Bob Anderson demonstrates the capabilities of the Sandia robot affectionately known as M2, for Mighty Mouse.
Real Robots Don't Look Like Humans
June / July 2007
Volume 5 Number 3
Haven't there been times when you wished you were a little taller, a little stronger, had a little more reach, or maybe you just needed something done that you didn't want to do yourself? Wouldn't it be excellent to have a surrogate that could stretch, lift, reach or perhaps maneuver something you didn't want to touch? Best of all, wouldn't it be great if you could teach it to use its own senses to adjust its behavior to varying situations?
Robots now can do all of that, but today's robotic reality isn't usually like a scene in a domestic-comedy movie set in the near future, where the door bell rings and the C3PO-like robot, perhaps wearing a frilly apron and carrying a feather duster, moves to the door. Generally, today's robots are less android-like and used for far more important chores.
Scientists and engineers at Sandia National Laboratories are designing and building machines that act in human-like fashion and can undertake tasks that humans might not be able or willing to do. Those tasks are not just limited to cognitive activities, like performing millions of calculations per second, but real, physical "pick it up from here and move it there" actions. Over the past several years, Sandia's robot makers have developed devices and systems to quickly and flawlessly paint, surveil, disassemble, collect and carry, among other chores. That particular ability to carry things recently provided the solution to a messy problem. A little mobile robot named Swarmy proved its worth when Sandia had a nasty cleanup job.
Boldly go —€¦
During a facilities cleanup operation, an old, below-ground storage tank had a thin layer of sludge in it that tested positive for extremely small but detectable amounts of hazardous materials. The sludge had to be removed before the tank could be closed and abandoned. The tank's shape, depth and position made clean-up efforts difficult. Its low-oxygen, confined-space environment precluded manned entry and inspection. The viscosity of the sludge also presented a technical challenge.
"We didn't want to send anybody down there," Sandia project manager Dan Borneo says. "Twenty feet below ground, you're going through a 30-inch diameter hole with no place to stand up. Digging the tank up would be extremely expensive, but we couldn't just leave it alone," Borneo says. He and his group looked at small robots attached to complex vacuums such as those being used at another DOE site for similar projects. Each of those systems cost approximately $300,000, plus another $300,000 or so to bring one to Sandia in Albuquerque and set it up.
Borneo decided to ask around. "The great thing about being at Sandia," Borneo says, "is if you can dream up a solution to something, someone has already built one and it's sitting on their shelf somewhere. The key is finding the right person and the right shelf."
For the tank cleanup problem, John Feddema, the laboratory's manager of intelligent systems controls, was the right person. He provided one small 10-inch by 2-foot cooperative or "swarm" robot, which was fitted with treads from old snow blower tires for extra traction. The robot's small size allowed it to get from one side of the 16-foot-long tank to the other. Once inside, Swarmy carried a tethered can to the farthest corners of the tank so that pulling the tether dragged the can through the sludge, collecting the material and allowing it to be removed from the tank and emptied for proper disposal. In a few weeks, Swarmy pulled out enough sludge to fill fourteen 50-gallon drums. The last little bits of goop were removed by fitting Swarmy with a modified shop vacuum.
Enough radiation to kill a robot
The ability to go where people can't is obviously a great strength of robots. At the White Sands Missile Range, a Sandia robot withstood lethal radiation to free a stuck cylindrical cobalt-60 radiation source. Gamma rays from the source, which was about the size of a salt shaker, could kill a man in half a minute.
Its radiation field was too deadly for a human, even in a protective suit, to get near enough to free it. The robot, for its successful efforts, was dubbed M2 for the cartoon character Mighty Mouse.
The lethally radioactive cylinder, used to irradiate circuit boards and vehicles to see how their electronic circuits stood up to radiation that would be present if a nuclear weapon detonated on or above U.S. soil, normally arrives and departs the test site through a metal sleeve, driven by pneumatic air. The method resembles that used by drive-up banks, where pneumatic air drives a cylinder containing transactions between the car and the bank. Once it was confirmed that the cylinder was stuck, the lab technicians gradually increased the air pressure in the sleeve from the standard 20 psi to 1,000 psi, but the cylinder wouldn't budge.
The range's management decided to call the local National Nuclear Security Administration's RAP team—€”the Radiological Assistance Program—€”headquartered at Sandia. Sandia manager Phil Bennett told the Range managers that his group had a robot that might do the job. The 600-pound, five-foot-long robot rolled on treads, could maneuver around obstacles, and had a long, multi-jointed gripper arm with the dexterity to reach into awkward places and apply force to drills and screwdrivers. But radiation that can kill a human also can wipe out a robot's electronics. Bennett estimated M2 could withstand intense radiation for only 50 minutes.
Awkward positioning of the stuck cylinder and radiation shielding required that M2 reach up and across, drill a hole through an oddly angled steel plate, and insert a wire through the drilled hole to move a switch in order to free the stuck cylinder. Aided by M2's video camera, Sandian Bob Anderson steered the robot around two free-standing radiation shields and stopped it at the work site. The robot drilled through the steel plate, opening a space for a probe to pass through and push down one side of the teeter-totter switch. But the switch would not move. M2 drilled two more holes, but still no success. The RAP team then directed the robot to yank on the switch wires, hoping to dislodge the switch from the housing —€”again no luck. After 90 minutes of effort, the switch was still stuck and M2 had received so much radiation that its lower portion had stopped responding to commands.
After winching M2 out of the danger zone, the team decided to try removing the entire 3/16-inch steel plate that covered the switch. They repaired or replaced parts to get M2 working again and devised a metal guide to keep a drill-driven screwdriver blade from slipping or stripping the threads of the six screws that held the plate in place. In the re-build, M2 was fitted with a "trigger finger" to operate the drill. At last, with a combination of the trigger finger and a special gripper tip for M2, the team removed the plate and used air pressure to blow the switch out of the cylinder's pathway, freeing the lethal radiation source to return to its safe storage position.
Ready for my close-up
Robots like Swarmy and M2 can operate independently, if not autonomously, to solve unique problems. More often, though, multi-component robotic systems are developed to perform in projects that require high-quality repeatability, automated record-keeping and the assurance of following prescribed processes.
When those tasks are in hazardous environments where errors can have high-consequence outcomes and the robots must deal with gradations or nuances in components, Sandia's experience as a systems integrator for hardware, software, and command and control instructions is invaluable.
One such project is the disassembly of retired Vietnam-era munitions. Demil (for demilitarization) is a robotic system that includes three robots, which use 22 cameras and a variety of sensors. Demil, currently in final testing at the McAlester Army Ammunition Plant in Oklahoma, disassembles 155 mm projectiles that contain air-dispersed mines.
The disassembly process begins when a robot picks up a projectile approximately 32-inches long, weighing 103 pounds and containing 36 mines stacked in layers within the projectile. Each projectile has eight layers: four layers with four mines and four layers with five mines. This variability adds complexity to the robotic system's task.
A recent commercial for a car manufacturer showed a gangly yellow robot making a mistake and being banished from the shop floor. It was heartbroken. Life had lost all meaning. After a series of menial jobs (holding a speaker at a fast-food restaurant), suicide seemed like the only answer. Then, in the best tradition of Hollywood and Madison Avenue, the mistake and the subsequent disgrace were revealed as the robot's fantasy, a dream sequence from a device that cannot dream. But, that yellow machine made of metal, servos, sensors, cabling and programming struck a chord with many viewers. There was something very, very human about the machine's movements and expressed emotions. Some of us who saw the poor thing felt sorry for it, felt its pain even though we know that feeling pain isn't in its repertoire, and in fact that difference between us and it—€”and Swarmy, M2 and Demil—€”is one of its greatest strengths.
Margaret Lovell covers Sandia National Laboratories for Innovation.
Copyright © 2012 | Innovation America