Ergonomics Report June 15, 2005

From ErgonomicsReport.com

NEWS
Robots-human cooperation The Quest for a More Equal Partnership
By Jennifer Anderson
2005-06-15 00:00:00.0

Robots built on the master-slave model are destined for obsolescence, to be replaced by a generation that will cooperate with man on a more equitable basis. It will be just as critical in this new phase of robotics that designers get the ergonomics right that they build the partnership between man and machine on a clear understanding of what each can do best.

Interviewed for The Ergonomics Report™, three leading robotics experts talked about the kind of robots needed for successful man-machine partnerships and the challenges to designing them.

Dr. Bill Clancey, Chief Scientist at the Human-Centered Computing division of NASA Ames Research Center in California, describes success as "developing a robot with its own research goals and competencies, so it could contribute a personal point of view, like a colleague." He noted it could take decades to understand how to develop this "conscious" robot.

Partnering for a defined task, such as a construction job, would require the ability to work cooperatively, according to Dr. Clancey. "This requires the ability to understand people's explanation of goals, beliefs, and plans, plus to negotiate both physical-interactive and strategic ways of working together." It will be necessary to have a system that can interpret commands intelligently and adapt to unexpected circumstances, especially when people are not supervising, he added.

Dr. Martin Haegele, head of the Robot Systems division of the IPA Fraunhofer in Germany, used household tasks to explain the challenges. He noted that robot lawn mowers and vacuum cleaners are designed to carry out one task only, and that pressing a button is enough to activate the device. Future robot partnership should go beyond that, he said. "We dream of robots carrying out a broad spectrum of tasks and supporting a human in his or her daily living, at home or at work."

A robot executing a broad spectrum of tasks in everyday environments has to be intuitively instructed, Dr. Haegele explained. "Instruction schemes depend on voice, gesture, tactile guidance of arms etc. A dialogue between robot and machine might be necessary to resolve difficult situations." He explained that researchers are still far from being able to instruct a machine the same way they can instruct a person, making "multi-modal user interfaces or instruction schemes a prime research topic."

Dr. Haegele's "dream" robot would be sent to work in homes and offices. IPA Fraunhofer is developing Care-O-bot, a mobile service robot that interacts with and assists humans in typical housekeeping tasks.

Dr. Clancey's robot-human partners would be sent to work on other planets or the Moon. For the present, they are earthbound, participating in experiments in a Utah desert environment that resembles distant Mars. Vulnerability to the hazardous conditions on other planets can be built out of robots, but can't be built out of humans. Spacesuits are buffers, but impair astronauts' mobility, dexterity, communications, visibility and strength. Dr. Clancey's NASA study takes these circumstances into consideration when designing experiments for the human-robot teams.

The design starts with determining "what functionality one wishes to emphasize," he said. Robots might be required for heavy lifting, or transporting loads, or working in areas deemed unsafe for humans, or exploring and surveying, or simply for working alongside humans on certain tasks. He explained that each function amounts to a different technical challenge, "involving mobility - ability to travel over given topography - physical interaction in real-time with people, perhaps involving understanding a person's speech and intentions." The long list of challenges also includes planning traverses, adapting to terrain obstacles and route finding. "Each of these goals is relevant to the work we do in the Utah desert using the Mobile Agents system," he said.

Robots on the move

Dr. Clancey ranked mobility as the most taxing challenge. He described it as "the ability of the robot to safely and reliably move around on its own without being led like a 1-year-old child in the kind of terrain encountered near Hanksville (the site of a current NASA study), and likely to be found in many areas of the Moon or Mars."

The current body of robotics literature points where robotics research is headed, pinpointing mobility and communications as primary areas of interest.

Researchers from Cornell University and the Massachusetts Institute of Technology in the United States and Delft University in the Netherlands have tackled the mobility issue on less challenging terrain with some success. A Cornell press release in February reported that a trio of biped robots developed at the universities performed for scientists at the 2005 meeting of the American Association for the Advancement of Science in Washington DC, ambling along with instinctive, humanlike co-ordination.

Robosapien also walks well and does much more. PC Magazine reported in June that Wow Wee, a company launched by a former robotics physicist at NASA's Jet Propulsion Laboratory in California, Mark Tilden, will introduce Robosapien V2 at the 2005 Consumer Electronics Show in Las Vegas. The movement of the 32-inch tall robot is fluid and agile, according to the magazine article.

"Conversations" between humans and robots

The Wow Wee company has also tackled the communications issue, the main line of investigation for many other researchers in human-robot cooperation. The June article reported that Robosapien V2 will feature enhanced sensing and communication abilities, and will be able to recognize objects and even skin tones, and know when its owner walks into a room. A laser-tracking feature will aid Robosapien's mobility, and it will be able to "speak" more naturally than its predecessor.
Humans and robots need to communicate ergonomically -- efficiently and without misunderstandings -- and the robotics literature suggests that many experts see this kind of two-way understanding as the biggest challenge.

Scientists at the University of Karlsruhe in Germany are developing a robot that communicates implicitly with its human partner. The means? Small biofeedback sensors that detect human stress by monitoring the activity of the human's heart. Their KAMRO robot detects panic, fear, anxiety or stress, then takes action to help. At some point in the future it will also be able to infer the human's state by reading gestures, facial expressions and intonation.

KAMRO is also equipped with tactile, acoustic and vision sensors to allow the robot to explore and analyze its environment so it can behave more "intelligently."

A robotic child minder called PaPeRo, built by NEC in Tokyo and introduced to the world at Expo 2005 in Japan in February, looks more like a plastic toy than a nanny. It uses facial- and audio-recognition technology to recognize children and call them by name, as well as react to nuances in their facial expressions and voices.

The Institut für Neuroinformatik (INI), a research institute at the Ruhr-Universität Bochum in Germany, is developing communication systems based on human gestures and the imitation of human actions. Providing the robot has a vision system that is sufficiently sensitive, it can understand sign language. The robot observes a human performing a particular task, such as grasping an object, and learns how the to perform the task from the human example.

Researchers at the University of Southern California are experimenting with body language to solve communication challenges. They are extending a robot's model of interaction with humans so it can induce changes in a human' s behavior and express its intentions in a way people can understand easily. The task involves designing a subset of body movements and behaviors for the robot that are easily recognized by humans.

Reproduction a one-sided affair for robots

A study at Cornell University has robots replicating themselves, an ability that could one day enhance the human-robot partnership. Dr. Hod Lipson's robots consist of a stack of three white plastic cubes that stand as a simple tower. Each cube has an electronic brain holding a blueprint for building new towers. Feed a robot tower new cubes and it will make a copy of itself in a couple of minutes. Lipson says self-replication could have major implications for how robots are used in environments like Mars or the bottom of the ocean, where repairing them is difficult, and that self-replication is the ultimate form of repair.

Talking about the potential impact of the new technology on human-robot cooperation, Lipson told The Ergonomics Report™ that self replication is an extreme example of the broader concept of self repair. When many people imagine long-term robotics, they imagine durable machines with sturdy construction, he said, but biology offers an alternative: most animals, for example, are not made of durable materials, but they can self-repair. "The ability of robots to repair themselves may ultimately lead to more sustainable human-robot environments that do not require humans to constantly support the machines."

The ergonomic model

The research suggests that new-generation robots will be required to interact and communicate with humans and each other, while responding to unstructured, changing and often difficult environments. To enhance the ergonomics of the interaction, some researchers are placing most of the emphasis on teaching robots to understand humans. Past generations only required humans to understand their "slave." Successful interaction is also likely to require both sides of the partnership to behave autonomously when needed, and assume more equal responsibility for making decisions.

Robots are quicker and more reliable than humans, but relatively inflexible. Human are flexible but prone to errors, fatigue, memory lapses and unreliability. If designers pay their dues to ergonomics, the human-robot partners of the future will complement each other's strengths and offset each other's limitations.

Sources: Dr. Bill Clancey; Dr. Martin Haegele; PC Magazine; NEC; INI, Cornell News