Intel combines Robots, WLANs, and Linux

An anonymous reader writes "This article by a researcher in Intel's Emerging Platforms Lab details some of Intel's current research into wireless, mobile robotics technology. A key goal of the effort, according to the article, is to efficiently combine the two technologies -- mobile robotics and wireless networks -- so that mobile robots can serve as gateways into wireless sensor networks. The Intel project is providing robotics researchers with a robotics development package that includes standardized silicon, a Linux-based open-source operating system, and open-source software drivers for robotics applications. Additionally, Intel has released a test version of a technical library for building Bayesian networks, which will help advance the ability of robots to navigate their environments, and pilot systems based on Intel's open-source packages are already being deployed in a variety of flexible environments in agricultural, security, and military applications."

Comments

Thought process

[worst_name_ever]

I imagine their thought process went something like this:

"Robots are cool. Wireless networking is cool. Linux is cool. So logically, wireless Linux robots would be the coolest thing ever!!!"

The only downside I can forsee is that imagining a Beowulf cluster of those might lead to a Matrix-esque apocalypse for us outmoded carbon units, which would be less cool.

Good idea!

[inaeldi]

Now we can provide Internet access to nuclear fallout regions.

Ok, ok, I'm sure there are plenty of good reasons for this, but I still like my idea more. I want to play my UT2003 after a nuke attack dammit!

Use?

[Dashmon]

Woohoo - run Apache and OpenOffice on your housekeeper! :P

Seriously, I wonder what use this... you don't need *mobile* network gateways that actually *think*, do you?

Re:Use?

[larien]

You want to be able to control your robots without trailing wires and receive telemetry/feedback; why not use consumer technology (Wifi) which is becoming standard on laptops and becoming more common? Given the choice would you rather (a) install a Wifi hub in your home to control robots which can also be used as a computing network or (b) install something proprietary (IR? some form of radio?) which is incompatible with you PC?

As for linux, well, you can strip it down and work with it a lot more easily than windows, I'd imagine, so it's a natural choice. The open source nature is also ideal for research.

Re:Use?

[worst_name_ever]

Seriously, I wonder what use this... you don't need *mobile* network gateways that actually *think*, do you?

Yeah, really - and if they're not programmed with the Three Laws Of Robotics, they might rat you out to your ISP for running a NAT gateway...

Is it linux based or RT linux based

[Angry White Guy]

I thought that robotics apps usually picked RT Linux for their core? Does RT linux still offer additional benefits to robotics?

Re:Is it linux based or RT linux based

[larien]

It depends what you're doing, I guess; if any of the tasks are time-critical (i.e. missing a command by a few milliseconds is a problem) then you want RT linux. It may well be that the time tolerances are such that you don't care about a few ms here and there.

RT operating systems (whether linux or something else) will always have a place in certain tasks. It says something for linux that it can be modified to suit different tasks.

What happened to Windows for Robots?

[ites]

Didn't Microsoft release a "Windows for Robots" OS some time ago? I seem to remember that no-one would buy robots based on RoboWin because they only ran for three hours before needing a recharge. So Microsoft had to build their own Robots and sell them to people who never actually used them but thought they were cool 'coz they could read Excel documents. And then Sony brought out their range of household robots running on PalmOS, which was cool because the robots could recognize script and you could give them to your mother to use an she'd never call for help. But IIRC the final straw was that virus that infected every WinBot and turned it into a homicidal home-recipe machine, producing endless and ultimately fatal lunches of Belgian Waffles with corn syrup.

Uh... I'm sorry. I must stop with the blue pills. Does any company on earth (except MS and Nokia and Palm) bring out a new device that does _not_ run Linux?

Re:What happened to Windows for Robots?

[Anonymous Coward]

Now I'm really glad I took the red pill.

The Three Laws of Robitcs

[march]

The Three Laws of Robotics....

1. A robot may not install Windows products, or, through inaction, allow a Windows products to be installed.
2. A robot must obey the orders set forth in the GPL except where such orders would conflict with the First Law.
3. A robot must protect the open source initiative so long as such protection does not conflict with the First or Second Law.

:-)

Re:The Three Laws of Robitcs

[Capt. DrunkenBum]

You forgot:

4. Profit.

Re:The Three Laws of Robitcs

[Anonymous Coward]

4. A robot must always refer to its operating system as "GNU/Linux" and not "Linux"; failure to comply will activate a nonrevocable self-destruct sequence (aka the "RMS", or "Robotic Mandatory Selfdestruct").

Looks interesting

[Timesprout]

I'm wondering though since they are not actually all that interested in the physical capabilities of the robots, concentrating instead on group intelligence why do they actually build the robots?(OK I know geeks and their toys).

Surely the robot controller code could be emulated purely in software to determine how the robot will respond, a much more sophisticated version of the recent Java battle bots if you will.

Is there some benefit to physically building the robot when researching group intelligence ?

Re:Looks interesting

[Anonymous Coward]

Is there some benefit to physically building the robot when researching group intelligence?

Yes - you get to see what an actual group of robots do, instead of what some simulation says they might do.

Re:Looks interesting

[Anonymous Coward]

My undergraduate research advisor, a leading researcher in the field of non-linear controls applied to robotic problems, had a saying which will answer your question pretty well: "Simulations are doomed to succeed."

Here at Carnegie Mellon, most of our mobile robots used for research are controlled in one of two ways:
1) a pc/104 stack has been added onto the robot.
2) some poor graduate student's laptop has been tied down on top of the robot.

Putting laptops on top of robots is a nice hackish solution, but it's not really optimal if you're not doing massive computations. pc/104 stacks are neat, but, for a very small robot, they can take up a lot of room. Combining several functions of the pc/104 stack into one card is key here. Pretty much every robot that goes scurrying through the halls of Newell Simon hall has an orinoco wireless card sticking out of it.

We already do that.

[grub]

We have a merging of bio-mass, networking and telephone services where I work: when people think the network is getting flaky, my phone starts to ring.

WLANs and Linux...

[jkrise]

Intel isn't very clear with Linux on Centrino, their WLAN offering - funny to see them offering exotic stuff on Linux. Intel seems to resemble MS more and more these days. Time to clip wings, perhaps?

Research cool, not consumer cool

[JasonFleischer]

So everyone's first instinct is to make some disparaging remark about how combining buzzwords --> profit!!!! I don't think that's what this is about. This has nothing to do with consumers, and presumably therefore little in the way of profit for Intel. This is about adapting a consumer technology for a research area in a highly useful way.

Mobile robotics has been hard hit recently, when one of the main companies making robots (Nomadic Technologies) was acquired by 3COM in 2000 for their wireless networking technologies. Obviously 3COM had no interest in research robots that cost thousands but sell only hundreds of units. Since then there's been a bit of a hole in the market for somebody to sell prepackaged wireless robot stuff to researchers, especially those that work in the software/AI/algorithms end of things don't care to spend effort developing hardware.

Intel's Centrino blah blah is supposed to make connected mobile computing easy and increase battery life. Well guess what drives my ancient Nomad Scout robot? A laptop connected to the robot's power supply in a hack'd fashion, communicating using a USB-driven RF link. This platform could have saved a couple of months development of things which aren't exactly shining examples of engineering anyway.

This hardware isn't the sort of thing that the average /.'er is going to drool over and plot how to justify purchasing it to their spouse. But it is very useful for the couple of thousand mobile robotics researchers around the planet.

Re:Research cool, not consumer cool

[Evil Adrian]

Come on now, "spouse"? Slashdotters can't even get dates, how the hell are we supposed to get married??

Re:Research cool, not consumer cool

[Hayzeus]

Since then there's been a bit of a hole in the market for somebody to sell prepackaged wireless robot stuff to researchers, especially those that work in the software/AI/algorithms end of things don't care to spend effort developing hardware.

• ActivMedia
• iRobot
• Arrick Technologies Trilobot (not WLAN-based, tho -- uses an OCR-LAWN II last I checked)
• a host of others

Nomads seem to have had a lot of what little market share there was for research robots, but there are a number of other companies that offer similar products. I'm wondering, in fact, if the market isn't really too crowded.

Of course, any truly MANLY researcher would build their own.

Free robot Mind is available

[Mentifex]

Mind.Forth is free AI source code for a robot AI Mind in Win32Forth.

Mind-1.1 in JavaScript is the AI Tutorial version of the same robot Mind software for true artificial intelligence.

AI4U: Mind-1.1 Programmer's Manual is the textbook of artificial intelligence describing the Robot Mind-1.1 software of the Mentifex AI project as listed in the Free Software Donation Directory.

Technological Singularity is happening right now.

Re:Free robot Mind is available

[guacamolefoo]

I had started doing some research into AI for a variety of reasons recently, not least because I was interested in what intelligence is and whether true AI could be created. It led me into issues involving software, hardware, and differences in basic approaches to AI. I went down a lot of roads - parallel distributed processing, Marvin Minsky, etc. Fascinating stuff.

I thought I had an interesting idea when I started to think about an AI developed as open source versus a closed-source AI developed by a software firm. Hah. Nothing new under the sun, evidently. Still thinking about the possibilities of an AI run through a distributed computing effort. The processing power is there, but the coordination efforts would be nightmarish.

In any case, this Intel research seems to be leading down an interesting path, although robotics is definitely not AI. Some convergence is likely, but anytime I see someone pursuing a project like this with (more or less) commodity hardware and free tools, I am encouraged.

GF.

Re:Free robot Mind is available

[Leers]

I have been thinking lately that to create a true AI, one would need to give it a robotic interface to our world. Perhaps even raise it like a child. This would be so we would have enough shared experiences with it that we could communicate with it enough to know it was alive. I guess computer AIs could serf the internet to get access to people, but if all it did was that we would just have an AI geek. ;)

Re:Free robot Mind is available

[guacamolefoo]

I have been thinking lately that to create a true AI, one would need to give it a robotic interface to our world. Perhaps even raise it like a child. This would be so we would have enough shared experiences with it that we could communicate with it enough to know it was alive. I guess computer AIs could serf the internet to get access to people, but if all it did was that we would just have an AI geek. ;)

How the robot raised by geeks would turn out would be interesting. Nevertheless, you have an interesting point about "raising" a robot. One of the reasons that humans have prolonged childhoods is probably that there is an advantage to the genes from a natural selection standpoint.

With an AI, you would need to do this only once, however, or only a relatively limited number of times. Then copy the thing to many bodies.

There is an extant effort to codify "common sense" rules for dealing with the real world, btw. There is a company out there that is collecting into a database a variety of situations and trying to use that as a basis for creating at the least a very "sensible" program. While it may not "think" in a pure sense (do we?), it may at least act prudently (as a general rule, don't jump off cliffs for instance).

I sometimes think that any element of a "man-like" AI (which is extremely limiting) must contain some sort of feedback/self-programming process. It sort of makes me think of collaborative programming in an OS model -- try, revise, try again. The goal orientation (ego) and learning process (self-programming) seem to me to be good places to start.

As I mentioned with the Intel program, the cool thing is the OS software and the (relatively speaking) commodity hardware. Wireless is simply another cool part of the hardware set. A really, really smart AI could probably even use the Wi-fi for radar as well as communication.

On a final note, the gist of many AI projects I have seen (and the Intel one seems to be leaning this way) is on group action and intelligence. A "hive mind" as opposed to self-contained individual members. Wi-fi would be beaucoup for doing this on the cheap (blah blah blah Bluetooth blah blah blah).

GF.

Re:Free robot Mind is available

[zephc]

you have a LOT of -1s in your list of posts. You would probably do better if you didn't spam the board with the same links over and over. You should just offer people good keywords to google for. Sorry to be blunt, but the Novamente and SIAI stuff is a lot more sophisticated that what you have, and I suggest you read a lot of that stuff.

offloading the brain

[bloosqr]

I've been thinking about this for a while. The whole thing about stamp/robots is that they run on CPU's that most of
us are no longer used to. What would be excellent is if you could "offload" the brains via high speed wireless. 45mb wireless to high speed processor(s) I would think offer a much different version of robot programming than the current set. I would also think this would use less power than lugging a laptop around on the robot like the kit you can buy at compusa (let alone minaturization possibilities)

-avi

Re:offloading the brain

[Anonymous Coward]

I somehow doubt our brain works in a digital boolean pattern, so I don't think there will be any uploading of brains anytime soon.

Re:offloading the brain

[The Kryptonian]

You have a really good point here - historically one of the biggest problems in robotics is that robots usually would benefit greatly from having bigger brains than they could actually carry. Making use of a wireless connection would certainly solve that problem. It would also create the possibility of "robot hives", where one master brain of very high computing capacity could control a small fleet of robot body surrogates.

This concept has been explored in great detail in science fiction; those of us who actually read the stuff are already aware of the potential. Hurray for Intel! Real, practical humanoid robots can't be far off now (by which comment I most definitely imply that I don't think they're there yet).

Re:offloading the brain

[bloosqr]

Thats exactly what i was thinking as well.. hive minds and robots. I keep wanting to get into robot building but the whole pic/stamp mentality just doesn't do anything for me. (Perhaps because i'm a software geek rather than a hardware geek). But on the one hand we have the AI kids building nns or even brute forcing ala kasparaov sucking up huge amount of CPU cycles and these dinky robots that do "light detection" and a few 100 assembly commands and need to be low power. No power problems via wireless :). Glue this into that other slashdot article from a week or so ago about "power transmission" via wireless and you'd really be set ;)

-avi

Kill them now!

[mental_telepathy]

Sorry, just finished watching the preview for T3

Intel Stayton boards

[CTho9305]

These boards are really cool (Stayton is used on the CMU TagBots). We (CMU Robotics Club) normally use a board designed by robotics club members to control robots, but they are based on 20MHz PICs, and don't have and wireless support (at least presently). When combined with the Intel board, however, the big processing can be done there, and the Cerebellum can just be used as a smart motor driver and sensor interface board.

This lets the robots run more complex code and communicate with each other wirelessly. Intel has provided CMU with enough boards for a LOT of cool projects.

Some of this has been done before

[Steve1952]

The idea of using robots to communicate to wireless sensors has been around for a while. See, for example, USPTO patent application 20020173877.

Re:Some of this has been done before

[hklingon]

Whoa! This is cool. I'm a college student and I might have prior art on this one. In 1999, 2000 I developed a 6 legged walker that had 2.4ghz video back to one desktop computer, and a 2400 baud serial link back to another. The robot had tilt, temp, range, IR, hall-effect etc. sensors on board, and the 2400 baud link went to another computer that was the real brains.. the video processing happened on one and the electronics on the other. All wireless. It was a mobile, wireless sensor platform for students to experiment on, etc. Before that, I was modifying RF remote controls to transmit information... I wonder if this patent has a bounty . . . ... .. ... bwa!

5-step program.

[Anonymous Coward]

Step 1: Build robot.
Step 2: Appease linux zealots and put redhat on it.
Step 3: Wirelessly network said robot.
Step 4: ???
Step 5: Profit!!!

Half-Life Skills

[bsd-mon]

Finally we'll be able to apply our escape techniques in a real life nuclear disaster!

Woo hoo but not new

[Anonymous Coward]

Irobot has done it and sold it, http://www.irobot.com
I too plot to do it, though I haven't had much time while raising a few organic children these past few years :) you can use a notebook motherboard for example, but activewire (usb micro controller) Atmel avr and a few other technologies should be combined with laptop like mainboards running wifi links to the network for control / instruction loading..

gimi 2mil and I would make all kinds of units like this while showing the children of the world how too. Personally this is where things will go, but before we get to far we need AI the likes of COG and better. Bots with brains that run on linux clusters :)
COG is a project at MIT,, I forgot the link,, but its Unix driven most massive civilian AI I have ever seen.

cy83rpunk2002@netscape.net

Intel + WLAN + Linux == unsupported

[SilverSun]

I find this article very interesting, especially since I just recieved an e-mail from Intel custumer support, telling me that my centrino based laptop's wireless network adapter (Pro100) is not supported with linux. After asking when I can expect drivers, I got a friendly e-mail, telling me that linux support is not planned in the near future....

I think this is symptomatic for many big companies. The left hand doesn't know what the right hand does...

Cheers, Peter

Now to really own the RIAA

[illumina us]

You can take your bot and have it sit right outside of teh RIAA buildings and d/l their songs and pwn their networks, or you could just run CS IRL on it. Though providing networks to nuclear fallout regions is still cool.

I've been doing this for 3 years now!

[MrJerryNormandinSir]

I run redhat on a TriM systems embedded controller and I use a Siemens 802.11 adapter. Gee...
If Intel tries to patent the idea it will piss me off. I did not feel this was worthy of a patent.
I'm driving servos with a pontech controller, I've monitoring Analog ports, I'm processing ultrasonic
ranging data. I've got some of my robots at
www.nfnnet.org

Short Ciruit..

[Anonymous Coward]

So what, they took Johnny Five and put a wireless network card in him. Wait, he already had wireless communications, so all they had to do was upgrade the firmware...

Intel, Linux and WLAN, yeah right

[rjkm]

Some others already commented on the situation with lacking Centrino (power saving and WLAN) support. I did not even get any answer from their "customer support". I'll only buy Transmeta in the future. I just got a Libretto L5 and it is MUCh better regarding heat and noise (none besides the harddisk).

But I also still have the Centrino notebook (Samsung X10). Does anybody know if there are Linux supported mini-PCI WLAN cards which I could use instead of the Intel card?

Control and simulation tools

[rtv]

Simulation is used extensively. Several of the researchers mentioned in the article use the (GPL'ed - try 'em yourself) Player/Stage tools for robot control and simulation.

But using real robots is a vital reality check.

First "SkyNet" reference!!!

[t0ny]

Soon, Intel will merge with CyberDyne Systems Corporation to take advantage of some of their bleeding-edge AI research.

Soon after, they will receive a government contract to create a unified defense infrastructure, and merge it with their wireless networked robots. Soon after, it will determine that humans are outdated.

Let the games begin!

Re:First "SkyNet" reference!!!

[t0ny]

um, moron moderators, how is it redundant if its the first...

Re:Russia?

[evilviper]

Sometimes moderators see the humor, sometimes they don't. I guess not too many people caught-on to this one.

Sounds so good, we've been doing this, too!

[campgod]

Check out the program and video from MIT's MASLab 2003. Our robot's use a very similar setup. Geode+Orinico+RedHat. Next year we're moving to Eden+WindowsXP Embedded :) Can you guess why?

From the program:

The electrical components used in MASLab are quite different from other
contests. At the heart of each team's robot is the "Geode," a 300 MHz
x86-compatible processor. With 256 MB of RAM, a 6GB hard disk, wireless
networking, and a full complement of peripherals. This PC runs an
unmodified installation of RedHat Linux.

The Geode itself cannot control motors nor interface with sensors, so
the MASLab staff designed and manufactured a robotics controller board.
We call the controller the "ORC", for "Our Robotics Controller". This
board serves as a slave to the Geode, executing simple commands under
the direction of a program running on the Geode. The Geode and ORC
communicate over an RS-232 serial link. It contains an LCD display,
support for four 12V motors, integrated battery charger, and power
regulation. The ORC board features several Cyprus Microsystems
Programmable System-on-a-Chip (PSoC) parts connected by a serial bus.
The PSoCs are configured to support three servos and an array of analog
and digital sensors including ultrasound and optical encoders.

While the usual assortment of robotics sensors are available (ultrasound
range finders, infrared range finders, momentary buttons), MASLab
additionally includes a web camera. This color camera has a resolution
of 640x480 and serves as most robots' primary sensor, scanning the
playing field looking for targets and scoring areas.

Power is the problem

[Anonymous Coward]

The key hurdle to everything in the robot world remains power. I'm not talking processing power, but power sources and consumption. General processors like P4s and XScale consumer have poor processing / power consumption ratios, making them poor canidates for robot designs. Networked microcontrollers (or embedded CPUs like Hitachi SH series) coupled with FPGAs or ASICs are still the best way to go. Better yet embedded microprocessors with synthesizable cores. But thats not Intels way of business.

Re:My thoughts on this idea

[RandomUsername99]

Aparently we at "harvard" are not supposed to use spellcheckers... I did not recieve such a memo. Also, with this particular subject, you probably would want to lie about being from the college up the street, MIT. Not only do they have a much better robotics program, It is much easier to spell.