June 25, 2012 12:21 PSP Book - 9in x 6in 02-Junichi-Takeno-c02
10 Story of Robots
used for industrial purposes. Humanoid robots are not yet good for
practical use. There are many reasons for this. Two major reasons
that are generally cited are (1) the effectiveness of bipedal walking
is not yet established, and (2) the safety is uncertain.
Obviously, traveling on wheels is more efficient than walking as
far as moving is concerned. Moreover, the robot’s gait is unstable
because a complex mechanism is used to achieve bipedal walking.
If the robot falls over while working, human operators could be
injured. The robot itself might also be damaged.
A humanoid that is capable of ukemi (literally, a quick response
to any action) when falling down on the ground is reportedly to be
developed soon. In the future, this problem of falling over will be
solved (Fig. 2.3).
We now go on to a story about industrial robots. A floor-
mounted arm mechanism called a manipulator is the basic form of
an industrial robot. Since the manipulator is mounted on the floor,
it cannot change its position or move. In a mass-production plant,
products are transferred on assembly conveyors one after another.
Figure 2.3. HRP Robot (at the International Robot Exhibition).
June 25, 2012 12:21 PSP Book - 9in x 6in 02-Junichi-Takeno-c02
Story of Robots 11
Figure 2.4. Industrial robots at work (at the International Robot
Exhibition).
Any mechanism designed to process products transferred on the
assembly line does not need to move by itself. A manipulator is a
mechanical structure that simulates the structure of a human arm,
elbow, wrist and hand. So we see a number of artificial arms working
at plants (Fig. 2.4).
The Unimate robot, developed in the United States in 1961, was
the world’s first practical industrial robot. As a typical industrial
robot, Unimate affected the development of industrial robots
thereafter.
The remarkable development of industrial robots is due to the
working principle of “teaching playback.” An operator performs a job
step by step and the robot memorizes the entire procedure. After
learning, the robot repeats what it has memorized by itself. Human
operators perform the work just once to teach the robot. Thereafter,
the robot plays back the steps repeating the job endlessly until the
power is turned off. One might say that the robot performs a kind of
“imitation learning.”
June 25, 2012 12:21 PSP Book - 9in x 6in 02-Junichi-Takeno-c02
12 Story of Robots
Industrial robots perform spot welding, painting, and assembly
more safely and better than humans. Spot welding is a technique
for joining the metal of the car body not by welding linearly but
in dotted lines. Linear welding provides stronger structures than
spot-welding but requires higher welding accuracy. When welding
structures, spot welding in the required locations by the required
number of welds is better suited for robots.
Arc welding is a popular technique for welding metals. An arc
is an electrical discharge like a spark. High voltage is generated
on metal surfaces to produce sparks, and the resultant high
temperature is used to fuse the metals.
Assembly robots are extensively used in the semiconductor
industry for inserting electronic components into substrates. This
job was done by humans in the past. The robot used for this
purpose is generally called a SCALA robot (Selective Compliance
Assembly Robot Arm). A SCALA robot is a simplified version of
the manipulators already mentioned. A SCALA robot is called a
horizontal articulated robot and consists of a vertical linear axis and
one each rotating axis in two perpendicular directions. Its features
include high positional accuracy in a three-dimensional space and
the capability of moving parts from one position to another at high
speed. These are the reasons why such robots are good at inserting
devices.
A large number of manipulator and SCALA industrial robots
have been installed in assembly plants throughout Japan to produce
low-priced and high-quality products despite the disadvantage of
variable-model short-run production, making Japan a nation with
great robotics expertise and an economic giant.
Having introduced industrial robots, I will now proceed to
the story of robots developed by researchers. Robots created by
researchers are deeply related to the study of artificial intelligence
(AI). Let us first review the study of AI.
The subjects in the AI studies curriculum at the Massachusetts
Institute of Technology (MIT) Artificial Intelligence Laboratory in
the United States include language, deduction, learning, vision, per-
ception, operation, programming, architecture, and expert systems.
Of these, vision, perception, deduction, programming, learning, and
operation relate to the study of AI and robots. This suggests that the
June 25, 2012 12:21 PSP Book - 9in x 6in 02-Junichi-Takeno-c02
Story of Robots 13
study of robots shares many common research areas with the study
of AI.
Themes in AI studies include intelligent vision systems, percep-
tion function of robots, robot programming and AI, robot hands and
tactile sensors, and autonomous mobile robots (Grimson and Eric,
1987).
The study themes for the intelligent vision system include
technologies to represent image information captured by vision
sensors as line drawings, detection of 3D shapes through the
combination of lasers and vision sensors, and detection of 3D shapes
by irradiating moir
´
e interfering light and random dots in place of
lasers. Another important theme is the development of an automatic
system to represent an object viewed by vision as line drawings;
decompose the line drawings into basic figures; and describe the
relationships between the figures’ connections.
The central theme in the study of the perception function of
robots is stereovision. In the study of stereovision, at least two
vision sensors set apart a certain distance are used to observe
the object. The difference of how the object is viewed between
the two images captured by the two sensors (generally called
disparity) is calculated to determine the distance to the object
by triangulation. This technique is simple in principle, but the
problem is how to find the correspondence relationship between
the images required for calculating disparity. The image correlation
method is extensively used, but a decisive solution has not been
yet established. The currently available solution is to increase the
number of vision sensors for observation to enhance the reliability
of the measurements. Most stereovision algorithms succeed in
bounding the image domain to be used when searching for the
correspondence relationship between images using an epipolar
constraint. The epipolar constraint means that in stereovision,
assuming a common plane passes through the two imaging planes of
the two vision sensors and the object, the image information of the
object present on the common plane is always reflected on the image
planes of the two vision sensors where these image planes intersect
the common plane and nowhere else. Other viable methods include
optical flow to measure the movement of the sensors by tracking the
images appearing on the sensors.
June 25, 2012 12:21 PSP Book - 9in x 6in 02-Junichi-Takeno-c02
14 Story of Robots
Robot programming and AI are described now.
The playback method is generally used when building and
programming industrial robots. Programming a robot means that
an operator does the job in the machine shop and the process is
recorded in the memory of a robot for learning. With this system, the
robot can flexibly respond to uncertain events and adapt to sudden
changes in the environment such as an unexpected object moving
in its area during operation. To respond to a variable environment,
the robot needs to incorporate an environment observation system
such as vision sensors and run a program capable of driving the
manipulator flexibly.
Studies on sensitive adaptation to a variable environment belong
to the domain of AI study and may be said to be equal to the
study of intelligent robots. Many research papers were published
in this particular field between 1970 and 1990. Some researchers’
activities during this period are introduced here in detail.
Study themes in this field include planning problems, rule-based
systems, model-based systems, and motion planning.
Early famous studies on planning problems include GPS (General
Problem Solver) by Newell and Simon (1963) and STRIPS (Stanford
Research Institute Problem Solver) by Nilsson et al. (1971). Using
GPS, given the start state S and goal state G, the procedure from S to
G is automatically generated. The monkey and banana problem is a
famous toy problem in the study of artificial intelligence. Like GPS,
STRIPS calculates the procedure from S to G automatically based on
hierarchical planning.
Studies involving mobile robots include the famous SHAKEY
developed (1966–1972) by the Stanford Research Institute (SRI) in
the United States and the SHRDLU program at MIT that puts blocks
together.
SHAKEY, the mobile robot, is connected to computers, has a TV
camera and range finder (distance meter) serving as vision sensors.
Driven by two electric motors, the robot moved around the room and
was able to avoid obstacles of a simple shape. The behavior of the ro-
bot was automatically calculated by STRIPS. For example, for a robot
in room1, which is given a goal state of in
room(room5) or moving to
room 5, STRIPS writes a program reading go
through(door1, room1,
room2), go
through(door2, room2, room4), and go through(door3,
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.