ALBUQUERQUE, N.M. Inventors at Sandia National Laboratories have devised a robotic control scheme that relieves operators of tedious and error-prone jobs. The first assignment for the Sandia Modular Architecture for Robotics and Teleoperation, or Smart, is to speed bomb defusing and disposal.
"Smart starts with full autonomy, and then backs off to the degree of autonomy with which a teleoperator is comfortable," said Sandia project leader Phil Bennett. "In tests against even the most highly skilled teleoperators, Smart was able to shave valuable minutes off the time needed to defuse a bomb."
Smart was unveiled at the recent International Association of Bomb Technicians and Investigators here, which drew some 700 law enforcement officers worldwide. Smart was demonstrated on its prototype platform, a Wolverine robot manufactured by Remotec Inc., a subsidiary of Northrop Grumman Corp., Smart's first licensee. Smart is available for licensing by any U.S. ally.
"Since we are a national laboratory we are not in the business of licensing technologies for a profit, but we hate to see technologies that could be useful in non-military applications just sitting on the shelf," said Bennett. Bennett believes Smart-based robots will be useful to emergency response teams in such endeavors as cleaning up chemical spills; for security applications like patrolling perimeters; for space exploration; and for combat engineering tasks such as laying barbed wire, removing mines and assisting the injured.
Since 1986 Sandia has operated a 226-acre Robotic Vehicle Range of rolling hills and open desert, as well as a garage, machine shop, observation tower, and six assembly and test labs.
Practically all of its efforts, however, have hitherto been to support military applications like the Fire Ant, a "tank-killing" robot that can sneak up on enemies and launch a rocket. Smart unites military requirements and civilian applications.
Bob Anderson, who developed the Sandia software, "came up with the idea" for Smart, Bennett said. "He said a lot of the same development effort had to be repeated each time we tackled a new robotic project, whether it was for the military or NASA or some civilian application. So the first objective with Smart was to simplify rapid prototyping from diverse components."
One of the greatest problems in assembling robotic systems from both prototype and off-the-shelf components is controlling all the hardware in such a manner that not only are the design objectives met, but also the overall system is stable under all operating conditions.
Anderson devised and patented for Sandia a master control algorithm for Smart that simplifies the system integration of diverse components manufactured by different suppliers while guaranteeing overall stability regardless of environmental obstacles. "Stackable" software modules permit each robot component or function to be separately enabled without affecting the stability of the overall system.
"The speed with which you can add new components has really got the law enforcement community excited," said Bennett. "A sergeant can look at the problems officers are encountering and add new tools and behaviors to the robot without worrying about system integration issues."
Bennett also predicted that licensees will be able to field new capabilities for their commercial robotic systems just weeks after they are developed, rather than wait months for engineers to resolve system integration issues. For instance, new three-dimensional controllers for video reality applications, such as data "gloves," could enable future operators to "feel" bombs.
"We will have standard driver modules available for all the off-the-shelf joysticks, spaceballs and haptic controllers. We already have modules for all the popular gaming controllers," said Bennett.
In the field, low-level integration issues are aided by the Smart robots' simpler graphical user interface, compared with normal remote controllers. Smart enables the robot itself to make the "how-to" decisions while the operator decides what to do next. For instance, operators often have to move an arm while looking at the camera view upside down and backward, and 2-D video cameras make it difficult to judge distances.
"Sometimes it's like playing a videogame with a seven-lever joystick sitting upside down, with one eye closed, and with your boss looking over your shoulder," said Bennett. "Operators might think they're about to bump an object but they're really three feet away. Or they don't know if the robot will be able to fit between two cars or climb a flight of stairs until they try."
Bennett observed expert FBI teleoperators to determine which robot tasks were frequently encountered. He concluded that Smart's goal should be to free up the human operator from the monotonous and sometimes confusing details of moving joints, changing tools, accurate placement and aiming.
For instance, Smart's "visual targeting" module utilizes two video cameras for 3-D vision, so that teleoperators can click their mouse on a bomb and command the robot arm to extend to that location. This would ordinarily have to be performed by a teleoperator who commands each of the robot's joints to turn a certain amount. "When you use visual targeting, the robot moves all the joints in whatever way is necessary to get the tool in the right place," said Bennett.
Another module, called "reachability," can be invoked before the arm moves. Reachability determines whether the arm can reach a clicked-on spot from where it is or whether you need to move it closer first.
This hierarchy of modules gives Smart high-level understanding of how to accomplish frequently encountered goals.