Slashdot Mirror
Xerox PARC Working On Modular Robots
An reader writes "I was watching Discovery Channel Canada last night, and they had a story about modular robotics that is being researched at Xerox PARC. Rather than build a single, large robot, project leader Mark Yim is working on small, autonomous bots that can work together to achieve a desired goal. When many of this bots are linked together, they call the result a polybot. "
doesn't anybody remember the constructicon devastator... the first polybot
Sounds a lot like those transformer toys where you assemble many small robots to get one big robot. It would be funny to have those small bots assemble to make car or trucks, or aircrafts.
Will Bandai or another of the compagnies that built those toys sue those guys? I suppose someone has already patented the idea of Robot changing into transportation mean.
Although you may not realize it, Microsoft apps also tend to be many smaller pieces linked into a whole. Everything is a COM object. MSXML is an object that can parse XML, mshtml handles HTML. Word is a container for a bunch of COM objects (the word processor, the spell checker, the smart tags, etc), Excel is a container for a bunch of COM objects, IE is a container, and so on. Just because each one of these components is not a separate program does not mean that they cannot be reused or linked together.
The standard UNIX "method" is to take all those "single-purpose" apps like awk, or sed, or grep, and glue them together with shell script, or perl script, or tcl script, or some other scripting language. The same method applies to Windows as well. You can access COM objects via WSH (using JScript, VBScript, PerlScript, or any other language that has an implementation of a COM interface that can plug into the Windows Scripting Host), and then you can glue those objects all together into a larger whole. What's more, though, is that you can utilize these objects within actual compiled code, as well. C, C++, VB, C#, any language with a CLR target, ASP/ASP+ (in C#, VBScript, JScript, PerlScript, ...) all can reuse thse objects, unlike in the UNIX model wherein a C app would have to fork to spawn a new process and exec something like grep or sed, make a system() call, or "borrow" code from those tools. And in the code-borrowing case, you're limited to the language that the code was written in (and close relatives, as you can use C code in C++ and ObjC for instance).
Obviously, projects like KDE's KOM/KParts architectures borrow heavily from this idea, and succeed well. But these projects have much farther to go before they'll reach the level of binary reuse and interoperability that Windows has had for years. And yes, I know about CORBA, but compared to COM or KOM/KParts, or even Mozilla's XPCOM, it's cludgy, bloated, and nasty, and is a much heavier paradigm (for instance, it requires an object broker).
This project has been going on for some time now, it's been well publicized in the robotics community and is certainly an interesting offshoot of robotics. I saw a presentation at ICRA (International Conference for Robotics and Automation) last year. Very cool.
It strikes me as rather similar in approach to some of the nano-bot discussions I've read. I.e. build an "assembler" and use that to build other small pieces that can fit together modularly to do what you want. (that's kind of rough, but you get the idea) Interesting parallel in totally different research worlds, although modularity is hardly novel..
The approach does have it's challenges however, the number of independent modules arguably makes the complexity much higher per resultant functionality. A simpler robot could have achieved some of those configurations, and probably more efficiently (power,weight,computational, etc..). Similarly, most robots are rather specifically designed with some task in mind, making general purpose robots is an astoundingly difficult task because of the widely varying requirements of the physical world between different tasks.
I don't mean to bash, actually - I fully support this avenue of research and it's darn cool! Wish I had one..
Brett
UCSD Computer Vision and Robotics Lab Grad Student
Here's the CVRR web page if you're link-happy (no goats).
__ No registration required to read this message. They did it in the Matrix.
This research project is repackaged swarm robotics. Swarm robots have been around for years. The main problem with swarms is getting the power and leverage to manipulate large objects as the swarm is only as strong as it's weakest link. The main benefit is that they tend to be more fault tolerant than monolithic robots.
See:
Robotics portal
Swarm robotics google search
CMU Robotics Institute
Plus, based on their recent comments on IP, they seem like a pretty cool company.
shut up man
In one of Asimov's books about robots. Admittedly this was a client/server type robot, rather than a peer to peer one.
However, unless someone can prove otherwise, it looks like the original idea was Asimov's.
Remember that Bill Joy article speculating that humanity would be destoryed by a grey-slime of nanobots? Well, this is the type of work that's going to make them - imagine zillions of these things at nano-scale, each with it's own 'soft' (i.e using natural principals rather than heuristics) artificial intelligence all working together to provide a giant hive mind.
Asimov's laws of robotics seem ridiculous in this context.
Legobots.
There's a bit of a trademark issue, but still..
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
Robot football could be a good inspirational line of thinking for research in polybots. We have several competitions going into robot football. These games will probably contribute to the system design of polybots. The idea here is to have many (more or less) identical robots, who by working together can reach some sort of goal.
b al l/main.php
a ll .html
One of the things that have come from robot football events is that the individual robots need some sort of collaborative software to coordinate their efforts. So a polybot would proably benefit from a singe strong computer brain, distributing commands to the individual robots. Other approaches (manly in the nanobot research) move towards a simple set of rules that combined provides the desired effect.
Robot football can easily be found through the search engines, but here are some links:
Competition:
http://www.tech.plym.ac.uk/robofoot/
http://www-lce.eng.cam.ac.uk/projects/robotfoot
Some research:
http://www.cci.cse.dmu.ac.uk/cci/Projects/footb
-:) Oh no - not again.
www.rednebula.com
Rodney Brooks published a similar idea: Fast Cheap, and Out of control in 1989. (This is different than the movie.)
I particularly like a related, albeit less useful, plan of sending two robots to the moon, each covered in logos of various large companies. The plan was to finance the trip by selling the pictures the robots would take of each other.
I'm not the kind of OSS radical that says MS should open its source code, but it definitely should ship documentation of these tools with its OSes.
I found these guys searching for modular robots a few months ago. The ideas behind their stuff are very similar if on a larger scale. The videos seem fake though. Anyone know anything more about them or have any insight into thier validity?
It would be wonderful if we had some polybots to aid with the search and rescue effort going on at the wtc RIGHT NOW. It seems like we should be able to design small, tenacious bots that could go into the hostile, challenging terrain of the wreckage and find people below the surface. Each bot could sport one or more cameras and be driven remotely by rescue teams on the surface. Perhaps they could use gps and/or each other's relative location to pinpoint the location of folks they find in three dimensions.
+1 Horrible Pun
My only political goal is to see to it that no political party achieves its goals.