Robotics

• 1. Introduction

  ›

  Robotics asks a machine to do four things at once: hold power, take a physical shape, run a control system, and obey software. Pull any one of those out and the robot stops being a robot. A roboticist is the person who keeps all four working together. The goal sounds modest. Design machines that can assist humans across fields as different as agriculture and space exploration. But the same machines that help also displace. A 2017 study found that automation alone puts 47% of US jobs at eventual risk. That tension runs underneath everything that follows. How does a robot stay upright on one wheel? How does it grip a windscreen without cracking it? How does it learn to walk when no engineer told it how? And what happens to the people whose work it quietly takes over?

• 2. Muscles And Movement

  ›

  Actuators are the muscles of a robot, the parts that turn stored energy into movement. Electric motors are the most popular kind, rotating a wheel or gear, while linear actuators move in and out for the large forces that factory robots demand. Portable robots tend to use brushed and brushless DC motors. Industrial robots and computer numerical control machines lean on AC motors instead.

  Piezoelectric motors offer a stranger alternative. Tiny piezoceramic elements vibrate many thousands of times per second, stepping the motor around a circle or driving a screw. Their reward is nanometer resolution, speed, and force packed into a small frame. Series elastic actuation takes a different gamble, placing intentional elasticity between the motor and the load. That softness lowers reflected inertia, which improves safety during collisions and absorbs shock, and it has been used in walking humanoid robots.

  Pneumatic artificial muscles, also called air muscles, are tubes that expand up to 42% when air is forced inside. Muscle wire, a shape memory alloy, contracts under 5% when electricity is applied and has powered some small robots. Electroactive polymers go furthest, contracting up to 380% activation strain, and have been used in the facial muscles and arms of humanoid robots. The most striking candidate is still experimental. Elastic carbon nanotubes could replace a human bicep with wire 8 mm in diameter, feasibly letting future robots outperform humans.

• 3. Ways Of Moving

  ›

  Carnegie Mellon University's Ballbot balances on a round ball and stands at roughly the height and width of a person. Robots with one or two wheels gain efficiency, fewer parts, and the ability to slip through confined areas. Two-wheeled balancing robots use a gyroscope to sense how far they are falling, then drive the wheels up to hundreds of times per second to counter it, following inverted pendulum dynamics. NASA's Robonaut has been mounted to a Segway for the same effect.

  Honda's ASIMO walks using the zero moment point algorithm, keeping inertial forces exactly opposed by the floor's reaction so no moment tips it over. Some human observers say its walk looks like it needs to use the bathroom. In the 1980s, Marc Raibert built robots at the MIT Leg Laboratory that hopped like a person on a pogo stick, falling sideways and jumping to catch themselves. One of his bipedal machines ran and performed somersaults. The most promising path forward uses passive dynamics, where the momentum of swinging limbs powers walking, perhaps ten times more efficiently than the zero moment point method.

  A modern passenger airliner is essentially a flying robot managed by two humans, with autopilot handling takeoff, flight, and landing. Biomimetic flying robots borrow from bats, raptors, gulls, beetles, and dragonflies, with flapping wings that improve maneuverability over propellers. Climbing robots copy gecko toe-pads to scale vertical glass, one named Speedy Freelander. In water, some fish reach a propulsive efficiency greater than 90%, and in 2014 a robotic fish outperformed some real fish in average maximum velocity and endurance.

• 4. The Robot's Hand

  ›

  Matthew T. Mason described robotic manipulation as the robot's control of its environment through selective contact. A robotic arm is called a manipulator, and its working tip, whether a tool or a hand, is the end effector. Most arms carry replaceable end effectors, each suited to a small range of tasks. The most versatile manipulators, like a humanoid hand, can reach up to 20 degrees of freedom with hundreds of tactile sensors.

  Grippers are among the most common end effectors, often just two fingers that open and close on small objects. Friction jaws hold an object with the full force of the gripper, while encompassing jaws cradle it using less friction. Hands that behave more like a human hand include the Shadow Hand and the Robonaut hand. Suction end effectors, powered by vacuum generators, can hold very large loads as long as the surface is smooth enough for suction to grip. Pick-and-place robots use them for electronic components and for objects as large as car windscreens, partly because soft suction is less likely to damage what it lifts.

• 5. Perceive, Process, Act

  ›

  The control of a robot runs through three distinct phases: perception, processing, and action. Sensors report on the environment or the robot itself, that information is processed, and the result becomes signals to the motors that move the structure. At a reactive level, raw sensor data converts straight into actuator commands. At longer time scales, the robot builds a cognitive model that tries to represent itself, the world, and how the two interact.

  Current robotic and prosthetic hands receive far less tactile information than a human hand. One research design wraps a rigid core in conductive fluid held by an elastomeric skin, so that touching an object deforms the fluid path around the electrodes and maps the forces felt. In 2009, scientists from several European countries and Israel developed a prosthetic hand that lets patients write, type on a keyboard, and sense through its fingertips. Other sensing draws on lidar, which measures distance using reflected laser light, along with radar and sonar.

  Human speech is hard for a computer to recognize in real time because the sound of a word shifts with accent, acoustics, volume, and the speaker's health. The first voice input system was designed in 1952. By the end of the 20th century, the best systems could recognize continuous natural speech up to 160 words per minute with 95% accuracy. One of the earliest talking machines came in 1974, when Michael J. Freeman converted digital memory to rudimentary speech using pre-recorded computer discs, programming his robot to teach students in The Bronx, New York. Hanson Robotics later built expressive synthetic faces from elastic polymer skin animated by subsurface servos, and robots like Kismet produce a range of facial expressions for social exchange.

• 6. Work And Its Cost

  ›

  China had the greatest number of industrial robots in operation as of 2022, with 1.5 million units, and was increasing that figure by more than 20% annually. Robots have been used in manufacturing since the 1960s. According to Robotic Industries Association US data, the automotive industry was the main customer of industrial robots in 2016, taking 52% of total sales, and robots can perform over half the labor in the auto industry. By 2003, an IBM keyboard factory in Texas ran fully automated as a lights out facility.

  Spyce Kitchen ran two robotic food-bowl restaurants in Massachusetts between 2018 and 2022. Beyond food, robots handle palletizing, forklifts, nuclear cleanup, solar panel cleaning, Mars rovers, and robot-assisted surgery. The cost of all this efficiency falls on workers. Theoretical physicist Stephen Hawking observed in 2016 that the automation of factories had already decimated traditional manufacturing jobs, and that artificial intelligence was likely to push that destruction deep into the middle classes, leaving only the most caring, creative, or supervisory roles. Proposals such as basic income are raised to replace those lost wages.

  Not all of the substitution is a loss. Putting machines into unhealthy or dangerous environments is an occupational safety benefit, covering high-risk work in space, security, energy, logistics, and inspection. Humans remain better at light-duty work that demands creativity, decision-making, and flexibility. The need to work safely in close quarters has produced cobots, collaborative robots, and some European countries now fold robotics into national programs to promote healthy cooperation between robots and operators.

• 7. Designing The Future Robot

  ›

  In 1997, Professor Hans Moravec, principal research scientist at the Carnegie's Robotics Institute, predicted that robot intelligence would reach the capacity of a lizard by 2010, a mouse by 2020, then a monkey, and finally a human by around 2045. Much robotics research looks past specific industrial tasks toward new kinds of robots and new ways to build them. The study of motion splits into kinematics and dynamics. Forward kinematics manually controls joints to move end effectors, while inverse kinematics fixes the end-effector state and lets the joint values follow.

  Evolutionary robotics borrows from natural selection. A large population of robots competes, those that perform worst against a fitness function are removed and replaced by variants of the winners, and over many generations a satisfactory robot may appear without direct human intervention. Because so many generations must be simulated, the process often runs mostly in simulation before evolved algorithms reach real machines. Bionics applies animal physiology to design, as with BionicKangaroo, modeled on how kangaroos jump.

  Swarm robotics coordinates large numbers of mostly simple robots, where collective behavior emerges from local interactions between the robots and their environment. Quantum robotics studies running robotic programs on quantum computers, which will likely outperform digital ones. According to a September 2021 GlobalData report, the robotics industry was worth USD 45 billion in 2020 and is projected to grow at a compound annual growth rate of 29% to reach 568 billion by 2030. Two major academic gatherings, the International Conference on Robotics and Automation and the International Conference on Intelligent Robots and Systems, are where much of that future is argued out.