Chapter 3: Classification of Robots
The science club met in the computer lab, and one video played after another.
In the first clip, a small six-wheeled rover crawled slowly across grey dust. It was the Pragyan rover, which India landed on the Moon in 2023. In the next clip, a drone flew low over a green field, spraying a fine mist over the crop. Then came a boxy yellow machine with a bright lamp and a camera, sinking into dark water to inspect a pipeline lying on the sea floor. The last clip showed a machine with two legs and two arms, walking across a room and lifting a carton onto a shelf.
Ravi said all four were robots. Meera agreed, and then asked a harder question: if all four are robots, why does not a single part of one look like any part of another?
The answer is that a robot is built for the place it must work in and for the way it must be controlled. Fig. 3.0-1 places the four machines from those clips side by side.
Diagram
Fig. 3.0-1 – four robots drawn side by side in one strip: a six-wheeled rover on rocky ground, a four-rotor drone flying above a field, a boxy underwater robot with a lamp and camera below the water surface, and a two-legged, two-armed robot standing on a floor; label each one with the place it works in — ground, air, water, indoor floor
This chapter will answer all five of these questions.
3.1 Why We Group Robots
Walk into a hardware shop and ask for a screwdriver. The shopkeeper does not empty every shelf on the counter. He goes straight to one drawer, because the shop is arranged in groups — screws in one place, wires in another, paint in a third. Sorting things into groups is what makes a large collection usable.
Robots are a large collection too. A robot is a machine that senses something, works out what to do, and then acts on the world around it. That single description covers a rover on the Moon, a drone over a field, and a robotic arm bolted to a factory floor. Calling all three “robots” is correct, but it is not very useful. Two robots can share the same name and share almost nothing else.
So engineers sort them. Putting robots into groups on the basis of a shared feature is called classification.
What sorting gives you
Grouping robots is not just tidiness. It does three real jobs.
First, it helps you choose. If a job has to be done at the bottom of a lake, you can ignore every robot that was never built to be waterproof, and look only at the small group that was.
Second, it helps you compare. Two robots from the same group face the same problems, so it makes sense to ask which of them is faster, cheaper or steadier. Comparing a drone with a factory arm on speed would tell you nothing useful.
Third, it makes talk easier. When one engineer says “underwater robot” to another, a whole set of features — sealed body, thrusters, a camera with its own lamp — is understood without being listed.
The two ways of sorting
Robots are grouped in two main ways, and this chapter follows both.
The first is by field or terrain — the surface or space a robot is built to work in. Terrain simply means the kind of ground or surroundings a machine has to cope with. On this basis, robots are grouped as ground robots, aerial robots and underwater robots.
The second is by control — who or what decides the robot’s next move. On this basis, robots are grouped as manual robots and automatic robots.
See Fig. 3.1-1, which shows both ways of sorting on one page.
Diagram
Fig. 3.1-1 – a two-branch tree; the box at the top reads “Classification of Robots”; the left branch is labelled “Field or terrain” and opens into three boxes — Ground robots, Aerial robots, Underwater robots; the right branch is labelled “Control” and opens into two boxes — Manual robots, Automatic robots
These two ways of sorting do not compete with each other. They answer different questions: one asks where does it work, the other asks who is in charge of it. Later in the chapter you will also meet robots described by the living creature whose shape or movement they copy — a third kind of description, and one that fits alongside the other two rather than replacing them.
3.2 Classification on the Basis of Field or Terrain
You would not wear football boots into a swimming pool. The boots are not badly made — they are made for a surface that grips back. Water gives nothing to grip, so the studs become useless weight.
A robot faces the same problem, only worse. It cannot change its body halfway through a job. Whatever surroundings it will work in — a factory floor, open sky, deep water — have to be decided first, and the machine is then built around that decision. This is why the first way of sorting robots is by the field or terrain in which they operate.
Three things about a robot are settled almost entirely by its terrain.
The first is what holds it up. A machine on the ground rests on the ground itself. A machine in the air has to hold itself up every second it is switched on, using spinning rotors or wings. A machine in water is held up partly by the water pushing back on it.
The second is what it pushes against to move. Wheels push against the ground. Rotors push against air. Propellers and thrusters push against water. See Fig. 3.2-1.
Diagram
Fig. 3.2-1 – three panels side by side; panel one shows a wheeled robot on ground with an arrow at the wheel labelled “wheel pushes the ground”; panel two shows a four-rotor robot in air with downward arrows labelled “rotors push air down”; panel three shows a robot under water with a rear thruster and an arrow labelled “thruster pushes water back”
The third is what the surroundings do to the machine itself. Air and water carry dust, rain and salt. Electronic parts stop working when water reaches them, so an underwater robot needs a sealed body, while an indoor ground robot may need no sealing at all.
On the basis of terrain, robots are grouped as ground robots, aerial robots and underwater robots. The next three sub-sections take them one at a time.
3.2.1 Ground Robots
A cycle is beautiful on a smooth road and hopeless in a ploughed field. A tractor is the other way round. Both travel on the ground, yet they are built differently, because “the ground” is not one thing. It can be a polished factory floor, a rocky slope, loose sand, or a heap of broken concrete.
Robots that work on such surfaces are called ground robots. They are also called terrestrial robots, from a word meaning “of the earth”.
Because the surface is already holding the robot up, a ground robot can afford to be heavy. It can carry a large battery, a strong arm, or a load of goods. That freedom is the biggest advantage this group has over robots in air and water.
The four ways of moving
The real design question for a ground robot is how it will move. Four answers are in common use, shown in Fig. 3.2-2.
Diagram
Fig. 3.2-2 – four small robot bodies drawn in a row on the same rough ground line, labelled Wheeled, Tracked, Legged, and Wheeled-legged hybrid; the wheeled one has four round wheels, the tracked one has two continuous belts, the legged one has four jointed legs, and the hybrid has jointed legs ending in wheels
Wheeled robots use wheels. Wheels are simple to build, cheap, and lose very little energy on a hard flat surface, so this is the first choice indoors — in factories, warehouses and hospital corridors, where the floor is smooth and the paths are known.
Tracked robots use continuous belts, like the ones on a bulldozer, wrapped around a set of wheels. Tracks spread the robot’s weight over a much larger patch of ground, which is what keeps them from sinking into mud or sand.
Legged robots have legs with joints instead of wheels or belts. A leg can be lifted over an obstacle and placed on the far side of it, so a legged robot can cross a step, a gap or a pile of rubble that would stop a wheel.
Wheeled-legged hybrid robots combine both: legs with wheels at the ends. On smooth ground they roll, which saves energy, and on rough ground they walk.
Legs look like the best answer to everything, but they are not. Fig. 3.2-3 shows why the choice is a real trade-off, and Table 3.1 sets the two side by side.
Diagram
Fig. 3.2-3 – photo of a four-legged robot climbing a flight of stairs, with one leg raised above the step edge
Where ground robots are used
Ground robots are put to work in a wide range of fields.
Manufacturing: wheeled robots carry parts and cartons between machines in factories and warehouses, so that people are not spending the day pushing trolleys.
Agriculture: field robots move along crop rows to sow seed, pull out weeds, or spray only the plants that need it, which cuts the amount of chemical used.
Search and rescue: tracked robots are sent into collapsed buildings and mines, where a rescuer would be in danger, to look for people and send back pictures.
Space exploration: rovers such as Perseverance, which NASA landed on Mars in 2021, drive across the surface of another world and send back pictures and readings of the rock and soil.
Surveillance and security: slow patrol robots move along a fixed route in a campus or a large plant at night, carrying cameras and sending an alert when something is out of place.
Every one of these uses has the same limit behind it. A ground robot goes only where the ground goes. A gap, a flooded stretch, a wall or a staircase can end its journey, and if it gets stuck in soft sand there is usually nobody nearby to pull it out.
3.2.2 Aerial Robots
Put a ground robot down and switch it off. It stays exactly where it is. Do the same to a flying machine and it falls.
That single difference shapes everything about robots that work in the air. Air is thin and gives way, so a flying robot has to push a large amount of it downwards, without stopping, for as long as it is in the sky. Every extra gram it carries has to be held up as well, which is why these machines are built as light as their job allows.
Robots that fly and work in the air are called aerial robots. Since nobody sits inside them, they are also called unmanned aerial vehicles (UAVs), and in everyday speech they are called drones.
The four common designs
Quadcopters have four rotors. Because lift is shared between four small propellers close to the body, a quadcopter can hover in one spot and turn on the spot, which makes it easy to fly in a small space. This is the design used for most camera drones.
Fixed-wing drones have wings like an aeroplane. They do not hover; they must keep moving forward, and the air flowing over the wing holds them up. In return they use far less power for the distance covered, so they fly faster and much further than a quadcopter of the same weight.
Hexacopters and octacopters have six and eight rotors. More rotors mean more lift, so these machines can carry heavier cameras or spray tanks. They are also steadier, and with so many rotors the machine can usually stay in the air even if one motor stops.
VTOL (Vertical Takeoff and Landing) drones are a mixture of the first two. They lift off straight upwards on rotors, like a helicopter, then tilt into forward flight and let a wing carry them, like an aeroplane. So they need no runway and still cover long distances.
See Fig. 3.2-4 for the four shapes.
Diagram
Fig. 3.2-4 – four aerial robots drawn as simple top views in a row, labelled Quadcopter, Fixed-wing drone, Hexacopter, and VTOL drone; show four, six and eight rotor arms respectively on the first, third and fourth-from-left shapes, and a plain wing with a nose propeller for the fixed-wing drone
Where aerial robots are used
Aerial photography and videography: a camera drone lifts a shot high above a wedding, a stadium or a construction site, where earlier only a helicopter could go.
Mapping and surveying: a drone flies a fixed pattern over an area, taking overlapping photographs that software joins into an accurate map of the land.
Agriculture: drones spray fertiliser or pesticide over a field in a few minutes, and cameras looking down on the crop show which patches are growing poorly.
Environmental monitoring: drones count wildlife, watch forests, and reach places such as glaciers and volcano slopes that are hard or unsafe to walk into.
Search and rescue: after a flood or a landslide, a drone can cover in an hour the ground a rescue party would take a day to search on foot.
Infrastructure inspection: bridges, power lines and wind turbines are checked by flying a camera along them, instead of sending a worker up on ropes.
Flying comes with limits that no design removes. Battery life is short, the load a drone can carry is small, strong wind pushes it off course, and rain gets into it. Flying also happens over other people’s heads, so in India, as in most countries, there are rules about where a drone may be flown and permission is needed for many flights.
3.2.3 Underwater Robots
Water is the hardest of the three terrains, and not because it is wet.
Push your hand into a bucket and you already feel water pushing back. Go deeper and that push grows fast. Sunlight fades within a few tens of metres, so most of the ocean is dark. Salt water eats into metal. And a mobile phone stops working the moment it goes under, because the radio waves it depends on are absorbed by water within a very short distance. A machine sent down there has no signal, no satellite position fix, and often nothing to see.
Robots built for such places are called underwater robots.
An underwater robot is built inside a sealed, pressure-proof body. It moves not with wheels but with thrusters — small propellers that push water backwards so that the robot moves forwards. It carries its own lamps, because there is no light to borrow. Fig. 3.2-5 shows the two forms these machines take.
Diagram
Fig. 3.2-5 – a cross-section of the sea; a ship floats on the surface with a long cable running down to a boxy robot with a camera, lamp and arm hovering near the seabed, labelled “ROV — tethered, driven by a person on the ship”; to the right, a torpedo-shaped robot swims freely with no cable, labelled “AUV — no cable, follows a stored plan”
Why depth is a problem
The push of the water on every square centimetre of the robot’s body is called pressure, and it rises steadily as the robot goes deeper. In sea water, the pressure grows by roughly one atmosphere for every 10 metres of depth, where one atmosphere is the pressure the air already puts on us at sea level.
\[p = 1 + \frac{d}{10}\]
Here \(p\) is the total pressure on the robot in atmospheres, and \(d\) is its depth below the surface in metres. The 1 stands for the air pressing down on the water above it.
This is why an underwater robot is heavy and thick-walled while an aerial robot is light and thin. A rounded, sealed casing carries this squeeze far better than a flat panel, which is why so many underwater robots are shaped like a tube or a ball.
The two types
Remotely operated vehicles (ROVs), also called unmanned underwater vehicles (UUVs), are joined to a ship at the surface by a long cable called a tether. The cable carries electric power down and live video up, and a person on the ship drives the robot while watching that video. Because a human is watching every second, an ROV can do delicate work — turning a valve, cutting a rope, picking up one object from the seabed.
Autonomous underwater vehicles (AUVs) have no cable. They carry their own battery, and the whole mission is stored in their computer before they are lowered into the water: the path to follow, the depth to hold, the readings to record. The machine goes down, does the work on its own without anyone steering it, and comes back to the surface to be picked up.
Where underwater robots are used
Marine exploration and research: robots reach depths no diver can survive, and send back the first pictures of animals and rock formations there.
Oceanography: AUVs record the temperature, saltiness and movement of sea water over long routes, and that data feeds into weather and climate studies.
Environmental monitoring: sensors on the robot measure water quality and pollution, and cameras show the health of coral and other marine habitats.
Underwater archaeology: ROVs survey shipwrecks and sunken sites, photographing them without disturbing what is lying there.
Offshore industries: oil and gas platforms use ROVs to inspect and repair pipelines and structures on the seabed, work that would otherwise put divers in danger.
Search and recovery: after an accident at sea, robots are used to find and bring up objects from the bottom.
The limits are as clear as the uses. An ROV can never go further than its cable. An AUV can go far, but nobody can change its mind halfway, and if it fails there is no cable to pull it back by. And in muddy or churned-up water, even the best camera and lamp show very little.
3.3 Classification on the Basis of Control
A ceiling fan runs at whatever speed you set. If the room gets hotter, the fan does nothing about it — you have to get up and turn the regulator. An air conditioner set to 24 °C behaves differently. It measures the room temperature itself, and when the room warms up it cools harder without being asked.
Both machines end up doing the same job. What differs is who makes the decision.
The same difference splits robots into groups. Every robot works through three steps, over and over: it senses something about its surroundings, it decides what to do next, and it acts by moving a motor or a joint. The sensing and the acting are always done by the robot. It is the middle step, the deciding, that can sit either with a person or inside the robot’s own program. Who holds that middle step is what we mean by control.
Fig. 3.3-1 shows the two paths side by side.
Diagram
Fig. 3.3-1 – two loops drawn side by side; on the left, labelled “Manual control”, the loop runs sense → picture sent to a person at a screen with a joystick → person decides → command sent back → robot acts; on the right, labelled “Automatic control”, the loop runs sense → robot’s own program decides → robot acts, with no person in the loop
Notice what does not change. Both robots still have sensors, both still have motors, and both still act on the world. Control is not about what a robot is made of. It is about where the decision comes from.
On the basis of control, robots are grouped as manual robots and automatic robots. The two sub-sections that follow take each in turn, and you will see that the line between them is less like a wall and more like a slope, with some robots standing partway along it.
3.3.1 Manual Robots
A suspicious bag is lying at a bus stand. Nobody knows what is inside it. A police team drives up, unloads a small tracked machine with a camera and a gripping arm, and sends it towards the bag while everyone stands well back.
The machine does not decide anything. An officer watches its camera on a screen and moves it with a joystick — forward a little, turn left, lower the arm, open the gripper. Every single move is a human decision. The robot supplies the body; the person supplies the judgement.
Robots that work this way are called manual robots.
Because the operator is usually far from the machine, these robots are also called tele-operated robots, or telerobots for short. “Tele” means “at a distance”, as in television — a picture from a distance. Fig. 3.3-2 shows the arrangement.
Diagram
Fig. 3.3-2 – on the left, an operator seated at a screen and joystick; an arrow labelled “commands, by cable or radio” runs to the right towards a small tracked robot with a camera and a gripping arm; a second arrow labelled “camera view and sensor readings” runs back from the robot to the screen
Notice that information travels both ways. Commands go out to the robot, and pictures and sensor readings come back. Without that return path the operator would be steering blind.
Why hand the decisions to a person
Manual control is not a backward choice. It is chosen on purpose whenever the situation is one a program cannot be trusted with.
The first reason is an unpredictable job. No two collapsed walls look alike, and no two suspicious packages are packed the same way. A person can look at the picture and change the plan on the spot.
The second reason is responsibility. In surgery, a robot such as the da Vinci system holds the instruments and steadies them, but a surgeon sitting at the console moves them. The skill and the answerability stay with the doctor.
The third reason is distance combined with danger. Astronauts operate the robotic arm outside the International Space Station from inside the station. The arm goes where a person in a spacesuit would be at risk, while the decisions stay with the crew.
Manual control has its own costs. One trained operator is tied to one robot for the whole task, and tired operators make mistakes. The link can be lost — a cable snags, a radio signal is blocked by a wall — and a robot that has lost contact can do nothing but wait. And however fast the link, the human eye and hand cannot react in a few thousandths of a second, which is what some jobs need.
3.3.2 Automatic Robots
A floor-cleaning robot is left running in a hall and the family goes out. It rolls forward, meets a chair leg, backs off, turns, and carries on cleaning. Nobody is holding a joystick. Nobody is even in the room.
Robots that work like this are called automatic robots.
It is worth being clear about what “chooses” means here. The robot is not thinking. It is comparing a number from a sensor against a limit that a programmer wrote down, and then running whichever set of instructions matches. Every rule it applies was put there by a person, in advance.
Two levels of automatic control
Not every automatic robot is left completely to itself. There are two levels, and the difference between them is how much a human is still involved.
Semi-autonomous robots handle the routine part of the work themselves but keep a person watching, ready to step in. A survey drone flies a stored pattern over a field on its own, and the pilot standing at the edge of the field takes over when a bird crosses the path or the wind picks up. The robot handles the ordinary; the person handles the unusual.
A cobot, short for collaborative robot, is another machine of this kind. Cobots are built to work in the same space as people rather than behind a safety fence. They run their task on their own, they are made to move gently and to stop when they touch something, and a worker beside them guides and corrects them through the job.
Fully autonomous robots work without anyone watching over them at all. Once started, they sense, apply their rules and act, for the whole task. A floor-cleaning robot is the everyday example. Industrial robots that repeat the same welding or packing operation all day are another, and a self-driving car is an attempt at a much harder one.
Fig. 3.3-3 places all three kinds of control on a single scale.
Diagram
Fig. 3.3-3 – a horizontal arrow labelled at the left end “every move decided by a person” and at the right end “no person involved once started”; three markers along it, from left to right — Manual (bomb-disposal robot on a joystick), Semi-autonomous (survey drone flying a stored route with a pilot watching), Fully autonomous (floor-cleaning robot); above the arrow, a caption reading “the amount of human decision falls from left to right”
What automatic control buys, and what it costs
Handing the decisions to the program has clear gains. One person can look after many robots instead of one. The robot reacts in a fraction of a second, faster than any human hand. It works through the night without tiring, and it repeats the same movement to the same accuracy the thousandth time as the first.
The cost is that the robot can only handle what its rules cover. Show a cleaning robot a spilt glass of milk and it will drive through it, because “wet patch” was never written into its rules. When something unexpected happens, an automatic robot does not stop and wonder — it applies the nearest matching rule, which may be the wrong one.
3.4 Bio-Inspired Robots
An engineer trying to build a machine that walks over loose rocks has a hard problem. A goat solves the same problem every day, on a hillside, without a single wheel.
That thought is behind a whole family of robots. Instead of starting from a blank sheet, the designer starts by looking closely at a living thing that already handles the situation well — how it is shaped, how it moves, how it senses what is around it — and then builds those ideas into a machine.
This is a different kind of description from the two you have already met. Terrain tells you where a robot works. Control tells you who decides its moves. This third description tells you where the design idea came from.
It is worth saying plainly what such a robot is not. A machine shaped like a snake is not alive, does not grow, and does not feel anything. It borrows a shape and a way of moving, nothing more. Copying part of a living thing is a design method, not a step towards making one.
Nature is a useful place to borrow from for a simple reason: living things have been shaped over an enormous stretch of time by the very conditions the robot will have to face. A fish is the shape it is because that shape moves through water at a low cost in energy. A gecko can hold on to a smooth wall. A bird’s wing works in gusty air. These are answers to engineering questions, already tested.
The three sub-sections that follow take this idea in order: what these robots are and what designers copy from nature, why they are built at all, and which ones have actually been made.
3.4.1 What Bio-Inspired Robots Are
Look at a housefly on a wall and then at a small drone. The fly weighs almost nothing, lands upside down, and takes off before your hand arrives. No machine of that size does all three. A designer who wants a machine to do even one of them will start by studying the fly.
Robots built this way are called bio-inspired robots. “Bio” means life, so the name simply says that the design idea came from a living thing. They are also called biomimetic robots, from a word meaning “imitating life”.
The word to hold on to is “copied”. A robot cheetah is metal, plastic and batteries. It has no muscle, no blood and no bones. What has been carried across from the animal is a shape and a pattern of movement, and nothing else.
Designers also copy selectively. They study one feature that solves their problem and leave the rest of the animal alone. A wall-climbing robot borrows a gecko’s foot without borrowing its tail, its eyes or its skin.
What is copied from nature
Ideas are usually taken from one of four things, shown in Fig. 3.4-1.
Diagram
Fig. 3.4-1 – four panels in a row, each with a living thing on the left and the robot part it inspired on the right; panel 1, a running four-legged animal beside a four-legged robot, labelled “Locomotion”; panel 2, a gecko’s foot beside a flat gripping pad, labelled “Body shape and materials”; panel 3, a cat’s whiskers beside a thin touch sensor on a robot’s body, labelled “Sensing”; panel 4, a line of ants beside a group of small identical robots, labelled “Group behaviour”
Locomotion. Locomotion means the way a living thing moves from place to place — legs, wings, fins, a wriggling body. This is the most copied idea of all, because animals get across broken ground, air and water using far less energy than most machines do.
Body shape and materials. An elephant’s trunk bends anywhere along its length and still lifts a heavy log. A gecko’s foot sticks to smooth glass and lets go again without glue. Such features suggest new shapes and new soft materials for robot bodies and grippers.
Sensing. Animals sense things our instruments handle poorly. A cat reads a narrow gap with its whiskers in complete darkness; a bat finds a moth by listening to its own echoes. Robots have been given whisker-like touch sensors and echo-based sensing for the same reasons.
Group behaviour. A single ant does very little. A colony of them finds food, carries it home and builds a nest, with no ant in charge. Robots built on this idea come in large numbers, each one simple and cheap, and they get a large job done together instead of one clever machine doing it alone.
Copying nature has a real limit, and it is honest to state it. Living muscle, bone and skin are still far ahead of the motors, metals and plastics available to us. A robot built to run like a cheetah is heavier, noisier and clumsier than the animal, and it runs for minutes rather than a lifetime. Bio-inspired robots take an idea from nature. They do not match it.
3.4.2 Why Bio-Inspired Robots Are Built
Copying an animal is not easy. Legs need more motors than wheels, a flapping wing is harder to build than a propeller, and a soft body is harder to control than a rigid one. So a designer only takes this road when there is something worth gaining. There are six such gains, and each answers a real problem.
Getting more out of less
The first is efficiency. Efficiency means getting the job done with less energy or less effort wasted. Animals are under constant pressure to be efficient, because an animal that wastes energy has to find more food. A fish that glides through water with a slow beat of its tail is showing an answer that took no engineer any effort to invent, and a robot shaped that way needs a smaller battery for the same distance.
The second is working in natural surroundings. Factories are built flat, square and predictable, and wheels do well there. Forests, riverbeds and earthquake rubble are none of those things. The animals that live in such places move through them all day, so their way of moving is a sensible starting point for a machine that must do the same.
The third is toughness. Toughness here means being able to take damage and keep going. Step on a beetle’s leg and it limps on with five. Most machines are not like that: one broken part and everything stops. Copying the way a living body spreads its work across many similar parts gives a robot the same ability to carry on, as in Fig. 3.4-2.
Diagram
Fig. 3.4-2 – a six-legged robot walking; one leg is drawn shaded and broken, with a small caption “leg damaged”, while the remaining five legs are shown in contact with the ground and an arrow shows the robot still moving forward
The fourth is better sensing. Some animals sense things our instruments handle poorly — a bat finding a small moth in the dark, a cat reading a narrow gap with its whiskers. Where a job needs that kind of sensing, copying the animal’s method is often easier than inventing a new one.
The fifth is blending in. A robot shaped and coloured like a fish can swim close to a shoal without scattering it, which is exactly what a researcher studying that shoal needs.
The sixth is treading lightly. A small, light, animal-like machine can move over a field, a reef or a nesting site without crushing what is there, which a heavy wheeled vehicle would.
3.4.3 Examples of Bio-Inspired Robots
The ideas in the last two sub-sections have been built and tested. Four families of bio-inspired robots are worth knowing: humanoids, birds, snakes and insects.
Humanoids
A humanoid is a robot built in the general shape of a human body — a torso, two legs, two arms, and a head carrying its cameras.
The reason for this shape is not that it looks friendly. It is that the world around us was built for our shape. Stairs are made for legs of a certain length, door handles for hands at a certain height, tools for a grip like ours. A machine that has to work in a house, a hospital or an office can either have all of that rebuilt for it, or it can be given a body that fits what is already there.
Honda’s ASIMO, one of the best-known humanoid robots, could walk, climb a staircase and run. Boston Dynamics’ Atlas is a later two-legged machine that walks over uneven ground and keeps its balance when pushed. See Fig. 3.4-3.
Diagram
Fig. 3.4-3 – photo of a two-legged humanoid robot standing upright, with its head, arms and jointed legs clearly visible
Balance is the difficulty. A body standing on two legs is always close to falling, so a humanoid has to correct itself many times a second just to stay upright. That takes many motors, a great deal of computing, and a large share of its battery — before it has done any useful work at all.
Birds
Bird-inspired robots fly by flapping and twisting their wings instead of spinning a propeller. Festo’s SmartBird, modelled on a gull, is the well-known example: it takes off, flies and lands using wings that beat and twist much as a real bird’s do.
A flapping wing has advantages a propeller cannot offer. It is quieter, which matters when studying nervous wildlife, and it handles gusty air near buildings and trees better. It is also far harder to build, because a wing has to change its shape through every stroke, and it carries less weight than a rotor of the same size.
Snakes
A snake has no legs and no wheels, yet it gets through a gap barely wider than its own body. That is precisely what a rescue team needs when someone is trapped under a collapsed slab.
Snake robots are built as a chain of short segments joined end to end, each joint driven by its own small motor. Running a wave of bending along that chain pushes the robot forward, and the same chain can climb a pipe or wrap around a bar. Modular snake robots developed at Carnegie Mellon University have been taken to disaster sites to search through rubble that no wheeled machine could enter. Fig. 3.4-4 shows the arrangement.
Diagram
Fig. 3.4-4 – a snake robot drawn as eight short cylindrical segments joined by jointed connectors, curved into an S shape, with one joint enlarged in a small circle to show the motor inside; an arrow along the body is labelled “a wave of bending travels from head to tail”
Being thin is also the weakness. A snake robot is slow, it has little room inside for a battery, and with its body pressed against dirt and broken concrete its joints wear quickly.
Insects
Insects are copied for two different reasons: how they run and how many of them there are.
For running, there is the VelociRoACH, a small six-legged robot built at the University of California, Berkeley and modelled on a cockroach. It is fast for its size and it rights itself and carries on after being knocked about, which is exactly the behaviour that makes cockroaches so hard to catch.
For numbers, there are the RoboBees developed at Harvard University: flapping-wing robots small and light enough to be measured in fractions of a gram. Kilobots, also from Harvard, are small, cheap ground robots meant to be used by the hundred, so that a group of them can carry out a task together in the way a colony of insects does. Robots used in large groups like this are called swarm robots.
Very small robots run into a very stubborn problem: power. There is almost no room for a battery, so many insect-sized flying robots have been flown with a thin wire supplying power from outside. Cutting that wire, and giving such a machine its own power, is still a live research problem.
Where these robots are put to work
Across all four families, the same fields keep appearing: search and rescue, agriculture, underwater exploration, monitoring and surveillance, environmental monitoring, medical robotics, and aerospace. What these have in common is that the work happens in natural, awkward or delicate surroundings — the very places where copying a living thing pays off.
3.5 One Robot, Two Labels
A single person can be a cricketer and a cyclist, and there is nothing odd about that. The two words answer different questions.
The groups in this chapter work the same way. Terrain answers where does this robot work. Control answers who decides its moves. Every robot therefore carries both labels at once, and asking which of the two groups a robot “really” belongs to is asking a question that has no answer.
Take the machines you have already met in this chapter and give each of them both labels, as in Table 3.4.
Read a row across and you learn far more than either column would tell you alone. “Underwater” and “manual” together tell you there is a cable, a ship above, and a person watching a screen. “Underwater” and “automatic” tell you there is no cable and no live picture at all.
The two labels are not equally fixed. Terrain is settled the day the robot is built — a sealed, thick-walled underwater body cannot be turned into a flying machine. Control is often a setting. Many drones are flown by hand for take-off, switched to a stored route for the survey, and taken back under hand control for landing. The same machine is manual, then automatic, then manual again, within one flight.
This is worth remembering the next time you meet an unfamiliar robot. Three short questions will place it: Where is it built to work? Who decides its next move? And was any part of its design taken from a living thing?
3.6 Summary
- Classification is the arranging of robots into groups whose members share a chosen feature. It helps us choose a robot for a job, compare robots that face the same problems, and describe a robot in a few words.
- Robots are classified in two main ways: on the basis of field or terrain, and on the basis of control.
- Terrain decides what holds a robot up, what it pushes against to move, and how its body must be protected. On this basis, robots are ground robots, aerial robots or underwater robots.
- A ground robot works on the surface of the earth or on any other solid surface. Since the surface holds it up, it can afford to be heavy.
- Ground robots move in four ways: wheeled, tracked, legged, and wheeled-legged hybrid. Wheels suit flat hard floors, tracks suit mud and rubble, legs cross steps and gaps, and hybrids do both.
- Ground robots are used in manufacturing, agriculture, search and rescue, space exploration, and surveillance and security. Their limit is that they go only where the ground goes.
- An aerial robot flies and works in the air with no pilot on board. Such robots are also called unmanned aerial vehicles (UAVs) or drones.
- The common aerial designs are quadcopters, fixed-wing drones, hexacopters and octacopters, and VTOL (Vertical Takeoff and Landing) drones.
- A quadcopter can hover and turn on the spot; a fixed-wing drone cannot hover but covers far more distance on the same battery.
- Aerial robots are used for photography and videography, mapping and surveying, agriculture, environmental monitoring, search and rescue, and infrastructure inspection. Short battery life, small payload and wind are their main limits.
- An underwater robot works below the surface of water. It uses thrusters to move, carries its own lamps, and needs a sealed body because pressure rises by about one atmosphere for every 10 metres of depth.
- Underwater robots are of two types: remotely operated vehicles (ROVs), which are tethered by a cable and driven by a person, and autonomous underwater vehicles (AUVs), which carry no cable and follow a stored plan.
- Underwater robots are used in marine research, oceanography, environmental monitoring, underwater archaeology, offshore industries, and search and recovery.
- Control means deciding what a robot’s next action will be. Sensing and acting are always done by the robot; the deciding may sit with a person or with the robot’s program.
- A manual robot has every action commanded by a human operator. Such robots are also called tele-operated robots, and they still carry sensors — the readings simply go to a screen for a person to read.
- Manual control is chosen for unpredictable jobs, for work where responsibility must stay with a person, and for dangerous places. It ties one operator to one robot and fails if the link is lost.
- An automatic robot chooses its next action by applying stored rules to its sensor readings. Semi-autonomous robots do the routine part while a person watches; fully autonomous robots work without anyone watching. A cobot is a collaborative robot built to work safely beside people.
- Automatic control saves human effort and reacts quickly, but a robot can only handle situations its rules already cover.
- A bio-inspired robot, also called a biomimetic robot, copies the structure, movement or behaviour of a living thing. Designers copy locomotion, body shape and materials, sensing, and group behaviour.
- Bio-inspired robots are built for efficiency, for working in natural surroundings, for toughness, for better sensing, for blending in, and for treading lightly on the places they work in.
- Well-known examples are humanoids, bird-inspired flapping-wing robots, snake robots, and insect-inspired robots including swarm robots used in large numbers.
- Terrain and control answer different questions, so every robot carries both labels at once. Terrain is fixed when the robot is built; control is often just a setting that can be changed during a task.
3.7 Key Terms
- Aerial robot — a robot built to fly and work in the air, carrying no pilot on board.
- Automatic robot — a robot that chooses its next action itself, by applying stored rules to its sensor readings, without a person commanding each move.
- Autonomous underwater vehicle (AUV) — an underwater robot with no cable, which carries its own battery and follows a plan stored in it before the mission.
- Bio-inspired robot — a robot whose structure, movement or behaviour is copied from a living thing.
- Biomimetic robot — another name for a bio-inspired robot, from a word meaning “imitating life”.
- Classification — the arranging of robots into groups, where all the robots in one group share a chosen feature.
- Cobot — short for collaborative robot; a robot built to work safely in the same space as people, guided by a worker beside it.
- Control — deciding what a robot’s next action will be, either by a human operator or by the robot’s own program.
- Drone — the everyday name for an aerial robot.
- Field or terrain — the surface or space a robot is built to work in, such as ground, air or water.
- Fixed-wing drone — an aerial robot held up by air flowing over a wing, which must keep moving forward and cannot hover.
- Fully autonomous robot — an automatic robot that works for the whole task with nobody watching over it.
- Ground robot — a robot built to move and work on the surface of the earth or on any other solid surface.
- Hexacopter and octacopter — aerial robots with six and eight rotors, able to carry heavier loads and to stay up if one motor stops.
- Humanoid — a robot built in the general shape of a human body, with a torso, two legs, two arms and a head.
- Legged robot — a ground robot that moves on jointed legs, so that it can step over obstacles and gaps.
- Manual robot — a robot whose every action is commanded by a human operator while the robot is working.
- Quadcopter — an aerial robot with four rotors, able to hover and turn on the spot.
- Remotely operated vehicle (ROV) — an underwater robot joined to a ship by a cable, driven in real time by a person watching its camera.
- Semi-autonomous robot — an automatic robot that handles the routine part of a task itself while a person watches and takes over when something unusual happens.
- Swarm robots — robots used in large numbers, each one simple and cheap, that carry out a task together.
- Tele-operated robot — another name for a manual robot, since the operator is usually at a distance from the machine.
- Thruster — a small propeller that pushes water backwards so that an underwater robot moves forwards.
- Tracked robot — a ground robot that moves on continuous belts, which spread its weight and grip soft or broken ground.
- Underwater robot — a robot built to work below the surface of water, such as in a sea, lake or river.
- Unmanned aerial vehicle (UAV) — the formal name for an aerial robot, meaning an aircraft with nobody on board.
- Unmanned underwater vehicle (UUV) — another name for a remotely operated vehicle.
- VTOL (Vertical Takeoff and Landing) drone — an aerial robot that lifts off straight upwards on rotors and then tilts into forward flight on a wing.
- Wheeled robot — a ground robot that moves on wheels, best suited to flat, hard and smooth surfaces.
- Wheeled-legged hybrid robot — a ground robot with legs ending in wheels, which rolls on smooth ground and walks on rough ground.
3.8 Objective Questions
An aerial robot that lifts off straight upwards on rotors and then tilts into forward flight on a wing is called a:
- quadcopter
- fixed-wing drone
- VTOL drone
- octacopter
Which one of these is a bio-inspired robot?
- A wheeled trolley robot carrying cartons in a warehouse
- A segmented robot that bends its way through rubble
- A tethered robot inspecting an underwater pipeline
- A robotic arm welding car doors on a production line
Fill in the blanks.
- An underwater robot joined to a ship by a cable and driven by a person on board is called a ______________.
- In sea water, the pressure on a robot rises by about one atmosphere for every ______________ metres of depth.
- A robot built in the general shape of a human body is called a ______________.
State whether each statement is true or false.
- A manual robot carries no sensors.
- A fixed-wing drone can hover over one spot.
- The terrain group of a robot is fixed when the robot is built.
Match each item in Column A with the correct description in Column B.
Column A Column B a. Tracked robot i. Pushes water backwards to drive the robot forward b. Thruster ii. Works in large numbers, each machine simple and cheap c. Cobot iii. Spreads the robot’s weight so it does not sink into mud d. Swarm robots iv. Works beside people in a shared space, guided by a worker Name the two bases on which robots are classified in this chapter.
A machine with a camera and a gripping arm is driven along a corridor by an officer using a joystick. Give its terrain group and its control group.
3.9 Concept Questions
Define a ground robot, and state one advantage it has over an aerial robot of the same size.
Differentiate between a wheeled robot and a legged robot on any three points.
Explain why an underwater robot cannot use satellite positioning once it is well below the surface, and state how such a robot works out where it is instead.
Automatic robots never tire and react faster than a person. Give a reason why a surgeon still controls a surgical robot by hand.
An autonomous underwater vehicle is working at a depth of 220 metres in sea water. Taking the pressure as rising by about one atmosphere for every 10 metres of depth, and remembering the air pressing down on the sea, work out the total pressure on the robot’s body in atmospheres. Show your working.
A small ground robot has one distance sensor pointing forwards and one motor on each wheel. Write, in numbered steps, the rules it should follow so that it keeps moving forward and turns away whenever something comes closer than a safe limit.
3.10 Case-Based Questions
A school on a hillside has two inspection jobs to get done. The outside walls of the three-storey building have to be checked for cracks, and the old open well in the school compound has to be checked for damage below the water line. The school decides to hire a robot for each job rather than send a person up a ladder or down into the well.
- Name the terrain group of the robot suited to each of the two jobs, and give one reason for each choice.
- For the well inspection, would manual or automatic control be the better choice? Give one reason.
- State one limitation of the robot chosen for the wall inspection.
A farmer wants to spray a liquid fertiliser over a large field, and also to pull out weeds growing between the crop rows. A dealer offers three machines, described in the table below.
Machine Where it works Working time on one charge Load it can carry P In the air, on four rotors 25 minutes 5 kg Q In the air, on a wing; cannot hover 60 minutes 2 kg R On the ground, on four wheels 4 hours 60 kg - Name the terrain group of each machine.
- Which machine would you choose for spraying the fertiliser? Use two figures from the table to support your answer.
- Which machine would you choose for pulling out weeds, and why are the other two unsuitable for that job?