Have you ever dreamed of building your own Linux-based droid [that could] roam around your home autonomously, intelligently obeying your commands?” asks Dafydd Walters. “You may now be able to finally fulfill your dream.” Walters founded and leads the Open Automaton Project (OAP), which aims to help enthusiasts assemble an intelligent mobile robot with stereo vision and state-of-the-art PC mainboard… technology — for about the cost of a good PC.
Unlike the RETF and OROCOS projects, which focus on devising open standards for robot hardware-software interfaces, OAP is all about implementation. OAP's objective is to create a reference design that bridges the gap between sophisticated robots used by researchers and the “small, rug-roving” critters typically assembled by hobbyists.
OAP's SourceForge-hosted Website provides circuit schematics, source code, and documentation for free download under an Open Source license, “to enable robotics enthusiasts to assemble their own intelligent mobile robot,” says Walters.
Specific project goals include:
- Design a coherent set of modular components (hardware and software) that conform to standards (where possible)
- Minimize cost to $1,500 – $2,000, about the cost of a good PC
- Develop a low-cost real-time vision system for use as the primary spatial sensor
“To keep the project within the financial reach of hobbyists, costs are minimized by using readily available “off-the-shelf” consumer-grade hardware components where possible, and designing custom electronic subsystems where ready-made components are either not readily available or are too expensive,” notes Walters. “For example, the vision system consists of readily available consumer FireWire webcams mounted on a Pan and Tilt head made using a pair of standard radio-control hobby servos.”
According to Walters, the advent of consumer-grade webcams and motherboards supporting FireWire should make low-cost stereo vision possible. The project chose the ADS Pyro 1394 WebCam, but any IIDC-compliant (a.k.a. DCAM) digital camera should work.
Twelve sonar detectors (Devantech SRF04s) controlled by the
oap-sonar commandline utility augment OAP's vision. One unit pans and swivels with OAP's eyes, while eight face outward around the circumference of the base to detect obstacles. Three downward-facing detectors listen for precipices.
An infrared sensor mounted on OAP's panning and swiveling head and “tuned to the human body's infrared emissivity” helps OAP identify humans.
OAP pans and tilts its head through an I2C interface controlled by the
oap-head command line utility. Two main DC drive motors provide mobility through the
oap-motor program. Wheel encoders capture odometry. A power management module monitors battery levels and can communicate via infrared with OAP's fixed docking/recharging station.
OAP's human robot interface (HRI) includes:
- Speech output
- Minimalist robot-mounted LCD and keypad
- Wireless and infrared remote controls
Speech recognition/input will not be attempted, initially, according to Walters.
For more detail: Article Meet OAP — an open robot reference design project