aichi_voice_rogo_02.gif cutting_edge.gif No_10_small.gif
[Enduring Traditions] [Feature] [Local Voice] [Exploring] [EXPO 2005] [Gifts of Nature]
[International Aichi] [Cutting Edge] [Time Traveller] [Friendship][Tasty Treat] [LINK PAGE]

arrow_left.gif BACKWARD arrow_right.gif FORWARDarrow_right.gif

Evolving Robots

The "multifingered robot hand," developed at Nagoya University, is being used for the study of such motions as adjusting the grip on a ball.

First-generation robots like the Magic Hand required continuous control by human operators, as these robots were merely extensions of human limbs. Robots of the second generation, including industrial robots for auto assembly, had some limited decision-making powers. Now, research and development is in progress worldwide for the third generation of robots, which will be able to deduce how they should respond to any given set of conditions. Aichi Prefecture has several R&D facilities at the forefront of robot development, an update on whose activities is given on the following pages.

Karakuri: The original robots

Patterning robots after living things

A different angle

Practical applications

Karakuri: The original robots

The karakuri ningyo, one of Aichi Prefecture's best-known handicrafts since around the seventeenth century, are mechanical dolls animated by the use of springs and gears. Those made in Aichi are of unparalleled sophistication; a tea-serving doll, for example, can carry a cup to a guest's seat and bring it back after the guest finishes drinking from it and returns it to the doll's tray. What makes it possible for the karakuri to perform such elaborate feats is the superior gear-making techniques developed in Aichi. Because the gears used in the early days of mechanical-doll making were made of wood, they were prone to warping, and getting them to perform fine movements was no easy matter. By carefully calculating the numbers of teeth, revolution speeds, and ranges of movement, and by making each gear from several different kinds of wood, gear makers compensated for the warping factor. The techniques they perfected were later applied to the production of metal gears and used in the development of spinning machines. And these technologies served as the basis for the second generation of industrial robots. Aichi, therefore, had a historic advantage in robot development.

This brachiating robot moves by swinging like a monkey from one branch to another.
Two kinds of technology were crucial in the development of second-generation robots: software (for computer control) and hardware (engineering technology to ensure that the robot faithfully followed commands). The same holds true for the robots of the next generation. No matter how highly evolved the artificial-intelligence or other software that goes into the robots, they cannot be developed for practical use unless the machine technology is there to convert the software commands into action.

Patterning robots after living things

Third-generation robots are distinguished mainly by their autonomy and self-sufficiency. Autonomy is the ability to make decisions leading to a goal, even under conditions other than those anticipated. And self-sufficiency means being able to move freely by means of an internal power source.

Hidenori Ishihara, an assistant professor of engineering at Nagoya University's Microsystems Engineering Department, is involved in the development of a next-generation robot that can swing from one branch to another.

"A brachiating robot," Ishihara explained, "moves like a monkey by swinging arm over arm from one hold to another. This robot needs an external electric power source for the 14 motors on its body, though, and so it's not self-sufficient. But it maintains an awareness of its own movements through two cameras and is able to figure out how it should move through these images. This ability makes the robot autonomous, though in a limited way." In the course of developing the brachiating robot, the research team made many trips to the zoo, where they spent hours observing monkeys. According to Ishihara, the question of how to reproduce a monkey's movements in a robot served as the departure point for their research.

"While a robot can never become a living thing, our ultimate goal is to create robots with the autonomy and self-sufficiency of animate beings. We have not tried to reproduce the muscular and joint structure of a monkey because that would be difficult with the present technology. Instead, as the first step toward re-creating a living thing, we decided to faithfully duplicate just the movements."

To swing from one branch to another, the brachiating robot needs only one to two seconds—a speed that belies its heavy, metallic appearance. This robot really does move like a monkey.

A different angle

Other robot researchers, meanwhile, are focusing on the parallel structure, rather than the direct movement, of animals. One of them is Luo Zhiwei, a doctor of engineering at the Bio-Mimetic Control Research Center within the Institute of Physical and Chemical Research in Nagoya. Luo is trying to apply what he knows about the self-organization of parallel structures of an animal's nervous system to developing a robotic control system.

"As an example," said Luo, "when we touch something hot, we withdraw our hand. This is not something we do consciously; rather, our low-level spinal nerves issue the command that leads to the reflexive action. Given that this is the case, we should be able to control a robot using a system that imitates the spinal nerves."

As the focus for his experiments, Luo chose a four-legged walking robot resembling a cat. On each of the four legs, he installed a small computer that runs a simple program causing a leg to move by monitoring the motion of the other three legs. Luo described his results as follows: "I've learned that the robot is better able to cope with a variety of environmental conditions if I install small, light computers on each leg rather than have one big, heavy computer to control all of the legs in a centralized manner."
Ishikawa climbs Gasherbrum II. This was the sixth successful 8,000-meter climb for Ishikawa, then 62 years old.

Suppose that the robot encounters an obstacle while walking. If the control system inside the robot is centralized, the robot searches among its pre-programmed patterns, finds one that most closely resembles the obstacle in front of it, and uses this information to determine the best way to get around the obstacle. But this has major drawbacks. The patterns must be programmed into the computer by a human being based on anticipated conditions, which naturally limits the number of patterns the robot can choose from. If it encounters a totally unfamiliar obstacle, it will not be able to cope.

In contrast, consider how a robot with a computer on each leg would handle the situation. If the robot has a simple program instructing it to avoid obstacles, the first leg that recognizes an obstacle will move to avoid it, and the other legs will follow. There is no need for the robot's program to have different patterns in its memory.

A robot that can thus respond to unanticipated environmental conditions flexibly in real time is essentially autonomous. Although it is generally believed that third-generation robots will need high-performance computers to achieve full autonomy, Luo is trying to bring the computing requirements more within reach.

Practical applications

The past few years have seen ever-more-spectacular advances toward third-generation robots in Japan. Bipedal walking robots and pet robots are coming out one after another. While Aichi's robot technology is at the leading edge, most of the prefecture's R&D efforts are being directed these days to less conspicuous developments. Toshio Fukuda, a professor at Nagoya University's Center for Cooperative Research in Advanced Science and Technology, explains: "This is because we've begun to set our sights on bringing third-generation robots into practical use. Our immediate objective is to develop care-giving robots."

Japanese society is graying at an unprecedented pace. By the first half of the next century, people aged 65 or older are expected to account for a fourth of Japan's total population. Accordingly, securing enough health workers to look after these aged citizens has become an urgent task. At this stage, of course, we do not know whether robots will perform all of the duties of human care givers or whether they will play only a supporting role—for example, assisting people with meals by carrying food to their mouths. Either way, Fukuda says, safety and reliability will be prerequisites.

"Whatever role a robot performs," Fukuda notes, "we can't have it breaking down or going out of control and injuring people. To make robots practical for everyday use, we have to make sure they're safe and reliable. Machines will need to be developed that can stand the test of repeated use—thousands or even tens of thousands of times—without breaking down. While there is a common understanding among researchers around the country that the greatest need for robots is in the area of elderly care, those in Aichi have exercised their characteristic foresight in focusing their activities on ensuring safety and reliability, such as by faithfully re-creating animal movements and developing pragmatic control mechanisms."

(Phots by Masatsugu Yokoyama, Text by Masahiro Ota)