Slashdot Mirror
Open Robotics Debuts at Penguicon 3.0
thgreatoz writes "While attending Penguicon 3.0 in Novi, MI, I came across an interesting project. Matt Switlik of Swittech aims to do for robotics what the GPL did for Open Source Software - a completely open robotics platform. Dubbed the Open Robotics Peripheral Platform, or O.R.P.P, Switlik and his partner Jason Hunt have taken a completely modular approach to robotics, with the goal of making robot development as easy as homegrowing a PC. Will we see fleets of ORPP robots plowing our streets and mowing our lawns in the future?"
That sounds good; however, if there's going to be an open robotics platform, does that apply to the actual software powering them? They don't necessarily have to have the exact same software in O.R.P.P. compliant robots, but just the same "kernel", so that extra code could just be modular: added in when needed.
Debugging? Klingons do not debug. Bugs are good for building character in the user.
Will the USA become a place where the only jobs needed will be thought based. No more jobs where a person is needed to do a repetitious task over and over? Will the next outsorcing be not out of the USA, but from human labour to robots?
I see so many problems here. What will people do for a living??
I don't want this to sound like trolling, but it will. There are enough people out there who are not made for work which requires too much thought. Not everyone can pass Chemistry 101. Some people require the factory jobs to make enough money to buy a house, and live a life. If we start lowering the value of those jobs, we will be shoving a whole class of people into poverty.
I also can't help but think of the horror of the next war we face. No more "human life lost", instead we'll send drone airplanes and robots to do the fighting. Mr and Mrs Redstate will no longer have to reconsider if a war is just when their child is killed ("Was it worth it?"). I wonder if we would have burned all of Vietnam down if we did not have to send any Americans, if we only had to send robots. We could declare the area too unsafe and keep the reporters out.
Rosco: "If brains were gunpowder, Enos couldn't blow his nose."
You would be very suprised at the power of a 286. It could easily run a robot. And if you have the math co-processor, you could probably program some AI. Now the CGA or EGA monitors sucked, and the sound sucked. But at its very basic level, it is more powerful that you think.
I bet you could control multiple motors with a 286. Simple on/off commands for moving N/E/S/W.
Rosco: "If brains were gunpowder, Enos couldn't blow his nose."
Ironically enough, about the same time that Weird Al's song "The White Stuff" came out (parody of "The Right Stuff", about the middle of Oreo cookies), I was reading a novel from a series my dad was into, The Destroyer (don't recall the exact issue), about a politician who was mixing trace amounts of cocain in Oreo cookies and giving them away at campaign meetings to get people addicted and "feeling good" at his meetings to entice them to vote.
Three cheers for the DoubleStuff!
Regarding the original post, I'd love to see a build-it-yourself standard akin to building your own PC to do various robotic tasks. I'm guessing the technology would be prohibitively expensive for a while, along with gov't regulations to make sure you don't make a robot with motion sensors chase your neighbors down and kill them or something ridiculous.
my geeklog
Good points. It's incredibly hard to make truly self-guiding vehicles, especially land-based ones where you have far more obstacles to avoid than air or sea. It's a combination of sensor input and AI.
To expand on this: read up on the DARPA Grand Challenge. The goal is to design and build a completely autonomous vehicle capable of navigating various kinds of terrain and obstables over long distances, using only a set of waypoint coordinates given to each team two hours prior to the competition. DARPA hopes to bootstrap the research into autonomous military vehicles, starting with logistics and supply and working up to battlefield weapons. GPS is permitted, but no remote control whatsoever (aside from a remote kill installed by DARPA officials), and you only have a corridor a few yards wide for maneuvering.
Last year, the course was out in Nevada, around 160 miles in length and with a time limit of 10 hours - including obstable navigation, you would expect an average speed of around 20 miles per hour. Out of the bunch of universities and companies that submitted various entries, the farthest anyone made it was only a few (ie, less than 10) miles; most of them stopped or died within sight of the starting line. It's that tough.
This year promises to show vast improvement - some of the demo videos show pretty decent speeds - but the proof is in the pudding.
P.S. Most of that was recalled from memory - I apologize for any errors of fact.
Unfortunately, useful robotics is a highly proprietary market... far worse than the dark Unix years. Nobody who has the hardware wants "open robotics" and worse tend to tie their proprietary hardware to even more proproietary software! The only way to do something like this is "gaurilla" type projects like this one, but unfortunately you need somebody with experience to build hardware.
The main thing you could try to do is come up with Open robot specs using common industry standard pieces as much as possible. Perhaps thru persuasion to support the "open platform" manufactures would open up their software and prices would come down.. therein lies the problem... cheap robots like mindstorms and robosapien just aren't useful...and the parts to make something useful are a thousand $$ leap at this point....
In another life I worked at JPL in robotics and still keep in touch with friends there. They have been struggling for many years to open source much of the robotics software both inside NASA and through all the contract money they send out to universities. Of course the beauracrats are making things difficult. Last I heard there was trouble due to the ban on export of certain technologies and part of it is NASA and University IP lawyers worried about loosing control of that one golden nuggest that might come out of all that $$. In the meantime, every NASA center and university has re-written the same basic robot kinematics, navigation, and machine vision algorithms a dozen times in multiple languages. Many of the guys who wrote a lot of this stuff 15 years ago are now managers and trying to push colaboration and open source using their projects, but it is still hard to move lawyers.
It is clear that computational intelligence will emerge through emergence and therefore be somewhat resilient to full analysis and control. That much is almost certain. For instance, how do you encode the three laws into heirarchical neural nets?
However, it is also true that a fully functional robotic system will contain more than simply a brain. It will, for instance, contain a power supply of some sort as well as actuators and other mechanical parts for movement.
Just as modern automobiles sometimes include speedlimiters to override the brain (the human) when he/she decides to push the vehicle into a dangerous speed range, so too could deterministic software detect and shutdown the emergent intelligent systems of a robot in situations considered dangerous.
Consider a robot that has effectively two different brains.
(Brain 1) This is what we normally consider a brain - the robot's personality, decision making system, memories and whatnot... Build as you like
(Brain 2) Akin to a speedlimiter, a programmed system that watches what the robot does and takes action to shut it down if those actions go outside of the boundaries of acceptibility.
Clearly, it would be difficult to build (brain 2) in such a way as to catch all foul play by (brain 1), because (brain 2) would be restricted by deterministic techniques (e.g. traditional programming techniques). However, encoding Assimov's three laws in some form in (brain 2) is certainly conceivable.
Just food for thought
The reason that it can be true that 1+1 > 2 is that very peculiar nonzero value of the + operator