MIT's version of the "robotoddler" is just the latest MIT entry in the world of robots that can move themselves in a variety of settings. There's still a long way to go before today's robots evolve into practical, everyday technologies, but even now, autonomous robotic vehicles developed at MIT are exploring uncharted or hazardous places, assisting troops in combat and performing household tasks.
On land
In addition to his well-known work on humanoid robots such as Kismet, Professor Rodney Brooks led the development of several robotic vehicles and co-founded a company, iRobot, that develops these machines commercially. iRobot produces Roomba, a disc-shaped robotic vacuum cleaner for home use, and PackBot, a small, tank-like battlefield robot that can climb stairs and right itself when it flips over. Troops in Afghanistan use PackBots to explore enemy caves, and soldiers in Iraq use them to detect improvised explosive devices and inspect weapons caches. iRobot has also partnered with John Deere to develop r-Gator, an unmanned jeep that can shuttle supplies to and from combat zones.
"In 20 years, we've gone from robots that can hardly maneuver around objects to ones that can navigate in unstructured environments," said Brooks, director of the Computer Science and Artificial Intelligence Laboratory (CSAIL).
He also pointed to the many applications for labor-saving robots, from autonomous lawnmowers to mobile "assistants" for the elderly. Brooks and his CSAIL colleagues are currently working on an experimental robotic assistant built onto a Segway transporter. However, smarter, multifunctional robots that operate usefully are still a ways off. They will require advances such as object recognition (for example, the ability to differentiate between a pile of salt and a crumpled ball of white paper), manual dexterity and interfaces that could make a robot as easy to use as a refrigerator.
Then there's the final frontier: space. With funding from NASA, CSAIL is developing prototypes of autonomous vehicles and humanoid robots for exploration on the Moon and Mars.
In the sea
Professor Chryssostomos Chryssostomidis, director of the Autonomous Underwater Vehicles Laboratory (AUV Lab), envisions "robots filling the vast void of oceans, roaming around, observing, communicating, and reporting back." His lab has spent the past 15 years developing AUVs that have carried out missions ranging from surveying shipwrecks to testing underwater navigation and communication software.
The lab developed the Odyssey class of submarine-like vessels, which evolved into AUVs produced commercially by BlueFin Robotics, a company that spun out of the AUV Lab and still works closely with it. BlueFin vehicles aid research, survey offshore oil fields, and assist the U.S. Navy in mine warfare and battlespace preparation.
The next generation of AUVs, said Chryssostomidis, will include smaller, more robust vehicles that could be tossed out of an aircraft; hovering AUVs that inspect ship hulls for mines; biomimetic AUVs that mimic marine animals (based on past MIT projects such as Robotuna); and surface crafts for applications such as hydrographic surveying and communicating with and shadowing AUVs.
The biggest challenge for AUV engineers is power generation. Most AUVs run on batteries, and current fuel-cell technology limits missions to hours rather than weeks or months of continuous underwater activity. Chryssostomidis and his colleagues are also working on underwater acoustic communication via modem and on software that enables high-level control of both communication and navigation.
In the air
Eric Feron and his research group in the Laboratory for Information and Decision Systems are working on several projects that may lead to more airborne robots. Those projects include intelligent aircraft, communication among multiple air vehicles, and automated takeoff and landing.
The group has already made progress in two of these areas. The "robochopper," a model helicopter outfitted with a sophisticated instrumentation box, can perform autonomous aerobatic maneuvers at the flip of a remote-control switch. Feron, an associate professor of aeronautics and astronautics, also led the development of an intelligent aircraft guidance system that allows a pilot in one airplane to guide another unmanned airplane by speaking commands in English.
An agile aerial vehicle such as the robochopper is better suited than a surface robot to some scenarios, said Feron, noting that it's easier to fly a miniature robotic helicopter through a chaotic urban environment than to deploy a land robot down in the streets.
Feron is taking on the challenge of autonomous landing. Unmanned aircraft presently use GPS (Global Positioning System) for navigation, but that technology is not reliable enough to manage the fine transition between air and ground. "We wouldn't want to put it in any of the critical tasks involved in landing," he said.
The solution, says Feron, is to mimic a human pilot's vision. He is developing what he calls a "collaborative vision scheme," in which the "eye" of a helicopter (a camera), looks at a specially designed target sitting on the landing area. The target allows the helicopter to obtain the position parameters in real time necessary for landing.
A version of this article appeared in MIT Tech Talk on March 2, 2005 (download PDF).
