Useful robots

Okay, what's the point here? When I shared this idea with Jimbo Wales, he wrote, "[you're saying that] we can get more people involved in home robotics if we can think of fun collective projects that will get more people interested. And I think that is 100% right." So, let's design a fun collective learning project for Wikiversity.

I see a nice synergy developing in the future between Wikipedia and home robots. But for now, we need a simple project that everyone can participate in and learn from. I propose that some of us buy cheap wireless hardware add-ons to upgrade the vision and mapping systems of cheap, sturdy robots. The obvious platform is robotic vacuums, because they're cheap, plentiful (around 4 million in US homes), and good for something...but mostly, because they really need our help: they would do a much better job if they could "see" and "learn". But any cheap robot could serve as a platform to allow someone to participate in the project.

People who don't want to buy hardware could work on classification of the images. The longer I spend designing the database for this, the more it seems like a game...a fun game. The 3 kinds of sensors that we're likely to use will detect ultrasound (usually to get a big but fuzzy picture, but they can be targeted too), or a narrow infrared beam (very accurate, but only one spot at a time, not too close, not too far), or a phase-shifted laser rangefinder (very accurate, great range, useful for accurately positioning the robot, but lasers drain batteries fast). So every time you see an image, you get to decide which "weapon" you want to use next...fire the "laser cannon" for best accuracy and drain the batteries a bit, or move around until the object is in the range of the infrared camera and take a few readings, or fire up the ultrasonics and take your best guess whether that fuzzy image up ahead is something you absolutely don't want to hit or suck up...a kitten, a tangle of wires, a sock, a toy. Your mission as a robot-in-training, should you choose to accept it, is to follow an intelligent path, cover all the floors, and find your recharging station whenever you have to, without getting yourself or anyone else killed.

I bet it will play a lot like a game, but this is serious business. This project teaches all about machine vision, and participants could also choose to learn about wikiversity, graphic design, usability, or database management. But if we're successful, robotic vacuums will outsell other vacuums, and we help build a bright, shiny future where every home has a robot in it. Then, as the market grows, competition will force robotic vacuums to evolve more intelligence, sensors, arms, grippers and other features, so that they will eventually perform all kinds of housework.

Even a little success could make a big difference in how useful robotic vacuums are. My iRobot Roomba spent 40 minutes cleaning a hallway that it could have done in 5 minutes the other day if it had been able to keep track of what it had covered and what was left. Judging from the forums over at RoombaReview.com, that's not unusual. Robots need explicit instructions, and people don't like trading one tedious problem...vacuuming, for instance...for another, training and re-training a stupid robot. That's why we need to develop a user community to do the training, and I'm optimistic that we can succeed.


 * Update: the most relevant collaborative work at the moment is going on at SocietyOfRobots.com, and I'm going to ask permission to bring the relevant work over here, particularly the Google Sketchup stuff. - Dan Dank55 17:00, 27 January 2008 (UTC)



Note: Because of the Dec 1 Wikimedia Foundation resolution to work towards a license that's compatible with both GFDL and cc-by-sa, and because some content here migrates in both English and German to de.wikiversity.org, which uses cc-by-sa, and because GFDL does not work well by itself for database apps and images, all content you enter on this page and every subpage is considered dual-licensed under GFDL and cc-by-sa, unless you state otherwise.