A Brief History of Robotics

You’d think the Robot Revolution would have
happened by now, with the self-aware robots taking over the world, becoming our overlords,
making us work in the tungsten mines. Well, the thing is, the Robot Revolution already
is happening. It’s been happening for decades. It’s just been a lot less bloody than the
movies made us think it would be, and the robots are not self-aware… yet, but in all
likelihood, your car was made by a robot. These days, there are robots helping spacecraft
dock at the International Space Station, and you might even have one faithfully vacuuming
your kitchen floor. Not exactly Skynet, but all of the convenience that we already enjoy
thanks to robots took decades of hard work by scientists, and some of the problems that
robotics engineers were struggling with fifty years ago still haven’t been solved, so there’s
a whole lot of history that went into the robots that we rely on today, and it just
might help you understand why we don’t have robots taking care of our every need, or forcing
us to mine their tungsten. [Intro] If we’re going to talk about the history of
robotics, first, we need to talk about what a robot actually is, and for such a common
term, it’s surprisingly slippery to define, but technically speaking, a robot is just
a machine designed to accomplish a task. That’s it. Now, it might sound like that would cover
everything from a four-function calculator to NASA’s Pleiades Supercomputer, but that
is not what we’re talking about here. When we talk about robots, we’re really talking
about machines that use their programming to make decisions. For example, if you, a
human, decide to pick up a coin from the ground, there are three main steps you have to go
through. First, your eyes need to see the coin and then send that information to your
brain. Then, your brain processes that input and uses things like previous experience to
make a decision to pick it up. Eventually, your brain sends messages to your body to
grasp the coin. Robots go through a similar process, but without
the getting the Slim Jim at the gas station. They can go through that process because most
of the time, they have the components that let them carry out each step. They have sensors
for input, control systems for decision-making, and end effectors for output. Sounds simple
enough, but developing each of these components can be challenging. Sensors have to be able to detect things like
images and sound accurately, effectors have to be flexible and fast enough to do what
we need them to do, and the control system has to make all of the decisions necessary
to get the sensors and effectors working together. Of course, there are so many different kinds
of robots that these components can vary considerably, but that’s the basics of robot anatomy. Now if we tried to talk about every advance
that’s been made in every subfield of robotics in just the past half-century, we’d be here
all day. But to understand how we’ve gotten this far and why robots haven’t taken over
the world yet, we first have to talk about how the development of industrial, humanoid,
and military robots got the field where it is today. Industry is a good place to start, because
that’s where robots first became useful. Since factory work can be repetitive and often involves
lifting lots of heavy stuff, it’s a perfect fit for a machine. The world’s first ever
industrial robot, Unimate, was installed on a General Motors production line in New Jersey
in 1961. Weighing in at nearly a metric ton, it was basically a giant robot arm. Its instructions
programmed on a huge magnetic drum told the arm to stack and weld hot pieces of metal
over and over again. Soon, other car companies got in on the game,
installing their own robotic arms. But this first generation of robots was still in its
awkward stage. The arms weren’t particularly flexible, they were often powered by clunky
hydraulics, and they were ultimately difficult to program. So when a robotic arm called IRB-6
came along in 1974, it was a pretty big deal. This was the first electric industrial robot
that was controlled by a microcomputer. It had 16 KB of RAM, it was programmable, and
it could display four whole digits with its LEDs. Developed by a Swedish engineering firm,
ABB, this robot was used to perform inauspicious tasks like polishing steel tubes, but it was
a crucial step toward developing robots that were easier to program. But while controlling robotic arms was getting
simpler, another issue came up. You can give a robot as much programming as you want, but
if it can’t see, it’s not going to be able to do even seemingly simple things, like figure
out which box should go where on a pallet. Crude visual scanners had been around since
the ’50s; they could only see black and white, and the resolution was worse than what you
get from a flip phone camera. But to give vision to industrial robots, engineers had
to tap into another field that would completely change the robotics game: artificial intelligence. Now, artificial intelligence, or AI, is another
broad, vague term used to describe any attempt to make a computer do something that we’d
normally associate with human intelligence, like translate languages or play chess or
recognize objects. In the 60s, the problem was that even though AIs were getting better
at complex reasoning tasks, like playing chess and proving mathematical theorems, it was
incredibly difficult to actually get the programs to interact with the real world. There’s a
difference, for example, between figuring out the ideal placement of wooden blocks in
a theoretical model and actually moving those blocks into place, because moving them involves
a whole series of discrete decisions and actions, and the robots at the time simply couldn’t
manage that. For robots, vision isn’t just about taking
pictures, it’s also about recognizing objects so that they can react to things and situations
in real time. By the late 1970s, engineers had developed new algorithms that allowed
cameras to recognize edges and shapes by using visual cues like highlights and shadows, but
these programs were still just experimental, stuck in research labs. That all changed in
1981, when the first industrial robot got the gift of vision. A General Motors factory was once again the
guinea pig, implementing a system called Consight, in which three separate robots could use visual
sensors to pick out and sort six different kinds of auto parts as 1,400 parts per hour
moved by on a conveyor belt. Progress! For the next two decades, technology kept improving
— industrial robots were able to see better, move faster, carry heavier loads, and handle
more decisions. These days, industrial robots are advanced
enough that it’s totally normal for a factory to install a robotic assembly line that handles
nearly all of its production, and some industrial robots are heading in the direction of a more
general purpose use, like Baxter, the humanoid industrial robot. Humanoid, yet again a very
subjective term, but it just means human-like. For robots, that usually implies that they’re
designed to look and act as human-like as possible. So Baxter, for instance, is nearly two meters
tall, weighs 136 kg, and has a screen for a face. It also has a vaguely human-shaped
torso and two arms. But more importantly, it can be quickly programmed to do practically
anything. If a factory worker or researcher has a task to do, Baxter can probably handle
it, as long as it’s lifting less than 2.2 kg, because its arms are not industrial strength. But it does need to be told what to do. If
you want Baxter to stack some products in a box, for instance, you’ve got to program
it by manually making it do what you want the first time, and then it’ll imitate that
task. And its shape may be loosely based on a human, but Baxter cannot walk or climb stairs
or talk, so getting a robot to act like a human has proved to be a whole other ballgame,
and it’s one that researchers have been working on for decades, and it’s been slow going. The Wabot I is usually considered to be the
first full-scale humanoid robot. Developed by researchers at Waseda University in Japan
in 1973, it had arms, legs, and a vision system. It could walk, it could pick things up with
its hands, it could even talk. Except that it could only reply with pre-recorded responses
to very specific statements, and it took 45 seconds to take one step. This bot and its
successor Wabot II were a really big deal in their day, but they also pointed out an
important fact: it’s just much easier to design robots to do one task at a time. So recently, the thinking has been if general
purpose humanoid robotics are out of our grasp, we might as well focus on making something
that can do at least one useful task. That’s why in the past 10 years, there have been
more household robots in use than ever, programmed to perform a single function like vacuuming
the floor, mowing the lawn, washing windows or cleaning the pool. They’re not quite Rosie
from The Jetsons, but they were all made possible by the advances that came before them, like
having the ability to sense their surroundings and make decisions in order to navigate the
world. And it’s not like researchers have given up
on the humanoid front. There are humanoid robots in development that can perform some
impressive feats. Honda’s Asimo, for example, can walk at speeds of almost 6 kilometers
an hour, climb up and down stairs, carry a tray, push a cart, among other things. So
again, progress. You just can’t buy one. For the time being, Asimo is basically a research
tool and a spokesmodel for what the future of robotics might look like. But probably the most cutting edge research
going on in robotics today is being done by the military. Take BEAR, a military robot
that’s been in development since 2005. Unlike with humanoid or even single task robots,
in the military, robot design is more about function than form. So BEAR has some humanoid
components, like a head and two arms, but instead of walking like a human, its legs
are covered in treads, like a tank. The legs have a joint that acts kind of like a knee
that can rotate all the way around. Using their special limb design, BEAR has proven
really good at moving through rough terrain, including stairs. It can climb through debris,
carry an injured soldier back to base, it can carry light loads; tell it to go to a
location and it’ll go there. That’s a lot more than most civilian robots can do. But
discussion of military robots would not be complete without talking about DARPA. A division of the US Department of Defense,
DARPA has been working for the last 50 years to turn even the wackiest of concepts into
working technology, and it’s been one of the most active promoters in the field of robotics.
DARPA has stoked innovation by holding robot competitions; it held contests for autonomous
vehicles where robotic cars compete in a race, hundreds of kilometers through the Mojave
Desert, and in the summer of 2015, 11 teams will compete in the DARPA Robotics Challenge
Finals, where human supervised robots will try to carry out complicated and kind of dangerous
tasks associated with disaster response, like scrambling over debris and carrying things
to safety. The aim is to develop robots that can be sent into dicey situations without
putting human lives at risk, and we’re getting there. So robot tech has come a long way since that
first robotic arm in 1961, it’s just that, as often is the case when it comes to humans
and technology, we can dream up awesome designs and uses for robots faster than we can actually
invent them. For now, I’m perfectly happy to have a clean kitchen floor and my freedom
from the tungsten mines while science takes robotics to the next level. Thanks for watching this SciShow Infusion.
If you have questions or comments, be sure to let us know on Facebook, Twitter, Tumblr,
or down below in the comments, and don’t forget to go to YouTube.com/SciShow and subscribe.

Leave a Reply

Your email address will not be published. Required fields are marked *