When Stanford University won the DARPA Grand Challenge in ’05 the team made clear their philosophy that autonomous navigation should be treated as a software problem. While other teams spent their resources on building custom hardware, Dr. Sebastian Thrun simply sent his Stanford team’s Volkswagen Touareg off to get converted to drive-by-wire using proven and commoditized technologies. When he got it back, all that was left was to plug in his computers and start coding.
Software is a cheap medium on which to try different things without fear of cutting something too big or too small or accidentally reversing the polarities of some highly sensitive and expensive electronic device. No. If something isn’t built correctly in software, you just open it up in an editor, fix it, and try it again.
And if you want to build a robot to interact in the world, you’re going to make it small. If you’re going to make it small, you’re going to care about things like computing power and battery life and other technical excise tasks. The solution to this is to utilize the substantial computing power on your desk. When SPC-101C came out last year, I lauded the client-server architecture that delegated the substantial computing requirements to the host computer. The robot plays a sensing and interacting role for the host computer and is what lets your omputer move about in the world.
Surveyor Corporation clearly gets all of this.
They just announced a computer vision platform called Surveyor Stereo Vision System (“SVS”) that sells for a paltry $550. It’s designed to let developers and researchers get straight to the business of stereoscopic vision research. “We realize that in order to help accelerate the adoption of computer vision into the mainstream we have to remove the obstacles that hinder innovation. The last thing a computer vision developer wants to do is waste his time measuring voltage levels,” says Surveyor Corporation founder Howard Gordon.
Surveyor SVS’s straightforward approach can be explained in over-simplified terms as two socket streams sending video through a 802.11g radio. This standard method of transmission allows for easy integration into lots of open source computer vision libraries including OpenCV. To normal people, this just means that the disembodied cameras act like two high-end webcams that can do cool stuff like render the world in a 3D image you can see through red and blue glasses like the kind you wear at the movie theater.
Getting a pair of integrated cameras that can transmit video signals digitally throug 802.11g wireless protocols for $550 is impressive enough, but Surveyor SVS is also a complete mobile robot platform. “Surveyor is providing a great service to the industry by offering high-end technology at prices even a hobbyist can afford. I’m not quite sure how they pull it off, but I’m glad they do,” says Matt Trossen, founder and CEO of Trossen Robotics.
By comparison, Point Grey Research’s Bumblebee Stereoscopic camera sells for $1,500 and requires a firewire connection to a computer.
Disclosure: Our affiliate sponsor Trossen Robotics sells Surveyor Corporation technology.