Slashdot Mirror
A Standardized OS For Robots
Hugh Pickens writes "The New Scientist reports that at present, all robot software is designed uniquely, even for parts common to all robots but that could be about to change as roboticists have begun to think about what robots have in common and what aspects of their construction can be standardized, resulting in a basic operating system everyone can use. 'It's easier to build everything from the ground up right now because each team's requirements are so different,' says Anne-Marie Bourcier of Aldebaran Robotics but Bourcier sees this changing if robotics advances in a manner similar to personal computing where a common operating system allowed programmers without detailed knowledge of the underlying hardware and file systems to build new applications and build on the work of others. 'Robotics is at the stage where personal computing was about 30 years ago,' says Chad Jenkins of Brown University. 'But at some point we have to come together to use the same resources.' This desire has its roots in frustration, says Brian Gerkey of the robotics research firm Willow Garage. If someone is studying object recognition, they want to design better object-recognition algorithms, not write code to control the robot's wheels. "You know that those things have been done before, probably better," says Gerkey, who hopes to one day see a robot "app store" where a person could download a program for their robot and have it work as easily as an iPhone app."
For a few years, MOOS has been developed at Oxford University, to separate low-level control issues from high-level issues. It runs on OS X, Linux, and Windows.
There's also IvP, an autonomous vehicle control system that gets uses MOOS to abstract away the low-level details of controlling the particular vehicle on which it's running.
I built a robot for my school (City College of New York) for the Intelligent Ground Vehicle competition, and we used an open source programming environment called player/stage from Carnegie Mellon, which already has a huge number of libraries, and has a standardized driver format for sensors and other devices. It gives stuff like abstraction of motion - in other words, you draw a map, and have your navigation algorithm try to go around the map and it gives you back simulated data from your sonar, scanning laser, GPS, etc... It did save us a huge amount of time... instead of figuring out how to construct the data flow for sonar sensors we could just drop their packets in a queue... which let us move on to openCV - again, another existing open source project that already is well developed and gets you 75% of the way there. We used it for the drive system, and the position control had all sorts of generic modes for tank mode vs car mode, etc. It even starts you with some algorithms like "laser obstacle avoid", i.e., use the scanning laser and try to get around a maze. Drivers are typically in C, other stuff was in C++. And yes, it runs on Linux ;).
Safety was taken really seriously. When testing, you'd set up a tripwire fence that'd shut the robot down if it were jiggled. Every single person inside that fence had to be holding a deadman switch - let go and the robot shuts down. When I saw one of those suckers casually drag around a 500lb steel table that hadn't been bolted to the floor I got respect fast. Thankfully nobody got hurt, but at a customer site once, a badly-maintained spot welder managed to attach itself to a truck body on an assembly line. The robot kept right on going and literally threw the truck body into the aisle.
Liability's kinda critical for something like that. For unarmed, relatively weak research robots, a common platform makes sense. For higher-powered industrial robotics, this ain't gonna fly.
PHEM - party like it's 1997-2003!