4 ms·

Am I the only one who thinks that iRobot is actually doing a very poor job with the Roomba cleaning algorithm? Why can't we have a robot that maps the surface

by jfabre 12y ago

Am I the only one who thinks that iRobot is actually doing a very poor job with the Roomba cleaning algorithm? Why can't we have a robot that maps the surface and then cleans accordingly? I always get irritated when I look at my Roomba doing it's job.

chiph 12y ago
The Neato vacuum does a back & forth pattern. But after living with a Roomba for a while, I've grown to appreciate their mostly-random pattern - it (eventually) makes sure that the nap of the carpet gets brushed in different directions, so your carpet doesn't look like the greens at the local golf course, with stripes in it.
t0mbstone 12y ago
Look at the Neato. I switched from Roomba to Neato and never looked back. It actually vacuums in an intelligent manner.
Luc 12y ago
Because you need more processing power and sensors to first build the map, and then know where you are on the map.
akramhussein 12y ago
Check out the upcoming Dyson 360 Eye - http://www.dyson360eye.com/ http://www.dyson360eye.com/ Disclaimer: I worked on it. Edit: To answer your question directly, you have 3 choices: 1. Don't map (or somewhat guess it) - Roomba. 2. Map first, then clean - Neato. 3. Map as you clean - Dyson 360 Eye. This is obviously a crude explanation but they get harder going down the list. If you are interested to know more, read up on SLAM (Simultaneous Localisation and Mapping) algorithms.
- robterrell 12y ago
  How is it better than Roomba and Neato -- what's the cleaning algorithm? Does it map the space? etc. Edit: thanks, you answered while I was asking. That's a very interesting use case for SLAM. Does it re-map every use (i.e. today there's toys on the floor, but tomorrow there's not)? Does it store the point cloud or convert it to something more manageable? Are you licensing someone's SLAM library or did you roll your own?
  - akramhussein 12y ago
    Unfortunately I'm not really able to go in to much more details, but this video explains what's public knowledge - https://www.youtube.com/watch?v=oguKCHP7jNQ https://www.youtube.com/watch?v=oguKCHP7jNQ Sorry!
berberous 12y ago
It does a pretty good job though: http://www.designboom.com/weblog/images/images_2/jenny/roombaart/roombaart01.jpg http://www.designboom.com/weblog/images/images_2/jenny/roomb... I tried a Neato but it was much more prone to getting stuck (every single time actually). Obviously that will vary depending on your furniture.
IndianAstronaut 12y ago
This is actually better done with a robot team or swarm. There are several 'frontier' algorithms which use basic particle swarm or metaheuristic algorithms to rapidly move across a landscape or frontier.
snowwrestler 12y ago
What's better, a brute force algorithm that is cheap and simple but slow, or a highly optimized system that is expensive and complex and fast? The answer is of course "it depends." Certainly the Roomba could be smarter, faster, and more efficient, but the question is how much more are customers willing to pay for those.

by jfabre 12y ago

chiph 12y ago

The Neato vacuum does a back & forth pattern. But after living with a Roomba for a while, I've grown to appreciate their mostly-random pattern - it (eventually) makes sure that the nap of the carpet gets brushed in different directions, so your carpet doesn't look like the greens at the local golf course, with stripes in it.

t0mbstone 12y ago

Look at the Neato. I switched from Roomba to Neato and never looked back. It actually vacuums in an intelligent manner.

Luc 12y ago

Because you need more processing power and sensors to first build the map, and then know where you are on the map.

akramhussein 12y ago

Check out the upcoming Dyson 360 Eye - http://www.dyson360eye.com/ http://www.dyson360eye.com/ Disclaimer: I worked on it. Edit: To answer your question directly, you have 3 choices: 1. Don't map (or somewhat guess it) - Roomba. 2. Map first, then clean - Neato. 3. Map as you clean - Dyson 360 Eye. This is obviously a crude explanation but they get harder going down the list. If you are interested to know more, read up on SLAM (Simultaneous Localisation and Mapping) algorithms.

robterrell 12y ago
How is it better than Roomba and Neato -- what's the cleaning algorithm? Does it map the space? etc. Edit: thanks, you answered while I was asking. That's a very interesting use case for SLAM. Does it re-map every use (i.e. today there's toys on the floor, but tomorrow there's not)? Does it store the point cloud or convert it to something more manageable? Are you licensing someone's SLAM library or did you roll your own?
- akramhussein 12y ago
  Unfortunately I'm not really able to go in to much more details, but this video explains what's public knowledge - https://www.youtube.com/watch?v=oguKCHP7jNQ https://www.youtube.com/watch?v=oguKCHP7jNQ Sorry!

berberous 12y ago

It does a pretty good job though: http://www.designboom.com/weblog/images/images_2/jenny/roombaart/roombaart01.jpg http://www.designboom.com/weblog/images/images_2/jenny/roomb... I tried a Neato but it was much more prone to getting stuck (every single time actually). Obviously that will vary depending on your furniture.

IndianAstronaut 12y ago

This is actually better done with a robot team or swarm. There are several 'frontier' algorithms which use basic particle swarm or metaheuristic algorithms to rapidly move across a landscape or frontier.

snowwrestler 12y ago

What's better, a brute force algorithm that is cheap and simple but slow, or a highly optimized system that is expensive and complex and fast? The answer is of course "it depends." Certainly the Roomba could be smarter, faster, and more efficient, but the question is how much more are customers willing to pay for those.