Robotics Fundamentals
A Short History of Robotics
From the word “robot” in a 1921 play to today's warehouse fleets and walking machines. The milestones that shaped modern robotics and why each one mattered.
beginner · 12 min read
Lesson content
Robotics has two origins. One is in fiction. The other is in factories.
They took nearly forty years to meet and they have never fully agreed since. That gap explains a lot about the field, including why most people still picture a robot the way a play from 1921 described one.
Definition
Robot
The word was coined by Czech writer Karel Čapek in his 1921 play R.U.R. (Rossum's Universal Robots), from robota, meaning forced labour or drudgery.
Formal definition
The Robot Institute of America's working definition: “a reprogrammable, multifunctional manipulator designed to move material, parts, tools or specialized devices through variable programmed motions for the performance of a variety of tasks.”
Read that industrial definition again slowly, because every word in it was fought for.
Reprogrammable rules out a machine that does one thing forever.
Multifunctional rules out a single-purpose tool.
Variable programmed motions rules out a cam-driven mechanism that traces the same path because its shape says so.
What is left is basically a list of everything the first industrial robots could do that ordinary automation could not.
The turning points
1921: the word
Čapek's R.U.R. gives us the word and also the plot in which the robots rebel. Both have stuck. Isaac Asimov coins "robotics" in 1941 and publishes his Three Laws in 1942. Fiction, but influential fiction.
1954–1961: the first industrial robot
George Devol patents a programmable manipulator in 1954 and coins "Universal Automation", shortened to Unimation. With Joseph Engelberger he builds Unimate, installed at a General Motors plant in 1961 to unload a die-casting machine, which was hot and dangerous work. This is the moment robotics becomes an industry.
1969–1978: arms get smart
The Stanford Arm (1969) is the first computer-controlled, all-electric arm designed for computer control rather than adapted to it. In 1978 Unimation and GM produce the PUMA (Programmable Universal Machine for Assembly). The SCARA is invented in Japan the same year for fast flat assembly. The shapes of modern arms are now set.
1980s: the field becomes a discipline
Industrial growth is rapid. This is also the decade universities start teaching robotics across mechanical, electrical and computer science departments. The textbooks this curriculum draws on are written in this decade.
1986–1997: behaviour and autonomy
Rodney Brooks argues that robots should react to the world rather than model it first. He calls it the subsumption architecture and mobile robots start actually working. In 1997 NASA's Sojourner drives on Mars and Honda reveals the P2 humanoid.
2002 onward: robots leave the factory
The Roomba (2002) sells in millions. The DARPA Grand Challenges (2004–2007) turn self-driving from a curiosity into a field. Kiva's warehouse fleets (acquired by Amazon in 2012) put thousands of mobile robots into daily commercial use. ROS is released in 2007 and becomes the field's common software.
2010s to now: learning and legs
Deep learning solves perception problems that resisted forty years of hand-crafted methods. Collaborative arms come out from behind fences. Reinforcement learning makes legged robots genuinely capable over rough ground. Large language models start being used as task planners.
Why did it take from 1961 to about 2002 for robots to leave the factory?
Because a factory quietly removes almost every hard problem.
Parts arrive in the same place every time. The lighting never changes. The floor is flat, clean and empty. Nobody wanders through. A robot in that room needs no perception and no judgement at all. Being repeatable is enough.
Walk out of the building and all of that vanishes. Now there is clutter, moving people, changing light and uneven ground. Handling that needs sensing and computing. Neither was affordable until the 2000s.
So the thing that finally let robots leave was not better motors. It was cheap cameras, cheap LiDAR, cheap MEMS gyroscopes and processors fast enough to make sense of them.
| Era | The enabling technology | What robots could suddenly do |
|---|---|---|
| 1960s | Hydraulics and magnetic drum memory | Repeat a taught sequence reliably |
| 1970s | Minicomputers, electric servo drives | Be programmed in a language, not by wiring |
| 1980s | Microprocessors, cheap encoders | Closed-loop control at every joint, at scale |
| 1990s | Cheap sensors, probabilistic methods | Estimate where they are while moving |
| 2000s | LiDAR, MEMS IMUs, open-source software | Navigate unknown buildings and roads |
| 2010s | GPUs and deep learning | Recognise objects reliably enough to act on it |
| 2020s | Learned control, foundation models | Walk over rubble and follow an instruction in plain language |
Example
Unimate's first job and why it was chosen
The first Unimate did not assemble anything delicate.
It lifted hot die-cast parts out of a machine and stacked them. Dangerous, repetitive and, above all, identical every single cycle.
That choice tells you exactly what 1961 robotics could do: move to a taught point, close a gripper, move to another taught point.
No sensing. No adaptation. Nothing else.
Every capability added since has been about relaxing one of those constraints.
Two threads run through the whole century and both are still running.
The first: robots take the jobs that are dangerous, dull or dirty. That framing has held from hot die-casting in 1961 to nuclear decommissioning today.
The second: the hard part keeps moving up the stack. First the hard part was mechanics. Then it was control. Then perception. Now it is judgement and interaction.
Here is the evidence for that second one: the mechanics of a robot arm have barely changed since the PUMA in 1978. Everything that got better got better above the metal.
Reading Asimov's Three Laws as an engineering standard.
They were a plot device and most of the stories are about how they fail. Real safety comes from ISO 10218 and ISO/TS 15066: force limits, separation distances and risk assessments.
Assuming progress has been steady.
It has been lumpy. Each jump followed a component getting cheap rather than an idea arriving. SLAM was understood long before LiDAR was affordable.
Treating humanoids as the goal the field has been heading toward.
Humanoids are one branch. The commercially transformative robots have been arms bolted to floors and boxes on wheels.
Key takeaway
- “Robot” comes from Čapek's 1921 play and the Czech robota, meaning forced labour. Asimov coined “robotics” in 1941.
- Unimate (1961) was the first industrial robot and it did a dangerous, perfectly repetitive job.
- Factories removed the hard problems, which is why robots stayed in them for forty years.
- Cheap sensing and computing let robots leave, not better mechanics.
- The hard part keeps moving up the stack: mechanics, then control, then perception, now judgement.
Finished this lesson?
Guest progress is saved on this device.