What Is a Robot?
Robots are a diverse bunch. Some walk around on their two, four, six, or more legs, while others can take to the skies. Some robots help physicians to do surgery inside your body; others toil away in dirty factories. There are robots the size of a coin and robots bigger than a car. Some robots can make pancakes. Others can land on Mars.
This diversity—in size, design, capabilities—means it’s not easy to come up with a definition of what a robot is.
Founder and CTO, Rethink Robotics
CEO, Toyota Research Institute
In fact, the term “robot” means different things to different people. Even roboticists themselves have different notions about what is or isn’t a robot. And for most of us, science fiction has strongly influenced what we expect a robot to look like and be able to do.
So what makes a robot? Here’s a definition that is neither too general nor too specific:
A robot is an autonomous machine capable of sensing its environment, carrying out computations to make decisions, and performing actions in the real world.
Think of the Roomba robotic vacuum. It uses sensors to autonomously drive around a room, going around furniture and avoiding stairs; it carries out computations to make sure it covers the entire room and when deciding if a spot needs a more thorough cleaning; and it performs an action by “sucking dirt,” as roboticist Rodney Brooks, one of the Roomba creators, explains.
But no definition is perfect. You may argue, and perhaps rightly so, that the definition above could very well describe a dishwasher, a thermostat, an elevator, an automatic door, and many other systems and appliances around us. Take, for example, cruise control in cars. It senses how fast the vehicle is going, compares it to a preset speed, and accelerates or brakes as needed. Is cruise control a robot?
For his part, Brooks is not enthusiastic about considering dishwashers a class of robot. But other roboticists are less strict. A home thermostat can measure the ambient temperature, check a prestored schedule, and turn on the heating or cooling system accordingly. For Gill Pratt, another roboticist, that’s enough to call a thermostat a simple robot.
The thing to keep in mind about this or any other definition is that robots can typically do three things: sense, compute, and act. These three components vary widely from robot to robot. To sense the world, some robots use simple devices, like an obstacle-detecting sonar, while other robots rely on multiple sensors, including cameras, gyroscopes, and laser range finders. Likewise, the compute part can involve everything from a small electronic circuit to a powerful multicore processor or even a cluster of networked computers. As for the action, this is where robots vary the most: Some robots can move around; others can manipulate things. Some robots can move around and manipulate things. Some are designed to perform specific tasks, while others are more flexible and can do many different things.
But although robots vary in how they sense, compute, and act, they all operate in a similar way: Their sensors feed measurements to a controller or computer, which processes them and then sends control signals to motors and actuators. A robot is constantly repeating this sensing-computing-acting cycle, in what roboticists call a “feedback loop.” If you want to take away one thing from this tutorial, it’s this: Feedback makes machines smart, and almost every robot uses feedback.
To make things more concrete, consider BigDog, a rough-terrain quadruped robot developed by U.S. firm Boston Dynamics. BigDog uses sensors to measure the position of its leg joints and the forces applied on them. It also uses gyroscopes and an inertial measurement unit (IMU) to keep track of its position in relation to the ground. Based on that information, BigDog’s computer calculates which hydraulic actuators it should activate to move the robotic legs.
Tap on the buttons to see BigDog’s sensing, computing, and actuation systems.
As BigDog takes a step, it’s continually (several thousand times per second) updating its sensor, computer, and actuator information in a feedback loop that allows the robot to autonomously walk, trot, climb hills, and step over obstacles. Its creators have even kicked BigDog while it was walking and the robot didn’t fall down.
To build BigDog, Boston Dynamics engineers studied how real animals run and balance, and they used some of those ideas to design the robot’s sensing, computing, and actuation systems, combining these three components in a feedback loop.
Let’s go back to our original definition. Another key concept that we should mention is the notion of autonomy. We said robots are autonomous machines, but the level of autonomy differs from robot to robot. Some robots are controlled remotely by human operators. Other robots can run without any kind of human intervention. And a large number of robots rely on both remote control and autonomous behavior.
Again, people will disagree on how much autonomy a machine needs to be called a robot. You can try to tweak the original definition to suit your own opinion on this issue, but the fact is, most definitions will never be perfect. When asked to define a robot, robotics pioneer Joseph Engelberger once said, “I don’t know how to define one, but I know one when I see one!”
Maybe that is the perfect definition of a robot.
So where’s my robot?
Another question you may be asking yourself is, Where is my robot? Where are all those helpful robotic systems and humanoids that science fiction promised we’d have by now? Why can’t a robot fold my clothes for me?
Professor of robotics, MIT
CEO and cofounder, iRobot
Founder, DEKA Research and FIRST
Professor of robotics, UC Berkeley
The reality is that there remain huge challenges ahead for robotics, and practical home robots are still many years away. Many of the same problems that kept robots limited to factories and research labs are still around today. Two main issues are cost and complexity. Robotics components, especially actuators to power wheels and robotic arms, are still too costly. As for the complexity problem, when you combine sensors, computers, actuators, software, and user interfaces into a robot and try to operate it in the real world, things still don’t work perfectly. The robot stalls. Or it operates too slowly. Or it acts in an unsafe manner. To put it another way, things don’t work as well as they would need to in order to turn that robot into a practical commercial system.
Now here’s the good news: Progress in solving those challenges is not only happening—it’s happening faster and faster. Colin Angle, CEO and cofounder of iRobot, says that robotics has advanced more in the past three years than in the past five decades! Areas that are seeing promising breakthroughs include robot vision, learning, and navigation. Robots are getting better at recognizing objects and people, mapping indoor and outdoor spaces, and moving through real-world human environments. Robot manipulation and biped locomotion are advancing too, though more slowly.
A key development is that the technological advances in processors and sensors that made computers and smartphones better and cheaper are also benefiting robots. It’s getting easier to equip robots with powerful sensing and computing systems. Another benefit is that researchers don’t have to keep reinventing the wheel when it comes to assembling a robot, and that means they can pay more attention to robotics software.
Robotics software is a big deal. Without effective and robust algorithms and AI, a robot will never be able to accomplish much. So it’s a good thing that tools for robot simulation, control, and communication are getting better at a fast rate. Major robot makers are relying on open (or mostly open) software platforms, tapping on their user communities to develop capabilities they’d never be able to develop on their own. Hardware and software standards that would allow different robots to interface more easily are still lacking, but today’s robots are not the one-of-a-kind laboratory contraptions they used to be. They use similar components and rely on open-source operating systems and software packages.
Clearly, things are moving in the right direction for robotics. “So where’s my robot?” you continue to ask.”
The answer is that if we want capable, affordable robots to help us in the future, we need more people to develop such robots. That’s right. In the end, it all depends on you. The future of robotics is not going to happen; it’s going to be built. And you can be part of it. Check out the “Getting Started in Robotics” section at the top. Then go join a robotics club or competition, take a robotics course, and become a roboticist.
And maybe then you will build a robot that folds your clothes. And mine too, please.
Text by Erico Guizzo
BigDog illustration courtesy of Boston Dynamics. Photos (from top): David Yellen; Janey Pratt; MIT; iRobot Corp.; DEKA Research; Robert Vente.