects that aren't there.
I have barstools that are of the single column type with the circular base and it identifies these as pet waste, obviously they look nothing like pet waste.
Samsung seems to be aware of some of these limitations as the manual mentions adding no go zones over rugs, depending on their size, with the size dependent on which cleaning mode you have selected. The manual is hard to follow, almost as if it has been written by someone who's first language isn't English.
The carpet detection (for variable suction power) is also poor. I've heard the motor ramp up when it gets close to hard floor edges which I'm not sure is a feature to get more edge dust or if the bot has detected more dust, but the motor never seems to ramp up when it gets on carpet.
Again, the manual mentions that detection won't work on certain light coloured carpets which quite frankly, seems ridiculous. So features won't work with light carpets and if you have too dark carpets, drop sensors detect these as a potential drop. So you must have a very specific shade of carpet?!
Lastly, I've noticed the bot doesn't like going under furniture despite the fact it can definitely fit under with enough of a gap between the top of the lidar and the bottom of the furniture.
For £1,200 I would expect better to be honest.
It would help if Samsung were easier to contact about these issues and give feedback to engineers who could improve the AI. I've actually found work arounds for some issues such as my bot getting stuck in the corner of a room because it thinks the stools nearby is pet waste and nearby rug is a towel.
I worked around this issue by adding a no go zone around the stools, but before I deduced what was causing the bot to get stuck, I reported the issues to Samsung via web chat and it was a nightmare.
They force you to go through a script of questions that aren't related to issues you're experiencing. "Have you cleaned the filter?". No, the bot is literally a day old and how is the dust filter going to impact the bots ability to navigate?!
Thanks for heads up. I liked the older Powerbot with camera guidance, but this sounds like another over-reach by designers. I was dubious about the switch to lidar itself, with mechanical wear. Maybe lots or returns and weak sales will bring them to their senses. I liked the durability of the electronics from a major electronics firm compared to other brands, so this is a disappointment. Best obtained from sources with a generous return policy. Going to depend on details of each environment.
In favor of Lidar over camera found this remark in reviews:
"After doing some research on the roborock vacuums and roomba there seems to be a general concensus on both sides that LIDAR atm is much better than Vslam (visual simultaneous location and mapping, camera) for navigation and object detecting (vs plain SLAM)." No references were provided.
Goes with previous post about precision of location with lidar vs camera; still Roomba uses camera.
Of course besides the mapping detection Samsung includes a lot of short range IR proximity detectors around the bot, behind the black IR transparent windows.
[edit] An old problem with Lidar is glass doors, mirrors and reflective chrome furniture legs that don't return laser beam to detector.
Advances in LIDAR
The new model lidars may differ from Neato Robotic's original design twenty years ago. The SPARKFUN.com hobby store sells a small short range lidar specified as "time of flight" like radar vs the parallax range finder method of the original Neato device -- twenty years of advances in electronics could make the difference. A hobbyist here used this SPARKFUN product with a spinning mirror for scanning, instead of the bulky NEATO spinning module with spinning electrical connections subject to wear etc. SAMSUNG details are unknown but they could be more of this type more durable, and lower cost.
Testing results for the top Jet Bot AI+ (square shape, without side brush) : brief test of a unit found unusual things.
1. The brush has no bristles, and is a velvet like material with small 1/8 in. vanes, compared to bristles on the Powerbot series. It also has their hair cutting feature. There is a question whether this can clean plush carpet without a bristle brush, but I did not have the extensive time required to measure dirt collection (and to get a dirty floor...).
Customer service could not comment, having access only to published product description.
2. As specified it has 30 air watts suction compared to 40 on Powerbot, but the brush width is 9 in. vs 12 so is suction effectively the same for the intake area, per square inch. Samsung seems the only maker who gives an air watt measure.
3. It has a minimum run time of 1hr (and increased max run 90 minutes), and goes over smaller areas several times. The run would have to be manually interupted for shorter run times, should a room need to be used etc.
4. Unlike the Powerbot, it has a wall following feature, and small wheels on the front corner sides similar to older smaller models, to roll against the wall. In the general "go everywhere" run it first did inside walls and then later went around following walls. The Powerbot does not do this, and created some problems with cord guarding fixtures which would need to be additionally secured to avoid getting shoved around.
The proximity sensors work very well to keep it away from the walls, turning before bumping.
5. Most unusual the navigation would not travel underneath some cabinets on 6 in. legs under which the Powerbot always cleaned under -- just passed them by. It audibly announces when avoiding an obstacle with its new camera detection, and made no remarks about the cabinets, just avoided going under. This appears to be by design.
6. It is very quiet despite the high suction, compared to noisy Powerbot, perhaps the motor is enclosed in sound deadening material or something.
7. The entire front case cover is a spring mounted bumper, instead of small strip bumper on Powerbot, so it has full side bumper detection.
8. The obstacle avoidance feature has settings in the smartphone app for size of obstacle avoided. Wall following also adjustable on/off. Complex controls. Virtual boundaries, no-go-zones included.
9. There is no display panel on the bot since the info is complex enough to require a smartphone screen. There is also no IR remote which would also be too simple. Just a fancy led indicator sweeping light from the top edge, and three buttons on top: start, stop, go-to-base.
10. The drive wheels are only 2.5in. wide, compared to 4 in. Powerbot and 3in. most smaller bots, but have large 1/4 in. spike treads.
11. Odd behavior at a 1/4 in. thick wood chair mat for desk: when first encountered it would not climb but go around it -- without any obstacle announcement -- but then later came back and covered it. So very sensitive floor detection, impressive.
[edit] 12. Unusual detail: on the side with the brush drive connection, there are two short half inch high, thin rubber blades at the corner, despite the drive fully enclosed in a case, maybe keeping dust out of that area.
[edit] 13. Magnetic boundary strip sensing less than on Powerbot -- strip needs to be closer to the bot. A strip positioned behind a 1/4 in. carpet to hard floor transition, behind a metal molding, worked on Powerbot, and was sensed at some spots by Jet Bot but got crossed at another point. The strip would need to be put atop a 1/4 in. molding strip to work, or secured atop the carpet (and the Jet Bot will bump on the ends of it...). Virtual boundaries are the new preferred thing, on other brands as well.
These typical tests do not test carpet cleaning properly in my opinion, which is difficult to do, because it requires dust ground into the plush of pile carpet, and measuring the extraction by weight with a gram scale. Still higher performance in the typical surface debris test can be suggestive of better performance on the real dirt, but I still wonder with this machine about its unusual brush lacking bristles or blades.
