The future of Roomba

[sean_gmu]

The way I see it, the innovation that is key to Roomba's growth and success in the future is the development of a roomba that can sense its surroundings, know where it is, and where it has already been. I would think such a roomba should be able not only to know how to systematically vacuum a room, but to know what room it is in. Why?

1. Durability. You may wonder what durability has to do with this. Think about it. Irobot says that its blind cleaning algorithm makes it such that roomba cleans all areas of a room an average of 4 times to ensure complete coverage. In other words, the robot is wearing out four times faster than it would if it vacuumed the way we do, covering the whole room only once. No wonder, from what I gather, these things tend to break down quickly. Despite what Angle has said about these blind vacuums actually doing a better job because they don't "see" obstacles that really aren't obstacles, such as curtains or furniture that is low to the floor, I would bet a lot of money that irobot is working feverishly to develop a robot with a better sense of its surroundings to save itself warranty costs in the future.

2. Cleaning ability. If irobot developed a robot that cleaned a room only once, or even twice with overlapping paths, it could have a vacuum motor that is much more powerful and cleans better. Obviously, it would not run as long, but it would not have to to cover the same amount of territory.

3. Amount of floor cleaned: This one really depends on what future advances in cheap battery technology will be. To have a truly autonomous vacuum, it would be nice if roomba could cover an entire floor of a house. Just let it go, and watch it do the living room, hallways, and bedrooms. Even with the current battery, and if you kept the current vacuum motor, with a roomba that keeps track of where it is, you could easily clean 2-4 more times as much carpet, depending on whether you want the vacuum to cover each area once or twice, which from what I understand would be enough to cover an entire floor of a good sized home. Of course, if lithium ion batteries continue to get better and cheaper, my understanding is that they last twice as long as NIMH, so you could again make the vacuum twice as powerful and still be able to vacuum a much larger area than is presently possible.

Of course, it goes without saying that something that can automatically dump the dirt would be necessary to make these changes practicable, because the robot will be picking up a lot more dirt with each use.

I wonder if irobot realizes that it has reached a plateau with the technology they currently have? Most people, including myself, are not will to buy a robotic vacuum with the current limitations, because it really isn't all that autonomous, and from what I have read, are either not durable enough, don't clean well enough, or both. The company seems concerned about making a robot that costs too much, but I think they could go higher on the price to get some of these features. Look at how popular the Dyson is, and those sell for $500-$600. I would think that a robot vacuum that does most of the things I have talked about here would easily be worth that, and Dyson has shown that such a price does not take you out of the mass market if you have a useful product.

My suspicion is that a vacuum with the ability to to all or even some of what I have mentioned here simply costs more than $500. If so, hopefully the time isn't too far off when costs will come down enough, and then irobot will have true mass market appeal. 1.2 million sold so far sounds good, but it is a tiny percentage of the 300 million people in the U.S.

[Jobot]

Yeah, that claim: "Cleaning an area up to 4 times!!" Is so dubious, it's laughable, really. Creative, but laughable. What they really mean is: once, twice, three, four, five... you get the picture. I've watched my Discover wall crawl the same edge six times during one run... before I stopped counting.

Here's a slight dilemma, though: To add mapping, you'd probably have to slow the unit down quite a bit. And add a lot more heavier components to allow mapping to occur, don't you think? Just sounds like the ability to more efficiently clean a room with mapping technology may not -- entirely -- be the end all solution. 'Course, we're speculating, so it's easy to "speculate out" those other variables (weight, velocity).

Just start with some basics: I've had a Roomba now for two years, it it's just perfect. But, I've also read the horror stories, experienced a few dancing, coughing units, returned them, so start with durability. Man, IRobot, hellooooooo. Actually, they appear to have tried.

Overall, tell you what they're gonna do: Incrementally add features over the years: More durability, more user control variables (Hey, if I only want to clean the Room for 20 minutes before automatically heading home, right now, I can't do that. I have to let the unit decide, or drive it there myself.)

Anyway, to continue: What I really believe the next big advance, after the little features above... (You heard it here first, folks!!): Will not be mapping: It'll be a redesigned Roomba that includes a vacuum assist automatic bin emptying home base. Betcha.

[RoombaRules]

Interesting ideas Sean. I agree with some of what you said, but cheapness is a much bigger consideration in my opinion. Oh and Lithium Ion batteries do not last twice as long as NiMH batteries. There are a few kinds and most of them actualy have shorter life spans than NiMH, this is the leading complaint about the iPod, it uses a Lithium Ion battery that's not even replaceable.

Jobot: Yes I agree with that completely, in fact I spoke to an engineer that works for iRobot a little while ago, he said that they have already begun working on designing a dustbin emptying system for Roomba 3.0

[sean_gmu]

Certainly cheapness is a concern. That's why I am not advocating trying to compete with the ridiculously high priced Electrolux. That said, for a vacuum that can do even half as good a job as a Dyson, why wouldn't someone be willing to pay $500-600 for it, since you don't have to lift a finger? Time is money.

I am not so sure you'd have to slow down the robot to have it map the room. Look at the recent Grand Challenge, where vehicles were able to hit 30-40 miles an hour, driving completely autonomously, in an environment designed to trip up the robot vehicles, with gullies, sandtraps, tunnels, trees, potholes, and dusty roads. Those vehicles were successful with something like 4 or 5 Pentium M's as their mental horsepower. A roomba goes 100's of times slower, and could presumably do its mapping with 100's of times less processing power. That's without even accounting for the fact that a home is much less dynamic an environment than the desert of Nevada.

[celron]

I recently made enquiries into purchasing the robot Dyson and was told that it was discontinued.... Can anyone shed some light on this?

Cheers,

Aaron

[smorris]

Mapping a room is very complex. You have to have either an immutable zero point or a continous location reference. One problem with a zero point is wheel slippage on the floor, it is variable and as a consequence you can't use travel as a reference. Also the unit would have to start from the same point everytime or it has no reference.

A better system would be continous location reference, which has a minimum of 3 points for spatial reference. These could be IR sources it triangulates or radio becons which it measures the time differential in the signals. Then all the machine has to remember is the coordinates of collision points to effectively create an internal map. Using this technique the system could determine what is a wall and what is meerly a point obstruction using straight forward algorithms. Basically you have to build the equivalent of a chartplotter into the machine, the good news is it only has to be wireframe.

But again it isn't trivial. System development = cheap, fast, good...you can only pick 2.

[RC Eden Carpet Cleaning]

I recently made enquiries into purchasing the robot Dyson and was told that it was discontinued.... Can anyone shed some light on this?

Cheers,

Aaron

Ya I know a little about this. From what I have been told by friends that work for Dyson, the robot dyson was discontinued because in order for it to do everything that Dyson wanted it to do it would have cost more than $6,000. Mr. Dyson didn't want to cheapen the vac so he abandoned the project all together rather than dumping some of the features. I'm told that as soon as the cost to build a Dyson robot vac goes down they will consider the project once again.

[RoombaRules]

Ya I know a little about this. From what I have been told by friends that work for Dyson, the robot dyson was discontinued because in order for it to do everything that Dyson wanted it to do it would have cost more than $6,000. Mr. Dyson didn't want to cheapen the vac so he abandoned the project all together rather than dumping some of the features. I'm told that as soon as the cost to build a Dyson robot vac goes down they will consider the project once again.

Wow, what on earth did they want it to be able to do? That's just crazy, $6,000!
Heh, I bet they wanted it to have wings and clean your ceiling...

[RC Eden Carpet Cleaning]

I think that my friend said somthing about it being as powerful as an upright and it would have an attachment to do stairs. It would also be able to map out the room it was cleaning and of caurse it used ultrasonic to navagate. I too can't believe a $6,000 price tag

[Jobot]

I think that my friend said somthing about it being as powerful as an upright and it would have an attachment to do stairs. It would also be able to map out the room it was cleaning and of caurse it used ultrasonic to navagate. I too can't believe a $6,000 price tag

There it is, right there: Why Roomba won't be mapping anytime soon. Or matching the power of an upright. Here's an existing company with the wherewithal to actually produce a mapping, upright-power-capable vacuum robot, and: there's your figure: $6,000. If they could do it for one-sixth that, maybe... you might be in the ball park. But... as much as our intuition might convince us that "mapping should be next" or "power," is within reach as the next iteration of the Roomba... nah, in't ginna happin' There's the dollars right there.

Also, what makes us all think that a unit like, say, the Electrolux Trilobite is any more durable than a Roomba? Not saying it isn't -- I know the materials, cladding, underside dust protection are better. But, hey, electonics go "down" all the time. I'd rather deal with a down Roomba then a down Electrolux.

I dunno, Roomba seems to have built a unit, slotted and marketed it perfectly for what it does. And sales of 3 million-plus... I wouldn't mind selling a $250/unit widget to 1 percent of the U.S. population...

[RoombaRules]

You want to know what I find odd? ...
If you look in the iRobot FAQ about Roomba one thing they say is that: at the point of contact with the floor, Roomba's suction power is equal to an upright vac.

[RoombaRules]

Here we go! I FOUND IT!:

Question:
How much suction power does Roomba have?

Answer:
At the point of contact on the floor, Roomba has as much suction as a standard upright.

Here's a link to it:
http://www.irobot.com/sp.cfm?pageid=46& ... tegoryid=6

[Kenwood True X]

better system would be continous location reference, which has a minimum of 3 points for spatial reference. These could be IR sources it triangulates or radio becons which it measures the time differential in the signals. Then all the machine has to remember is the coordinates of collision points to effectively create an internal map. Using this technique the system could determine what is a wall and what is meerly a point obstruction using straight forward algorithms. Basically you have to build the equivalent of a chartplotter into the machine, the good news is it only has to be wireframe.

I agree, triangulation would be the best, most accurate method. The beacons could be placed around the house into wall outlets. In fact the Nintendo Revolution uses a similar concept, although it uses IR to determine the position of the remote instead of RF.

[RC Eden Carpet Cleaning]

Here we go! I FOUND IT!:

Question:
How much suction power does Roomba have?

Answer:
At the point of contact on the floor, Roomba has as much suction as a standard upright.

Here's a link to it:
http://www.irobot.com/sp.cfm?pageid=46& ... tegoryid=6

I wander what iRobot considers a "standard upright." For all we know it could be the power of one of those mini vacs that think they are full size vacs, like the Eureka Boss.

[Kenwood True X]

I don't think the Roomba has much more "suction" then a dust buster. It mostly uses it's brushes to collect dirt and hair. That's why the Roomba works so well on hard floors and low pile carpet, but not as well on medium pile (also because it lacks a beater bar).

I think on hard floors and low pile the Roomba can outperform an upright because of its design, but it won't perform as well as an upright on medium pile and higher.

[Polly]

I think iRobot's statement is true, as far as it goes.

The roomba vacuum does have a powerful suction, but only at very close range and only in a tiny area. This suction comes the fact that it has an extremely narrow silicone nozzle which brushes against the floor.

On hard floors or low pile this works well, but in a deeper pile rug, there is a larger distance between the nozzle opening and the dirt, since the dirt is lying down in the pile.

This weakens the suction seen by the dirt considerably, and accounts for the roomba's reduced cleaning ability on these deeper rugs.

[Polly]

On mapping etc. -- hardware isn't the problem.

Ultrasonic sensors are not very expensive (although they are a lot more expensive than a switch and a spring). Here's a pair for $10 http://www.hobbyengineering.com/H1324.html. In quantity would probably be more like 50 cents to $1.

So what about processing power? CPUs fall in price every year, and so does memory.

I don't think this kind of mapping is really very CPU intensive, but if it is, a 200MHz processor should be more than enough. Here's a whole embedded computer for $65: http://www.glomationinc.com/product_9302.html

Or if it turns out that not so much CPU is needed, this one is $25 in quantity 1000 http://www.zworld.com/products/rcm2000/

What's left? Programming. Some people have suggested that absolute wheel registration is needed, but instead the unit could self-correct for drift by doing regular sensing and seeing how reference objects move.

But the programming is challenging.

This is the real cost of producing a mapping smart unit. The software. The rest of it isn't so expensive, but writing the code and making sure it really works is very tricky and time consuming.

So don't expect prices to drop so quickly on mapping units just because hardware prices are falling. Hardware isn't the problem.

[cyberman]

This is the real cost of producing a mapping smart unit. The software. The rest of it isn't so expensive, but writing the code and making sure it really works is very tricky and time consuming.

So maybe Irobot can finaly get round to releasing the programming interface
and we can do this bit for them

[madrid]

I would have serious reservations using Roomba around children or pets if it had 4x more suction power. There would also be more safety concerns. eg. area rugs, electrical cords, pieces of bedspread touching floor. I guess there are pro's and con's. My roomba is pretty forgiving if I don't do a perfect job of getting the room ready. I honestly don't think I would want it to have much more suction power for these reasons. However, if I had no children and could 100% control the environment of my home, it would probably be nice.