Background
The first successful microcomputer in the early '80's
immediately fueled the introduction of several personal robot offerings.
Heathkit, in introducing the small Hero mobile robot, sold several
thousand in just a few years. Nolan Bushnell of Atari fame and fortune
founded Androbot, which produced the Topo and Bobo personal robots.
RB Systems produced a Hero-like robot called the RB5X. In the mid
'80's Arctec Systems (see www.robotswanted.com/robotgallery/arctec
for a technical description) produced perhaps the highest level of
autonomy robot, the Gemini. While few Gemini's were sold due to it's
$8,000 to $12,000 price tag, it could map an entire home, path plan
from room to room, and follow that path while avoiding previously
unknown, or unmapped obstacles. This level of autonomy is essentially
the same as Carnegie Mellon University's (CMU's - www.cmu.edu/home/news/nursebot.html)
nurse-bot prototype, Florence ("Flo"), developed over the
last few years. Applied Systems Intelligence, Inc. (www.asinc.com)
has been the lead in developing the high level of autonomy software
necessary for Boeing's new robot aircraft, the X-45. GeckoSystems
(www.GeckoSystems.com)
has achieved similar levels of automatic self-navigation, albeit with
lower power consumption and cost for their Personal Computer Robots
(PCR's).
At first glance it seems little has changed in mobile
robots over the past fifteen years. In many ways, this is true. However,
we will look at the technological requirements to achieve a high level
of autonomy in mobile robots in general and personal robots specifically.
The challenge of creating a robot is more software than hardware.
As the environment for the robot's operation becomes increasingly
dynamic, the software to manage
|
|
the near real-time flood of information
becomes even more complex and requires significant computational resources.
For example, the Gemini had five Apple computer CPU's on board primarily
running tens of thousands of lines of assembly language. CMU's Flo
has two Pentium class CPU's on board with additional computational
power off board. The hardware requirements for Applied Systems "cognitive
engine." are not available. These mobile robots, and most of
the others currently on the market (excluding the X-45, which has
not published its characteristics), only have continuous activity
for one to two hours before requiring a full recharge. Often the robot
inactivity cycle or recharge time is more than double that of its
useful cycle. This also holds for Sony's offerings such as the Aibo,
Activmedia's AmigoBot, iRobot's iRobot, Probotics' Cye, and Friendly
Machines' RoboMow. Even Eureka's and Husqvarna's dedicated vacuuming
robots only have one to two hours of battery life before needing four
to six hours of recharging. The GeckoSystems' CareBot PCR, an exception
to this trend, has thirty to thirty-five hours of usage with a battery
recharge time of four to eight hours.
All of the aforementioned robots, with the exception of
the Cye and CareBot, have "brains on board" (the computer
that controls the robot resides on the robot). This is an important
concept since the level and type of autonomy exhibited by a mobile
robot is directly proportional to its computational power. In general,
the more computer power available for a given robotic platform, the
higher the possible autonomy level.
Today's ubiquitous PC is truly astonishing in its ability
to offer a 1980-mainframe level of computing power at consumer electronic
prices. Our microcomputer industry has exceeded Moore's Law predictions,
the doubling of computer power every eighteen months, for nearly fifteen
years and in spite of |
