June 12, 2012 16:25 PSP Book - 9in x 6in 06-Junichi-Takeno-c06
92 Professor Rodney Brooks’ Robots
Environment
recognition
Action
planning
Sensing
(a)
(b)
Selecting
action
Drive
system
Figure 6.1. Mechatronics model.
repeated again (Fig. 6.1). A Cartesian robot thus repeats the process
from sensing to behaving. There is no mistake in this process. All
conventional robots are designed after this model.
However, many researchers, including the author, are not neces-
sarily completely satisfied with the mechatronics model. Typically,
no problems are experienced when researchers move a robot in a
laboratory for research purposes, but this serial processing system
turns into quite an annoying process once they attempt to develop a
practical or commercial robot.
In the laboratory, researchers simply pursue the study to achieve
their purpose and are not much concerned about the real-time
response. In addition, we should take into consideration that the
speed of computers was very much slower in the 1980s when
Brooks conducted his robot studies. At that time, if one adopted an
imaging system for the sensors, the turnaround time was terrible —
as long as several seconds to complete the processing for a single
screen. Nevertheless, researchers tried hard to improve the “real-
timeness” by, for example, introducing natural constraints, devising
bypass routing from environment recognition to the drive system,
and incorporating hardware to execute certain programs (Fig. 6.1b).
Natural constraints include the well-known epipolar constraint
explained in Chapter 2. In natural environments, for example, you
cannot see an object behind an obstacle.
Brooks introduced his new robot design methodology against
this historical backdrop. With the robot system conceived by
Brooks, parallel processing replaced the serial processing of the