← Back to Subjects ororo.world

Chapter 6: Identification of Robots

The Class 9 group from a school in Jaipur spent a morning at a robotics exhibition. In the first hall stood a bright orange arm bolted to a steel table. It had no head, no legs and no face. It simply swung, stopped, picked up a small metal block and set it down again, over and over, without ever moving from its spot.

In the next hall was a robot shaped like a person. It stood on two legs, turned its head towards whoever came close, and lifted a hand to wave. Behind it, on the floor, a low box on four wheels rolled quietly between the visitors, slowing down whenever someone stepped in front of it.

Three machines. Three completely different shapes. Yet each one was a robot, and each one was built out of the same kinds of parts. By the end of the morning the students could look at any of them and name what they were seeing.

Diagram

Fig. 6.0-1 – photo of a robotics exhibition hall showing an industrial robotic arm on a table, a humanoid robot standing nearby, and a small wheeled mobile robot on the floor

NoteCuriosity Corner

Q1. Why do robots look so different from one another? Q2. What is the part at the end of a robot arm that actually picks things up called? Q3. How does a robot arm know where its own joints are at any moment? Q4. Why do some mobile robots run on tracks instead of wheels? Q5. What does a walking humanoid robot need that a robot arm bolted to the floor does not?

This chapter will answer each of these five questions as you read through it.

6.1 A Robot Can Be of Any Size and Shape

Look at the tools in a kitchen. A spoon, a rolling pin and a pair of tongs are all made of metal, and all of them are held in the hand. But no two of them look alike, because no two of them do the same job. The shape of each tool comes from the work it has to do.

Robots are the same. A robot that welds car doors in a factory does not need legs. A robot that carries medicines down a hospital corridor does not need a face. A robot that has to climb a staircase cannot be a flat box on wheels. Once you know the job, the shape starts to make sense.

What Decides a Robot’s Shape

Three questions decide how a robot is built.

The first is the task — what exactly must the robot do? Lifting a heavy engine block needs a thick, strong arm. Picking up an egg needs soft, gentle fingers.

The second is the workspace — the region the robot can reach and work in. A robot arm fixed to a table can only work inside the space its arm can sweep through. A delivery robot’s workspace is the whole corridor, so it must be able to move about.

NoteDefinition

(The workspace of a robot is the region of space that the robot can reach and work in.)

The third is the environment and the difficulties it brings. A robot working underwater must be sealed against water and must be able to swim or crawl. A robot working on a farm has to deal with mud, uneven ground and rain. A robot working in a clean electronics factory has none of those problems, so it can be lighter and more delicate.

NoteExam Corner

Q. Two robots do exactly the same job — sorting parcels — but one works in a warehouse and one works on a ship’s deck. Will they look the same?

A. Probably not. The job is the same, but the environment is different. The one on the deck must handle a moving, wet, sloping surface, so it will need a steadier base and better protection from water.

Because of these three questions, robots come in an enormous range of sizes and shapes. Some are small enough to be held in one hand. Some are large enough to fill a room. Some look like a human being, some like an insect, some like an aeroplane, and many look like nothing at all except a machine built for one job.

Same Parts, Different Bodies

Here is the useful part. However different two robots look on the outside, they are built from the same short list of parts: a body that holds everything together, actuators that move it, sensors that tell it about its surroundings, a control system that works out what to do next, and a power supply that keeps it all running.

So identifying a robot is not about guessing what it is from its outline. It is about looking at a machine and naming the parts you can see, and saying what each one is there for. See Fig. 6.1-1.

Diagram

Fig. 6.1-1 – three robot outlines side by side, all drawn in the same simple style: (a) an arm bolted to a table, (b) a two-legged humanoid standing upright, (c) a low four-wheeled box. Under each, the same four labels appear: body/structure, actuators, sensors, control system

NoteCuriosity Corner

Q1. Why do robots look so different from one another?

A. Because a robot’s shape is decided by its task, its workspace and the environment it must work in. A robot is built in whatever form lets it do its job, so a welding robot, a walking robot and a delivery robot end up looking nothing like each other — even though all three are built from the same kinds of parts.

In the next three sections you will look closely at three very different robots — an industrial robotic arm, a humanoid robot and a mobile robot — and learn to name what each one is made of.

6.2 Identifying an Industrial Robotic Arm

Think about how you pick a book off a high shelf. Your shoulder swings the whole arm up, your elbow bends, your wrist turns so your palm faces the book, and your fingers close around it. Five or six separate movements, one after another, and the book is in your hand.

A factory needs that same set of movements, but repeated thousands of times a day without getting tired. That is what an industrial robotic arm is built for. It is a robot shaped roughly like a human arm, fixed to a floor, a table or an overhead rail, and used for jobs such as welding, painting, assembling parts, packing boxes and moving material from one machine to another.

NoteDefinition

(An industrial robotic arm is a robot built as a chain of moving segments, fixed in one place, and used to carry out repeated tasks in a factory or workshop.)

It is worth noticing what this robot does not have. It has no legs, no wheels and usually no face. It never goes anywhere. Everything it needs is brought to it on a conveyor, and it works only inside the space its arm can sweep through.

NoteExam Corner

Q. If the arm cannot move from its place, how does it work on a car that is much longer than the arm?

A. It doesn’t work on the whole car at once. The car moves past on a conveyor line, and several arms are placed along the line, each doing its own part of the job as the car reaches it.

You have probably seen these arms in videos of car factories — heavy, brightly painted, moving in smooth curves. See Fig. 6.2-1.

Diagram

Fig. 6.2-1 – photo of an industrial robotic arm on a factory floor, mounted on a fixed base, with a gripper attached at the far end

What to Look for on an Arm

When you look at a real industrial arm, or at a photograph or video of one, you can identify eight things. The next few sections take them one at a time:

1. the arm structure — the links and joints that make up its body
2. the end effector — the tool at the far end that touches the work
3. the actuators — the motors that move each joint
4. the control system — the hardware and software that decide the movement
5. the sensors — the parts that report back what is happening
6. the safety features — the parts that protect the people around it
7. the programming — how the arm is told what to do
8. the integration — how the arm is fitted into the rest of the factory

The first six are things you can point to. The last two are things you cannot see by looking, but without them the arm would just stand still.

6.2.1 Arm Structure: Links, Joints and Axes

Hold your arm straight out and try to touch your own shoulder. You cannot do it, because a straight arm has nowhere to bend. Now bend your elbow and it takes half a second. A rigid bar can only point; it is the bending places that let an arm reach things.

A robot arm is built on exactly that idea. It is a chain of stiff pieces with movable connections between them.

Each stiff piece is called a link. A link does not bend or stretch — it is usually a hollow casting of steel or aluminium, strong enough to carry the weight of everything beyond it. The bones of your arm do the same job.

Each movable connection between two links is called a joint. The joint is the only place where movement happens. Your shoulder, elbow and wrist are joints.

NoteDefinition

(A link is a rigid segment of a robot’s body. A joint is the movable connection between two links, and it is where the movement takes place.)

Two Ways a Joint Can Move

Robot joints move in one of two basic ways, and you can usually tell which by watching the arm work. See Fig. 6.2.1-1.

Diagram

Fig. 6.2.1-1 – two simple sketches side by side. Left: two links joined at a rotary joint, with a curved arrow showing one link turning about the joint. Right: two links joined at a linear joint, with a straight arrow showing one link sliding in and out of the other

NoteCompare Table

Table 6.1 – Rotary joint vs Linear joint | Feature | Rotary joint | Linear joint | |—|—|—| | How it moves | One link turns about the joint | One link slides straight in or out | | Everyday example | A door turning on its hinge | A drawer sliding out of a cupboard | | Path the far end takes | A curve | A straight line | | Where it is used on arms | Shoulder, elbow and wrist movements | Up-and-down or in-and-out reach |

Most industrial arms use rotary joints for nearly everything, because turning motors are simple, compact and reliable.

Counting the Axes

Every joint gives the arm one independent movement, and that movement is often called an axis. An arm with six joints is therefore called a six-axis robot, and six-axis arms are the most common kind in industry.

The number of independent movements a robot can make is called its degree of freedom (DOF). A six-axis arm has six degrees of freedom. Roughly speaking, three of them are used to bring the tool to the right place, and three to turn the tool to the right angle once it gets there.

NoteExam Corner

Q. Why not simply build an arm with twenty joints, so it can reach anywhere at all?

A. Every extra joint adds a motor, more weight for the joints below it to carry, more cost and more places where small errors can build up. Designers use the smallest number of joints that will do the job.

That is the honest limitation of arm structure: more joints mean more reach and more awkward angles the arm can manage, but also a slower, heavier and more expensive robot. A machine that only has to lift boxes from a belt onto a pallet may need just four axes, and building it with six would be a waste.

Reading an Arm from a Photograph

You can identify the structure of an arm in a picture without any special knowledge. Start at the base, which is bolted down, and follow the arm outwards. Each time the metal shape changes direction or there is a thick round housing, you have found a joint. The straight sections between them are the links. Count the joints, and you have counted the axes.

6.2.2 The End Effector

The links and joints of an arm can bring the far end of the arm to any point in the workspace, at any angle. That is all they can do. Getting to the right place is useless if there is nothing there to hold the object, weld the seam or spray the paint.

So a tool is fitted at the far end of the arm. This tool is called the end effector, because it is the part that produces the effect the robot was bought for. It is sometimes called the robot’s hand, and it is the only part of the arm that actually touches the workpiece.

NoteDefinition

(The end effector is the tool fitted at the far end of a robot arm. It is the part that interacts directly with the workpiece and carries out the actual task.)

One Arm, Many Possible Tools

The same arm can do completely different jobs simply by changing what is bolted on the end. A few common end effectors are shown in Fig. 6.2.2-1.

Diagram

Fig. 6.2.2-1 – four end effectors drawn side by side and labelled: (a) a two-finger gripper closing on a block, (b) a set of vacuum cups holding a flat sheet, (c) a welding torch with a nozzle, (d) a cutting tool

A gripper has two or more fingers that close on an object and hold it while the arm moves. A set of vacuum cups presses against a flat surface and holds it by suction — this is how sheets of glass, cardboard boxes and packets are lifted. A welding torch joins two metal pieces along a seam. A cutting tool trims material to shape. Paint sprayers, screwdrivers and glue nozzles are used in the same way.

NoteHow It Works

A two-finger gripper What it is: A simple end effector with two flat fingers that open and close, like a pair of tongs. How it works: A small motor or a compressed-air cylinder drives the two fingers towards each other until they press on the object. The friction between the fingers and the object holds it while the arm moves. Used for: Picking metal blocks off a conveyor and placing them into a machine. Watch out: The fingers are made for one shape and size. Give the same gripper a round bottle or a soft fruit and it will either fail to grip or squash it.

NoteCommon Error

assuming every end effector is a gripper What goes wrong: Students often describe the end effector as “the robot’s fingers”, so a welding arm or a paint-spraying arm seems to have no end effector at all. The fix: The end effector is whatever tool is fitted at the end of the arm. A gripper is only one kind. A torch, a spray nozzle or a cutter is just as much an end effector.

NoteExam Corner

Q. A factory arm has to lift both cardboard cartons and loose steel bolts. Can one end effector do both?

A. Not well. Cartons are flat and light and suit vacuum cups; bolts are small and irregular and suit a gripper or a magnet. Factories usually fit a tool changer so the arm can swap end effectors, or they use two arms.

NoteSafety Note

An end effector is never changed while the arm is live. The power must be switched off and, on an air-driven gripper, the compressed air supply must be disconnected and released first. A gripper that closes while a hand is between its fingers can cause a serious injury.

NoteCuriosity Corner

Q2. What is the part at the end of a robot arm that actually picks things up called?

A. It is called the end effector. A gripper is the type of end effector that picks things up, but the same position on the arm may instead carry a welding torch, vacuum cups, a spray nozzle or a cutter, depending on the job.

6.2.3 Actuators and the Control System

A joint is only a place where movement is possible. Nothing about a hinge makes a door swing; someone has to push it. In the same way, a robot’s joints will not move unless something supplies the force.

The part that supplies that force is called an actuator. On an industrial arm, each joint has its own actuator, sitting right at the joint or driving it through a gear.

NoteDefinition

(An actuator is the part of a robot that converts energy into movement, so that a joint or a tool actually moves.)

An actuator at a joint has to supply enough turning force to swing everything beyond that joint, including the end effector and whatever it is carrying. Turning force is called torque — the same thing you apply when you use a spanner. This is why the shoulder joint of an arm, which carries the whole arm’s weight, has a much larger actuator than the wrist joint.

Three Kinds of Actuator

Industrial arms use electric, pneumatic or hydraulic actuators. Table 6.2 compares them.

NoteCompare Table

Table 6.2 – Electric vs pneumatic vs hydraulic actuators | Feature | Electric | Pneumatic | Hydraulic | |—|—|—|—| | What drives it | Electric current through a motor | Compressed air | Oil under pressure | | Control over position | Fine, can stop anywhere | Coarse, usually only at the two end points | Good, but needs a pump and pipes | | Force it can give | Small to large | Small | Very large | | Everyday comparison | A ceiling fan motor | A cycle pump pushing a piston | A car’s hydraulic brakes |

Electric actuators are by far the most common, because they can be started, stopped and held at any position with great precision, and they need only a cable. Pneumatic actuators are cheap and fast but usually just push all the way out or pull all the way back, which suits a simple gripper. Hydraulic actuators are chosen when the load is very heavy.

Electric actuators come in a few types you will meet again in the laboratory: a plain DC motor, a geared DC motor that trades speed for torque, a stepper motor that turns in small fixed steps, and a servo motor that can be told to move to a particular angle and stay there.

NoteHow It Works

Servo motor What it is: An electric motor with a built-in position sensor and a small circuit that watches the position for it. How it works: The circuit is given a target angle. It compares the target with the angle the motor is actually at, and keeps the motor turning until the two match. If something pushes the arm away from that angle, the motor turns again to bring it back. Used for: Holding a robot joint steady at a set angle while the arm carries a load. Watch out: A servo can only correct as far as its motor is strong. Overload it and it will sit humming at the wrong position.

NoteExam Corner

Q. An arm is holding a heavy box in mid-air, perfectly still. Are its actuators doing anything?

A. Yes. The joints would sag under the weight if nothing resisted it, so the actuators are still supplying torque even though nothing is moving. Standing still is work for a robot.

The Control System

Actuators supply force, but they do not know when or how much. Something has to decide that. That job belongs to the control system, which is best described as the brain of the arm — not because it thinks, but because every decision about movement is worked out there.

The control system has two halves.

The hardware half is the controller — a dedicated computer or an embedded circuit board kept in a cabinet beside the robot, along with the input and output connections that carry signals to the actuators and bring readings back from the sensors.

The software half is a set of instructions, called control algorithms, that work out the movement. Given a target position for the end effector, the software calculates how far each joint must turn, how fast each actuator should run, and in what order, so that the arm moves smoothly and safely instead of jerking to a stop. See Fig. 6.2.3-1.

Diagram

Fig. 6.2.3-1 – block sketch of an arm’s control loop: a box labelled “controller” with an arrow out to boxes labelled “actuator at each joint”, an arrow from those to a box labelled “arm moves”, and a dashed arrow from “sensors” back into the controller

NoteExam Corner

Q. Why does the control system need to know how fast a joint should move, and not just where it should end up?

A. Because a heavy arm that starts and stops suddenly shakes itself and whatever it is holding. The software plans a smooth build-up and slow-down of speed so the load stays put and the joints are not strained.

6.2.4 Sensors and Feedback

Close your eyes and touch the tip of your nose. You can do it, because even with your eyes shut you know where your elbow is bent to and which way your wrist is turned. Take that sense away and your hand would wander about your face.

A robot arm has the same difficulty. The controller can send a command to a joint, but sending a command is not the same as the movement happening. The load may be heavier than expected, a gear may have a little play in it, or something may be in the way. Without some way of checking, the controller would be working blind.

So the arm is fitted with sensors — parts that measure something and report it back as an electrical signal. Information that travels back from a sensor to the control system is called feedback.

NoteDefinition

(A sensor is a device that measures a physical quantity — such as position, force or distance — and turns it into a signal the control system can read.)

The Sensor That Watches Each Joint

The most important sensor on an industrial arm sits inside the joint itself. It is called an encoder, and it reports how far the joint has turned.

NoteHow It Works

Rotary encoder What it is: A small disc fixed to the turning shaft of a joint, with a ring of evenly spaced marks or slots around it, and a light source and detector on either side of the disc. How it works: As the shaft turns, each slot lets a flash of light through to the detector. The detector produces one electrical pulse per slot. The controller counts the pulses, and since it knows how many slots there are in one full turn, it can work out the angle the joint has moved through. Used for: Telling the control system the exact angle of every joint, many times a second. Watch out: A simple encoder counts movement, not absolute position. If the power is cut and the arm is nudged by hand, the count is wrong, and the arm has to be driven back to a known “home” position before it can be trusted again.

Diagram

Fig. 6.2.4-1 – sketch of an encoder disc mounted on a joint shaft, with evenly spaced slots around its edge, a light source on one side, a detector on the other, and an arrow showing pulses going to the controller

The angle is worked out from a simple relation:

\(\text{angle turned} = \dfrac{\text{pulses counted}}{\text{pulses in one full turn}} \times 360^\circ\)

Here “pulses counted” is how many pulses the controller has received since the joint started moving, and “pulses in one full turn” is a fixed number decided by how many slots the disc has.

NoteExample 6.1

finding the angle a joint has turned through

Given: An encoder disc that produces 720 pulses in one full turn. The controller counts 180 pulses while the joint moves.

Find: The angle the joint has turned through.

Working:

\(\text{angle} = \dfrac{180}{720} \times 360^\circ\)

\(\dfrac{180}{720} = 0.25\)

\(0.25 \times 360^\circ = 90^\circ\)

The joint has turned through 90°.

NoteCuriosity Corner

Q3. How does a robot arm know where its own joints are at any moment?

A. Each joint carries an encoder. As the joint turns, the encoder sends a stream of pulses to the control system, and the controller counts them to work out the exact angle of that joint. With a reading from every joint, the control system knows the position of the whole arm.

The Other Sensors on an Arm

Three more kinds of sensor are common on industrial arms.

Force or torque sensors measure how hard the arm is pressing or pulling. An arm fitting a plug into a socket needs this — it must push firmly enough to seat the plug but stop before it breaks anything.

Proximity sensors detect that an object is nearby without touching it. They are used to check that a part has actually arrived on the conveyor before the arm reaches for it.

Vision systems are cameras with software behind them, used to recognise objects and find where they are lying. A camera lets one arm handle parts that arrive at slightly different positions each time, instead of needing them placed exactly.

NoteExam Corner

Q. An arm is programmed to pick a part from a fixed spot on a conveyor. The conveyor stops early one day and the part is 5 cm short of the spot. What happens?

A. With only encoders, the arm closes its gripper on empty air, because encoders report the arm’s own joints and know nothing about the part. A proximity sensor or a camera is what tells the robot whether the part is really there.

NoteEthics Watch

A vision system on a factory floor points at the work, but it also sees the people working nearby, and its recordings can be stored. Used carelessly, the same camera becomes a way of watching employees all day. The responsible practice is to tell workers what is recorded, keep the footage only as long as the job needs it, and use it for checking the process rather than the person.

6.2.5 Safety Features, Programming and Integration

An industrial arm is strong enough to swing a load of several kilograms at head height, and it does so quickly. It has no idea that a person has walked up behind it. Everything that keeps people safe around such a machine has to be designed in.

Safety Features

Four safety features are found on almost every industrial installation. See Fig. 6.2.5-1.

Diagram

Fig. 6.2.5-1 – top view of a robot cell: a robotic arm inside a fenced enclosure, a large red emergency stop button on the outside of the fence, a gap in the fence marked “light curtain” with dashed beams across it, and a conveyor running in through the gap

An emergency stop button is a large red button, easy to reach and easy to hit with the palm of the hand. Pressing it cuts power to the actuators at once, so the arm stops wherever it is.

Protective barriers are fences or panels around the working area, so that nobody can walk into the space the arm sweeps through while it is running.

A light curtain guards the opening where parts go in and out, because that gap cannot be fenced.

NoteHow It Works

Light curtain What it is: A row of light beams sent across an opening, from a transmitter on one side to a receiver on the other. How it works: While every beam reaches the receiver, the machine is allowed to run. If a hand or body breaks even one beam, the receiver notices the missing light and the robot is stopped immediately. Used for: Guarding the entrance of a robot cell where a conveyor must pass through. Watch out: A light curtain only guards the plane it covers. Someone can still climb over or under it, so it is used together with fencing, not instead of it.

Collision detection is the fourth feature. The control system watches the current drawn by each actuator. If the arm strikes something, the actuator suddenly has to work much harder to keep moving, the current rises, and the controller stops the arm before more damage is done.

NoteSafety Note

Never step inside a fenced robot cell while the robot is powered, even if the arm is standing still — a stationary arm may simply be waiting for the next part to arrive. Power must be switched off and locked off before anyone enters, and the same applies before rewiring or adjusting anything on the arm.

Programming the Arm

An arm does nothing until it is told what to do, and there are two usual ways of telling it.

The first is teaching. An operator holds a hand-held control box, drives the arm slowly to each position needed, and saves that position. Doing this for every position in the job builds up a stored sequence, which the robot can then repeat at full speed.

NoteAlgorithm

teaching a pick-and-place sequence to an arm Input: An arm with a gripper, a part sitting at a fixed pick-up point, and an empty tray at the drop point. Output: A stored sequence the arm can repeat on its own. Steps:

1. Switch the arm to slow teaching mode.
2. Drive the arm to a safe starting position and save it.
3. Drive the arm down to the part and save this position.
4. Save a "close gripper" instruction.
5. Lift the arm clear of the pick-up point and save that position.
6. Drive the arm across to the tray and save this position.
7. Save an "open gripper" instruction.
8. Return the arm to the starting position and save it.
9. Switch to run mode and play the sequence back at working speed.

Why it works: The arm only ever repeats positions a person has already checked, so nothing in the stored sequence takes it anywhere unexpected.

The second way is offline programming. The job is written and tested as a program on a computer, using simulation software that shows the arm moving on screen. Only when the program is correct is it sent to the real robot. This is done in robot-specific languages, or in ordinary programming languages such as C++ and Python. Offline programming means the production line does not have to be stopped while the robot is being taught.

Whichever way is used, the arm can be reprogrammed later for a different task. That is exactly what makes it a robot and not a fixed-purpose machine.

NoteExam Corner

Q. Why would a factory bother with simulation software when it could just teach the arm directly?

A. Because teaching ties up the real robot and the real production line for hours. Simulation lets the program be written and its mistakes found while the line keeps running, and it also shows collisions before they happen for real.

Integration

The last part of identifying an industrial arm is the part you cannot photograph. An arm on its own is of no use; it has to be joined to everything around it. This is called integration.

Integration means building the fixtures that hold the workpiece steady, setting up the conveyors that bring parts in and carry finished work out, and writing the software that makes the arm start at the right moment — after the conveyor stops, not before. Engineers who do this work are called integrators. A great deal of the cost of putting a robot into a factory goes into integration rather than into the robot itself.

NoteTry Yourself

Watch a short video of an industrial robotic arm working in a factory. Pause it and write down how many joints you can count, and name the end effector you can see.

6.3 Identifying a Humanoid Robot

Almost everything around you was built for a human body. Stairs are the height of a human step. Door handles are at the height of a human hand. Chairs, tools, switches, buses and taps all assume a person of roughly the right shape will come along and use them.

That is the strongest reason for building a robot in the shape of a person. A machine with a human-like body can move through a building made for people, use the same tools, and be understood by anyone who looks at it. Such a robot is called a humanoid robot.

NoteDefinition

(A humanoid robot is a robot whose body is built to resemble the human body, with a head, torso, arms and usually legs.)

A humanoid does not have to be a copy of a person. Some have a head, torso and arms but roll about on a wheeled base instead of walking. Some have a plain plastic face that only suggests eyes. What makes a robot a humanoid is that its overall body plan follows a human one. See Fig. 6.3-1.

Diagram

Fig. 6.3-1 – photo of a humanoid robot standing upright, showing head, torso, two arms with hands, and legs

Humanoids are used where a robot has to work among people or speak to them — greeting visitors at a counter, guiding people at an exhibition, helping in research on walking and balance, or reaching into places built for a human worker.

NoteExam Corner

Q. A humanoid is harder to build than a wheeled robot and falls over more easily. Why build one at all?

A. Because it fits the world we already have. A wheeled robot needs ramps, wide doorways and flat floors; a walking, human-shaped robot can use stairs, ladders and ordinary hand tools without anything being rebuilt for it.

What to Look for on a Humanoid

The parts of a humanoid are the same in kind as the parts of an industrial arm, but there are more of them and they are harder to fit in, because the whole machine has to carry itself about. Ten things make up a humanoid:

1. the frame and body structure
2. the joints, which give it its degrees of freedom
3. the actuators that move the limbs
4. the end effectors and hands
5. the sensors that tell it about its surroundings
6. the control system
7. the power supply
8. the communication system
9. the safety features
10. the customisation added for its particular application

The next four sections take these in order. As you read, keep comparing them with the arm in section 6.2 — the differences are where the interesting part lies.

6.3.1 Frame, Joints and Degrees of Freedom

An industrial arm is bolted to the floor, so its weight hardly matters. A humanoid has to lift its own body every time it takes a step. Every extra kilogram of body has to be carried by its own motors, powered by its own battery, all day. This one fact shapes the whole design.

The Frame

The frame is the skeleton of the robot. It gives the machine its overall shape, holds the head, torso, arms and legs together, and provides places to mount the motors, wiring and electronics inside.

NoteDefinition

(The frame is the rigid structure of a robot that gives it its shape and holds all its other parts in place.)

Frames are usually built from aluminium, carbon fibre or plastic. Aluminium is strong and much lighter than steel, so it is used for the load-bearing parts such as the legs. Carbon fibre is stiffer still for its weight, but costly, so it appears mostly in research robots. Plastic is used for outer covers and panels, which have to keep dust out and look tidy but carry no load.

NoteExam Corner

Q. Why not make a humanoid’s frame from steel, which is stronger and cheaper?

A. Because the robot has to carry its own frame. A steel body would need far bigger motors just to lift itself, and those motors would need a bigger battery, which would add still more weight. Lightness is worth paying for.

NoteSide Note

There is a curious effect called the uncanny valley. As a robot or animated face is made to look more and more like a human, people find it more likeable — up to a point. Very close to human, but not quite right, the feeling flips to unease. This is why many humanoids are given a plainly artificial face rather than a nearly real one.

Joints and Degrees of Freedom

Where an industrial arm has around six joints, a humanoid has joints at the neck, both shoulders, elbows, wrists, waist, hips, knees and ankles — and often in each finger as well. See Fig. 6.3.1-1.

Diagram

Fig. 6.3.1-1 – outline of a humanoid robot with small circles marking the joints, labelled: neck, shoulder, elbow, wrist, waist, hip, knee, ankle

Some of these are not single joints at all. A shoulder has to let the arm swing forward, swing sideways and twist, so a robot shoulder is built as two or three separate rotary joints stacked close together. Counted properly, that shoulder gives three degrees of freedom on its own.

NoteExample 6.2

counting the degrees of freedom of one humanoid arm

Given: A humanoid arm built with a shoulder that allows three independent movements, an elbow that allows one, and a wrist that allows three.

Find: The total degrees of freedom of that arm.

Working:

shoulder = 3

elbow = 1

wrist = 3

total = 3 + 1 + 3

7 degrees of freedom for one arm.

Add the second arm, the legs, the waist and the neck, and a full humanoid may have several dozen degrees of freedom. That is why it can move in ways an industrial arm cannot — it can crouch, turn its head, and shift its weight from one foot to the other.

All that freedom is expensive. Every degree of freedom needs its own actuator, its own sensor and its own share of the control system’s attention, and it adds one more part that can fail. It also makes balance harder, because the control system has to keep track of where the whole body’s weight is sitting while all those joints are moving at once.

NoteCommon Error

counting a humanoid’s degrees of freedom by counting its limbs What goes wrong: Students say a humanoid with two arms and two legs has four degrees of freedom. The fix: Degrees of freedom count independent movements, not body parts. One arm alone can easily have seven. Count the movements each joint allows, then add them up.

6.3.2 Actuators, End Effectors and Hands

A humanoid’s joints need the same thing every joint needs — something to move them. But the choice of actuator is much narrower here than on a factory arm.

Why Humanoids Use Electric Motors

A pneumatic actuator needs a compressor to make compressed air. A hydraulic actuator needs a pump, a tank of oil and thick pipes. Neither of those is easy to carry around on a walking machine, and both would have to be dragged along behind it.

So humanoids are driven almost entirely by electric motors, which need only a battery and a pair of wires. The motors chosen have two qualities: high torque, so a knee can hold up the body’s weight without collapsing, and high precision, so a hand can be placed exactly where it is wanted. A gearbox is usually fitted between the motor and the joint, because a small motor spinning fast becomes a strong, slow, controllable movement once it is geared down.

NoteExam Corner

Q. Where on a humanoid would you expect the largest motors, and why?

A. In the hips, knees and ankles. Those joints carry the weight of the whole body above them, and they have to lift it during every step, so they need far more torque than a wrist or a neck.

Actuators are not used only for walking and reaching. In some humanoids, small motors sit behind the face and move the eyebrows, eyelids and mouth. The robot is not feeling anything; the control system is simply driving those motors into set positions so that a person looking at it reads the face as friendly, attentive or surprised.

Hands and Other End Effectors

At the end of each arm sits an end effector. On a humanoid this is usually a hand — a set of fingers meant to pick up objects that were made for human hands. Some humanoids carry a simple two-finger or three-finger gripper instead, and a humanoid built for one particular job may carry a tool there.

The difficulty is weight. Everything in the hand has to be lifted by the elbow and shoulder above it, so packing several motors into the fingers makes the whole arm heavier and slower. A common solution is to move the motors out of the hand altogether. See Fig. 6.3.2-1.

Diagram

Fig. 6.3.2-1 – sketch of a cable-driven robot hand: three fingers with thin cables running from the fingertips down through the palm and along the forearm to a set of small motors mounted near the elbow

NoteHow It Works

Cable-driven robot hand What it is: A robot hand whose fingers are closed by thin cables instead of by motors placed inside the fingers. How it works: Each cable runs from a finger, through the palm and down the forearm, to a small motor near the elbow. When the motor winds the cable in, the finger curls; when it lets the cable out, a spring straightens the finger again. The idea is copied from the tendons in your own hand, whose muscles are in the forearm, not the fingers. Used for: Giving a humanoid several movable fingers without making the hand heavy. Watch out: Cables stretch slightly and rub over time, so the grip loses accuracy and the cables need checking and replacing.

Even the best robot hands are far behind a human hand. Yours has enough separate movements and enough feeling in the skin to pick up a needle, hold an egg without cracking it and turn a key, all without looking. A robot hand with three fingers and no sense of touch can manage the box but not the egg.

NoteCommon Error

thinking a humanoid’s hands must have five fingers What goes wrong: Students assume that because the robot copies a human body, the hand must copy a human hand exactly. The fix: The number of fingers is decided by the job. Many working humanoids use two or three fingers because that is enough to hold what they need to hold, and it is lighter, cheaper and easier to control.

6.3.3 Sensors, Control System and Power Supply

Stand on one foot for a minute. You will feel small corrections happening all the time in your ankle, and you are not deciding on any of them. Balance is not a single act; it is thousands of tiny adjustments, made continuously, based on the fact that your body knows which way it is leaning.

A humanoid that stands on two feet has to do the same. This is the deepest difference between a humanoid and a bolted-down arm: an arm that stops receiving commands simply stands still, while a humanoid that stops correcting itself falls over.

The Sensors

A humanoid carries several kinds of sensor.

Cameras act as its eyes, letting the control system find objects, people and obstacles. Microphones pick up speech and other sounds. Touch sensors in the hands and body report contact, so the robot knows it has actually gripped something or bumped into something.

Two more sensors are there entirely for balance. An accelerometer senses which way gravity is pulling, so the robot can tell how far it is leaning. A gyroscope senses turning.

NoteHow It Works

Gyroscope What it is: A small sensor chip, usually mounted in the robot’s torso, that measures how fast the body is rotating. How it works: Inside the chip is a tiny piece of silicon kept vibrating. When the chip is turned, the vibration is pushed slightly sideways, and the electronics measure that sideways push. The size of the push tells the control system the rate of turning, and the direction of the push tells it which way. Used for: Detecting the moment a robot starts to tip, early enough to correct it. Watch out: A gyroscope reports how fast the body is turning, not which way is up. Its readings drift over time, so it is always used together with an accelerometer, which does know which way gravity points.

Diagram

Fig. 6.3.3-1 – a humanoid drawn twice. Left: standing upright with a small box in its torso labelled “accelerometer and gyroscope”. Right: the same robot leaning slightly forward, with a curved arrow at the ankle joint showing the correcting movement

NoteExam Corner

Q. A humanoid is standing still on a flat floor. Are its balance sensors doing anything?

A. Yes, constantly. Standing still is not stable for a two-legged machine — it drifts slowly one way or the other. The sensors keep reporting the lean and the ankle and hip motors keep making small corrections.

The Control System

Everything those sensors report has to be turned into movement, and quickly. The control system of a humanoid is built around a microcontroller or a processor, together with software that reads the sensors, works out what to do and drives the actuators.

The word that matters here is real-time. If the robot is tipping forward, a correction that arrives half a second late is worse than useless. So the control loop runs over and over, many times every second: read the sensors, work out the correction, move the joints, read the sensors again.

The control system also handles the slower decisions — where to walk, when to speak, which object to reach for — but the balance loop underneath never stops running.

The Power Supply

An industrial arm plugs into the factory’s electricity supply and never has to worry about it. A humanoid cannot drag a cable behind it, so it carries its own power.

The usual choice is a rechargeable lithium-ion battery, the same kind used in a mobile phone or a laptop, because it stores a good amount of energy for its weight. The battery feeds the motors, the sensors and the control system.

This is also the humanoid’s plainest limitation. Motors that lift a whole body use a great deal of current, so a humanoid usually works for a limited time and then has to be recharged. Adding a bigger battery adds weight, which the motors must then carry — and that uses more power again.

NoteSafety Note

Lithium-ion batteries store a lot of energy in a small space. Never short the two terminals together, never puncture or bend a cell, and always use the charger meant for that battery. A damaged or wrongly charged cell can overheat and catch fire.

NoteCuriosity Corner

Q5. What does a walking humanoid robot need that a robot arm bolted to the floor does not?

A. Two things above all. It needs its own power supply, usually a rechargeable lithium-ion battery, because it cannot stay tied to a wall socket. And it needs balance sensors — an accelerometer and a gyroscope — with a control loop that reads them many times a second and corrects the ankle and hip joints, because a two-legged robot that stops correcting itself falls over. A fixed arm needs neither: it is plugged in, and it cannot fall.

6.3.4 Communication, Safety and Customisation

A humanoid has no keyboard on its back. Yet somebody has to load its program, change its settings, check how much charge is left in its battery and see what its camera is looking at. All of that travels over a communication link.

Communication

Communication on a robot means the paths by which it receives commands and sends information out. A humanoid usually has more than one.

A wireless link — Wi-Fi or Bluetooth — is the common choice, because a walking robot cannot be tethered by a cable. Over it, an operator sends instructions from a laptop or a phone, and the robot sends back its status: which joints are where, how the battery is holding up, what the camera sees.

A wired connection is still used for the jobs where reliability matters more than freedom to move — loading new software, or running tests in the workshop with the robot standing still.

A network interface lets the robot talk to other machines rather than to a person. A humanoid working at a hospital reception, for example, may look up an appointment on the hospital’s own computer system before answering.

Communication also runs the other way, towards the people around it. A speaker lets the robot speak, and a screen on its chest or in its hand can display what it is saying or offer choices to touch. See Fig. 6.3.4-1.

Diagram

Fig. 6.3.4-1 – humanoid robot outline with four labels pointing to it: wireless antenna in the head, speaker and microphone at the head, touchscreen on the chest, red emergency stop button on the back

NoteExam Corner

Q. Why keep a wired connection at all if the robot has Wi-Fi?

A. Because a wireless link can drop out, and some jobs must not be interrupted halfway — loading new software is the clearest example. A cable is also faster and cannot be interfered with from outside the building.

NoteEthics Watch

A humanoid working among people carries a camera and a microphone, and its wireless link can send everything it sees and hears somewhere else. In a school corridor or a hospital waiting room, that means recording people who never agreed to it. The responsible practice is to make it obvious when the robot is recording, to keep only what the task actually needs, and to switch the recording off where people expect privacy.

Safety Features

An industrial arm is kept away from people behind a fence. A humanoid is built to be near people, so its safety has to work in the open.

An emergency stop button is fitted where a person can reach it — usually on the back or shoulder — and pressing it cuts power to the actuators at once.

Collision detection watches for the sudden extra effort that means the robot has hit something, and stops the movement.

Compliance in the joints is a gentler idea. A stiff joint that is pushed simply pushes back, and whatever is in the way gets crushed. A compliant joint gives a little when it meets resistance, in the way your elbow yields if someone leans on it. So if a robot’s arm meets a person, the arm moves aside instead of forcing through.

NoteExam Corner

Q. A humanoid is walking down a corridor and someone steps into its path. Which of its safety features helps first?

A. None of them, if things go well — the camera should see the person and the control system should stop or step aside before there is any contact. Collision detection and joint compliance are what protect the person when that first line fails.

Customisation for the Application

The ten building blocks describe a general humanoid. A real one is always built for a particular job, and the last step is adding whatever that job needs.

A humanoid meant for medical work may be fitted with specialised sensors and tools. One meant to greet visitors may be given a better microphone array so it can pick out a voice in a noisy hall, and a larger screen. One used for research on walking may carry extra sensors in its feet that would be pointless anywhere else.

This is worth remembering when you identify a robot from a photograph. The blocks in sections 6.3.1 to 6.3.3 will be there in some form on every humanoid. Anything beyond them is a clue about what this particular robot was built to do.

6.4 Identifying a Mobile Robot

The arm in section 6.2 waits for work to be brought to it. Sometimes that is impossible. A warehouse is too big to bring to a robot. A field of crops cannot be put on a conveyor. A hospital’s medicines have to reach wards spread over four floors.

For work like that, the robot has to go to the job. A robot built to move from place to place under its own power is called a mobile robot.

NoteDefinition

(A mobile robot is a robot that can move about its environment on its own, instead of being fixed in one place.)

Most mobile robots are low, wide and unglamorous — a flat platform on wheels or tracks with sensors around its edges. You will find them carrying shelves across warehouse floors, delivering linen and medicines along hospital corridors, cleaning house floors, driving between rows of crops on a farm, and inspecting pipelines and drains. Flying robots and underwater robots are mobile robots too; only the way they move is different. See Fig. 6.4-1.

Diagram

Fig. 6.4-1 – photo of a four-wheeled mobile robot on a floor, with sensors visible around the front edge of its platform

The Problem That Defines a Mobile Robot

A fixed arm always knows where it is, because it never goes anywhere. Its encoders tell it the angle of each joint, and from those angles the position of the gripper follows exactly.

A mobile robot has no such luck. It sets off across a floor, and after a minute of turning and rolling it must still be able to answer two questions: where am I now, and what is in front of me? Everything else about a mobile robot — the sensors it carries, the software it runs — follows from those two questions.

NoteExam Corner

Q. A mobile robot counts its wheel turns carefully. Is that enough to know where it is?

A. No. Wheels slip on smooth or wet floors, and each small slip adds a small error that never gets corrected. After a few minutes the robot’s own count can be well off, so it needs other sensors to check its position against the real world.

What to Look for on a Mobile Robot

Ten building blocks make up a mobile robot:

1. the chassis or frame
2. the wheels or tracks that move it
3. the actuators that drive them
4. the sensors that watch its surroundings
5. the control system
6. the communication link
7. localisation, which is how it works out where it is
8. the safety features
9. the software framework it is programmed on
10. the customisation added for its particular job

The next four sections work through these, and by the end of them you will have met every part used in all three robots in this chapter.

6.4.1 Chassis, Wheels and Tracks

A cycle’s frame is not the part that does anything. It carries no load by itself, it has no moving parts, and nobody buys a cycle for its frame. But bolt the wheels, pedals, chain and seat to nothing at all and you have no cycle.

A mobile robot’s frame is called the chassis, and it plays that same quiet part.

NoteDefinition

(The chassis is the base structure of a mobile robot. It carries the wheels or tracks and provides the mounting for the sensors, actuators, control system and battery.)

The chassis of a mobile robot is usually a flat plate or a shallow box, and it is deliberately kept low and wide. A tall, narrow robot tips over when it turns quickly or crosses a bump; a low, wide one stays put. The battery, which is the heaviest single item, is normally mounted near the bottom for the same reason.

Wheels, Tracks and Legs

The chassis has to be carried across the ground somehow, and there are three usual answers. See Fig. 6.4.1-1.

Diagram

Fig. 6.4.1-1 – three simple side views of the same chassis: (a) on four wheels, (b) on two continuous tracks, (c) on four legs

Wheels are the common choice. They are simple, quiet and waste little energy, and they suit any hard flat surface — a warehouse floor, a corridor, a paved path.

Tracks are the wide rubber or metal belts that run around a set of wheels, like those on an earth mover. They spread the robot’s weight over a much larger area of ground and grip far better on loose or broken surfaces.

Legs are used where the ground is too rough or too broken for either — a legged robot can step over an obstacle instead of having to climb over it. Legs are much harder to build and control, which is why most working mobile robots still roll.

NoteCompare Table

Table 6.3 – Wheels vs tracks | Feature | Wheels | Tracks | |—|—|—| | Best surface | Hard, flat, smooth | Loose, soft, uneven | | Grip on sand, mud or gravel | Poor — the wheel digs in and spins | Good — weight is spread over the whole track | | Energy used | Less | More | | Speed | Higher | Lower | | Turning | Easy | Harder, since the tracks must scrape round | | Wear and maintenance | Low | Higher, as tracks stretch and come off |

NoteCuriosity Corner

Q4. Why do some mobile robots run on tracks instead of wheels?

A. Because of the ground they have to cross. A wheel presses the robot’s whole weight onto a small patch of surface, so on sand, mud or gravel it sinks in and spins. A track spreads that weight along its whole length and grips much better. The robot pays for it with lower speed, more energy used and more wear, so tracks are chosen only when the surface demands them.

NoteExam Corner

Q. A robot has to carry parcels along a hospital corridor and also up a short ramp at the entrance. Wheels or tracks?

A. Wheels. The floor and the ramp are both hard and smooth, so there is nothing for tracks to gain, and tracks would be noisier, slower and would scrape the floor when turning in a narrow corridor.

NoteCommon Error

describing the chassis as “the robot’s body cover” What goes wrong: The chassis is confused with the outer plastic shell, which only keeps dust out and makes the robot look tidy. The fix: The chassis is the load-bearing structure underneath. Take off the outer cover and the robot still works; take away the chassis and there is nothing to hold the wheels, battery and electronics together.

6.4.2 Actuators and Sensors

A car turns because its front wheels are steered by a mechanism connected to the steering wheel. That is one way to change direction, but it needs an extra mechanism and an extra motor, and it takes up room at the front of the vehicle.

Small mobile robots usually avoid it altogether, using a neater trick instead.

Driving and Steering with Two Motors

Most mobile robots use differential drive. Two wheels, one on each side, are each driven by their own motor, and a small free-rolling castor wheel holds the robot level. There is no steering mechanism at all.

NoteHow It Works

Differential drive What it is: A way of driving and steering a robot using two independently driven wheels and no steering mechanism. How it works: If both motors turn at the same speed in the same direction, the robot rolls straight. If the left wheel turns faster than the right, the robot curves to the right. If one wheel turns forward while the other turns backward, the robot spins on the spot. Rowing a boat works on the same principle — pull harder on one oar and the boat swings the other way. Used for: Almost all small indoor robots, including floor cleaners and delivery robots, because turning on the spot lets them work in narrow spaces. Watch out: The robot goes straight only if the two motors really do turn at the same speed. Small differences between motors, or a slightly under-inflated wheel, make it drift to one side.

Diagram

Fig. 6.4.2-1 – top view of a differential drive robot: two driven wheels on the left and right of the chassis, a small castor at the back, with arrows showing the left wheel turning faster than the right and a curved dashed line showing the resulting path to the right

Some robots use special wheels with small rollers set around their rims, arranged so the robot can also slide sideways without turning. This is called omnidirectional movement, and it is useful where a robot has to slot into a tight gap without swinging its body round.

NoteExam Corner

Q. What happens to a differential drive robot if one wheel slips on a wet patch of floor?

A. That wheel turns without moving the robot forward, so the robot swings towards the slipping side, and its own count of wheel turns says it went straight. The robot ends up somewhere other than where it thinks it is.

The Sensors It Carries

Because a mobile robot moves through a world it does not control, it carries more sensors than an arm does. Table 6.4 lists the common ones.

NoteTable

Table 6.4 – sensors found on a mobile robot and what each one measures | Sensor | What it measures | What the robot uses it for | |—|—|—| | Camera | Light, as an image | Recognising signs, objects and people | | LIDAR | Distance to objects all around | Building a map and finding a clear path | | Ultrasonic sensor | Distance to whatever is directly ahead | Detecting obstacles close in front | | Infrared sensor | Reflected light from a nearby surface | Following a line, spotting the edge of a step | | Encoder | Turns of each wheel | Estimating how far the robot has travelled | | Gyroscope | Rate of turning | Keeping track of which way it is facing |

Two of these are worth a closer look.

A LIDAR sits on top of the robot and spins, firing a laser pulse and timing how long the reflection takes to come back. Thousands of such measurements, taken all around, give the robot an outline of the walls and objects near it.

An ultrasonic sensor does the same thing with sound instead of light. It sends out a short pulse of sound too high for you to hear, and measures the time until the echo returns.

Notice that the encoder here is the same device you met on the arm’s joints in section 6.2.4. On an arm it counts the turning of a joint; on a mobile robot it counts the turning of a wheel. The part is identical, only its job has changed.

NoteExam Corner

Q. A robot has a LIDAR that sees everything around it. Why fit ultrasonic sensors as well?

A. LIDAR misses some things — clear glass lets the laser through, and a very shiny surface bounces it away at an angle. A cheap ultrasonic sensor near the front catches obstacles the laser has missed, so the two together are safer than either alone.

That is the honest limitation of all these sensors: not one of them works everywhere. Cameras need light, ultrasonic sensors are confused by soft surfaces that absorb sound, infrared sensors are fooled by dark or shiny floors, and encoders cannot detect a wheel slipping. A working robot carries several kinds so that the failure of one does not blind it.

6.4.3 Control System, Communication and Localisation

Sensors do not decide anything. A LIDAR reports distances, an encoder reports wheel turns, and that is the end of their work. Something has to look at all those numbers and choose what the robot should do next. That is the control system’s job on a mobile robot, just as it was on the arm.

The difference is what it decides about. An arm’s control system works out joint angles. A mobile robot’s control system works out where to go, how fast, and when to stop.

Turning Readings into Decisions

Start with the simplest useful rule. If something is close ahead, stop. If it is a little further off, slow down. Otherwise carry on. Written as a method, it looks like this.

NoteAlgorithm

deciding what to do from the distance ahead Input: The distance to the nearest object directly in front of the robot, in centimetres. Output: One of three actions — stop, slow down, or go forward. Steps:

1. Read the distance ahead.
2. If the distance is less than 20 cm, the action is "stop".
3. Otherwise, if the distance is less than 50 cm, the action is "slow down".
4. Otherwise, the action is "go forward".
5. Send the action to the wheel motors and go back to step 1.

Why it works: The three cases cover every possible distance and never overlap, so exactly one action is chosen every time the rule runs.

The same method written as a Python program is short enough to read in one go. A function is used so that the rule is written once and can be applied to any distance.

NoteProgram

choosing an action from the distance ahead Problem: Given a distance in centimetres, decide whether the robot should stop, slow down or go forward, and check the rule on a set of test distances.

def decide(distance_cm):
    if distance_cm < 20:
        return "stop"
    elif distance_cm < 50:
        return "slow down"
    else:
        return "go forward"

# test distances, in centimetres, used to check the rule
test_distances = [120, 64, 45, 28, 18]

for distance in test_distances:
    print(distance, "cm ->", decide(distance))

Output:

120 cm -> go forward
64 cm -> go forward
45 cm -> slow down
28 cm -> slow down
18 cm -> stop

On a real robot the list of test distances would be replaced by a fresh reading from the sensor, and the loop would never end — it would run over and over for as long as the robot was switched on.

NoteExam Corner

Q. Why does the control system need to run this decision again and again, instead of once at the start?

A. Because the world keeps changing. A corridor that was clear a second ago may have someone walking into it now. A decision made once would be out of date almost immediately.

Communication

A mobile robot is usually not working alone. In a warehouse, dozens of them share the same floor, and a central computer sends each one its next job and keeps them out of each other’s way.

That means the robot has to receive commands and transmit information — its position, its battery level, whether it has finished its task. The link is normally wireless, because the robot moves; a wired connection is used mainly when the robot is docked for charging or servicing.

NoteExam Corner

Q. The Wi-Fi in a warehouse drops out for ten seconds. Should the robots stop?

A. They should be built so that they can. Obstacle detection and stopping must be decided on the robot itself, not by the central computer, so that a lost connection never leaves a moving robot blind.

Localisation

The hardest question a mobile robot faces is the one from section 6.4: where am I? Working that out is called localisation.

Counting wheel turns gives a first estimate, but as you saw, every slip adds an error that is never corrected. So the robot compares what its sensors see with what it knows about the building.

The best-known method is SLAM (Simultaneous Localisation and Mapping). The name says what it does: the robot builds a map of its surroundings and works out its own position within that map, both at the same time. See Fig. 6.4.3-1.

Diagram

Fig. 6.4.3-1 – a robot in a room with LIDAR beams fanning out to the walls and a doorway, and beside it the outline map the robot has built so far, with a small marker showing where the robot thinks it is

The idea is one you use yourself. Walk into an unfamiliar house and you build a rough picture of it as you go — a corridor here, a turn there, a door on the left. A few minutes later you can find your way back to the entrance, because you know both what the place looks like and where in it you are standing.

The robot does it with numbers. Its LIDAR gives it the shape of the walls around it. As it rolls forward, it takes another set of readings and looks for the same corners and walls in the new set. From how much those features have shifted, it works out how far it really moved — which is far more reliable than counting wheel turns.

NoteExam Corner

Q. Where would SLAM struggle?

A. In a long, plain corridor with blank walls, every position looks the same to the sensors, so the robot cannot tell how far along it has come. Corners, doorways and furniture are what make a place recognisable.

6.4.4 Software Framework, Safety and Customisation

The program in section 6.4.3 was three lines of decision. A real mobile robot needs far more than that: something to read the LIDAR, something to run the motors, something to build the map, something to plan a path around an obstacle, and something to keep all of those running at once without getting in each other’s way.

Writing all of that from nothing, for every new robot, would take years. So teams build on a software framework.

NoteDefinition

(A software framework is a ready-made platform of tools and libraries that a robot’s own program is built on, so that common jobs do not have to be written again from the beginning.)

A framework gives three things. It gives libraries — pieces of working code for tasks such as reading a sensor or planning a path. It gives development tools for building, running and checking programs. And it gives a simulation environment, in which the robot and its surroundings appear on screen, so a new program can be tested before it is ever loaded into a real machine.

The best known framework in robotics is ROS, the Robot Operating System, which despite its name is not an operating system like Windows or Linux but a collection of tools and libraries that runs on top of one.

NoteExam Corner

Q. Why test a mobile robot’s program in simulation first?

A. Because a mistake in a path-planning program sends a real robot into a wall, or down a flight of stairs, at full speed. On screen, the same mistake costs nothing and can be watched closely and fixed.

NoteTry Yourself

Take the program from section 6.4.3 and change the two distances from 20 and 50 to 30 and 80. Write down what you think the output will be for the same five test distances, then run it and check.

Safety Features

A mobile robot’s danger is different from an arm’s. It is not usually strong enough to crush anything, but it moves about among people who are not expecting it, and some of them are carrying things or not looking down.

An emergency stop button is fitted on the top of the robot, where anyone nearby can reach it.

Collision detection is used here too. Many mobile robots also carry a bumper — a springy strip around the front edge with a switch behind it, so that any contact at all cuts the drive at once.

Speed limits matter as much as stopping. A robot moving through an area where people walk is set to a slow speed, so that it has time to detect someone and stop. The same robot may be allowed to move faster in an aisle closed to people.

NoteExam Corner

Q. A delivery robot’s LIDAR is mounted on top, well above floor level. What might it fail to notice?

A. Anything low down — a bag left on the floor, a step, an open drawer, a small child’s foot. This is exactly why low bumpers and downward-facing sensors are fitted as well.

Customisation for the Application

The ten building blocks give you a mobile robot that can move about safely and knows where it is. What it is actually for is decided by what is added on top. See Fig. 6.4.4-1.

Diagram

Fig. 6.4.4-1 – the same wheeled chassis drawn three times with different payloads: (a) a lifting platform raising a shelf, (b) a tank and spray nozzles for a farm, (c) a mast carrying a camera for inspection

A warehouse robot is given a lifting platform so it can carry a whole shelf. A farm robot is given a tank and nozzles, and sensors that can tell a crop from a weed. A search and rescue robot is given a thermal camera, which sees warm bodies in smoke and darkness where an ordinary camera sees nothing. An inspection robot is given a camera on a mast and lights of its own.

The chassis, wheels, motors, sensors and control system underneath may be nearly identical in all four. Only the payload changes.

NoteEthics Watch

A patrolling robot does not only record what its camera sees; it also records where it went and when, which over weeks becomes a detailed record of people’s movements through a building. Used carelessly, that turns a safety machine into a way of tracking staff or students. The responsible practice is to agree in advance which areas the robot patrols, to keep its logs only as long as they are needed, and to say plainly who may look at them.

6.5 Telling the Three Robots Apart

Three machines, all called robots, and no two of them alike to look at. An arm that never leaves its table, a two-legged machine that walks among people, and a flat platform that rolls down a corridor. Table 6.5 sets them side by side.

NoteCompare Table

Table 6.5 – Industrial arm vs humanoid vs mobile robot | Feature | Industrial robotic arm | Humanoid robot | Mobile robot | |—|—|—|—| | Body plan | A chain of links and joints on a fixed base | Head, torso, arms and usually legs | A low chassis on wheels or tracks | | Does it move about | No, it is bolted in place | Yes, usually by walking | Yes, that is its main purpose | | Workspace | Only what the arm can reach | Wherever it can walk to | The whole building, floor or field | | Power | Plugged into the mains supply | Its own rechargeable battery | Its own rechargeable battery | | Usual end effector | Gripper, welding torch, cutter, vacuum cups | Hands, or a simple gripper | Often none; a payload instead | | Sensors it cannot do without | Encoders at every joint | Accelerometer and gyroscope, for balance | LIDAR, ultrasonic and encoders, for navigation | | Its hardest problem | Repeating a movement exactly, thousands of times | Staying upright | Knowing where it is | | Typical use | Welding, assembly, packing | Meeting people, research on walking | Delivery, warehouse work, cleaning, farming |

What They Share

Read down that table and the differences are obvious. Now read across it for what does not change.

Every one of the three has a rigid structure that holds it together, whether it is called an arm, a frame or a chassis. Every one has actuators, and in all three cases they are mostly electric motors. Every one has sensors reporting back to a control system, and a control system deciding what the actuators should do next. Every one needs power from somewhere.

The parts are the same. What changes is how many of each there are, where they are placed, and which problem the design is fighting. An arm’s design fights the problem of precision. A humanoid’s design fights gravity. A mobile robot’s design fights not knowing where it is.

This is the idea worth carrying out of the chapter: a robot’s mechanical body can take any form at all. There is no shape that makes something a robot and no shape that rules it out. What makes a machine a robot is that it senses, decides and acts — and it may do so from a table, on two legs, or on four wheels.

Identifying a Robot You Have Not Seen Before

Because the building blocks are always the same, you can work out a great deal about an unfamiliar robot from a photograph or a short video.

NoteAlgorithm

identifying an unfamiliar robot from a picture or video Input: A photograph or video clip of a robot at work. Output: A description of what kind of robot it is and what it is built to do. Steps:

1. Look at the base. Is it fixed to something, or does it stand on
   wheels, tracks or legs? This decides whether the robot moves about.
2. Trace the structure outwards from the base and count the joints.
   The more joints, the more freedom of movement.
3. Find the far end of each moving part and identify the end effector,
   if there is one, and the payload, if there is one.
4. Look for the sensors: lenses, small round transmitters, a spinning
   unit on top.
5. Look for the power source: a trailing cable means mains supply,
   no cable means an onboard battery.
6. Put it together. The end effector says what work is done, the
   sensors say what the robot must watch out for, and the base says
   whether the work comes to the robot or the robot goes to the work.

Why it works: Every robot is built from the same short list of blocks, so naming the blocks you can see is enough to describe an unfamiliar machine.

NoteExam Corner

Q. You see a machine with a camera, a control box and a gripper — but it is bolted to a table and does exactly the same movement every fifteen seconds. What kind of robot is it, and what would you want to check?

A. It looks like an industrial robotic arm doing a pick-and-place job. Worth checking: how many joints it has, what the gripper is holding, and whether there is a fence or a light curtain around it — that last one tells you people work nearby.

6.6 Summary

1. A robot’s size and shape are decided by three things: the task it must do, the workspace it must work in, and the environment and its difficulties.
2. However different two robots look, they are built from the same kinds of parts: a rigid structure, actuators, sensors, a control system and a power supply.
3. An industrial robotic arm is fixed in one place and repeats a task. Its body is a chain of rigid links joined by movable joints.
4. Robot joints are either rotary, where one link turns, or linear, where one link slides. Each joint gives one independent movement, called an axis, and the total number of independent movements is the robot’s degrees of freedom.
5. The end effector is the tool at the far end of an arm — a gripper, welding torch, vacuum cups or cutter. It is the only part that touches the workpiece.
6. Actuators supply the force that moves a joint. They may be electric, pneumatic or hydraulic; electric motors are the most common because they can be positioned precisely.
7. The control system is the hardware controller plus the software that works out how each joint should move. It sends commands to the actuators and reads feedback from the sensors.
8. An encoder inside each joint counts pulses as the joint turns, so the control system always knows the angle of every joint. Force sensors, proximity sensors and vision systems are also used.
9. Industrial arms are guarded by emergency stop buttons, fences, light curtains and collision detection. They are programmed either by teaching positions with a hand-held control box or by writing and simulating a program offline.
10. Integration is the work of joining an arm to the conveyors, fixtures and software of the factory around it.
11. A humanoid robot is built to the plan of a human body, so that it can move through places and use tools made for people.
12. A humanoid’s frame is made of light materials such as aluminium, carbon fibre and plastic, because the robot must carry its own weight.
13. A humanoid has many joints, and some joint assemblies, such as the shoulder, give several degrees of freedom on their own.
14. Humanoids use electric motors, since pneumatic and hydraulic systems would need a compressor or pump to be carried along. Hands are often driven by cables from motors in the forearm to keep the hand light.
15. Cameras, microphones and touch sensors let a humanoid sense its surroundings, while an accelerometer and a gyroscope let it stay balanced. Its control loop must run in real time or the robot falls.
16. A humanoid carries its own rechargeable lithium-ion battery, which limits how long it can work.
17. Humanoids communicate wirelessly, by cable and over a network, and are made safe around people by emergency stop buttons, collision detection and compliant joints that give way when pushed.
18. A mobile robot moves about on its own, on wheels, tracks or legs. Its chassis is kept low and wide for stability.
19. Wheels suit hard flat surfaces; tracks spread the robot’s weight and grip better on loose or uneven ground, at the cost of speed, energy and wear.
20. Differential drive steers a robot using two independently driven wheels, with no steering mechanism at all.
21. Mobile robots carry cameras, LIDAR, ultrasonic sensors, infrared sensors, encoders and gyroscopes, because no single sensor works in every situation.
22. Localisation is the problem of working out where the robot is. SLAM builds a map of the surroundings and finds the robot’s position within that map at the same time.
23. A software framework supplies ready-made libraries, development tools and a simulation environment, so a new program can be tested before it runs on a real machine.
24. Customisation is the payload or tool added on top of a standard mobile robot, and it is what decides whether the robot lifts shelves, sprays a field or inspects a pipeline.
25. A mechanical body can take any form. What makes a machine a robot is that it senses, decides and acts.

6.7 Key Terms

1. Actuator – the part of a robot that converts energy into movement, so that a joint or a tool actually moves.
2. Axis – one independent movement of a robot, provided by one joint; an arm with six such movements is called a six-axis robot.
3. Chassis – the base structure of a mobile robot, carrying the wheels or tracks and holding the sensors, actuators, control system and battery.
4. Communication – the paths by which a robot receives commands and sends out information, by wireless link, cable or network.
5. Control system – the hardware and software of a robot that read the sensors, work out what should happen next, and send commands to the actuators.
6. Controller – the hardware part of the control system: a dedicated computer or embedded circuit board, with its input and output connections.
7. Degree of freedom (DOF) – the number of independent movements a robot can make.
8. Differential drive – a way of driving and steering a robot using two independently driven wheels and no steering mechanism.
9. Encoder – a sensor on a turning shaft that produces electrical pulses as it rotates, letting the control system work out the angle turned.
10. End effector – the tool fitted at the far end of a robot arm, which interacts directly with the workpiece.
11. Environment – the surroundings a robot must work in, and the difficulties they bring, such as water, dust or uneven ground.
12. Feedback – information that travels back from a sensor to the control system.
13. Frame – the rigid structure of a robot that gives it its shape and holds all its other parts in place.
14. Gripper – an end effector with two or more fingers that close on an object to hold it.
15. Humanoid robot – a robot whose body is built to resemble the human body, with a head, torso, arms and usually legs.
16. Industrial robotic arm – a robot built as a chain of moving segments, fixed in one place, used for repeated tasks in a factory or workshop.
17. Joint – the movable connection between two links, and the place where a robot’s movement happens.
18. Link – a rigid segment of a robot’s body, which does not bend or stretch.
19. Localisation – working out where a robot is within its surroundings.
20. Mobile robot – a robot that can move about its environment on its own, instead of being fixed in one place.
21. Sensor – a device that measures a physical quantity, such as position, force or distance, and turns it into a signal the control system can read.
22. SLAM (Simultaneous Localisation and Mapping) – a method by which a robot builds a map of its surroundings and works out its own position within that map at the same time.
23. Software framework – a ready-made platform of libraries, development tools and simulation software that a robot’s own program is built on.
24. Task – the particular work a robot is built to do, which is the first thing that decides its size and shape.
25. Uncanny valley – the unease people feel towards a robot or animated face that looks almost, but not quite, human.
26. Workspace – the region of space that a robot can reach and work in.

6.8 Objective Questions

1. (The part of a robot arm that touches the workpiece and does the actual job is called the:)

   1. link
   2. joint
   3. end effector
   4. controller

2. (Fill in the blanks.)

   1. A ________ joint allows one link to slide straight in or out, while a rotary joint allows one link to turn.
   2. ________ spread a mobile robot’s weight over a larger area of ground and grip better on mud and gravel than wheels do.
   3. The region of space a robot can reach and work in is called its ________.
   4. A humanoid robot carries its own rechargeable ________ battery, because it cannot be tied to a wall socket.

3. (State whether each statement is true or false.)

   1. An industrial robotic arm must be moved to a new place for each new job.
   2. Every degree of freedom of a robot needs its own actuator.
   3. A gyroscope tells a robot which way is up.
   4. The same encoder can be used to count the turning of a robot’s joint or of its wheel.

4. (Match each sensor in Column A with what it measures in Column B.)

   Column A	Column B
   a. Encoder	i. Distance to objects all around the robot
   b. LIDAR	ii. How hard the robot is pressing or pulling
   c. Accelerometer	iii. Turns of a shaft
   d. Force sensor	iv. The direction in which gravity pulls

5. (A differential drive robot has its left wheel turning forward and its right wheel turning backward, both at the same speed. The robot will:)

   1. roll straight forward
   2. roll straight backward
   3. spin on the spot
   4. stay still

6. (A mobile robot uses the rule from this chapter: stop below 20 cm, slow down below 50 cm, otherwise go forward. Write the action chosen for a reading of 35 cm.)

7. (Name the type of end effector that would be used to lift a large flat sheet of glass.)

6.9 Concept Questions

1. (Define a link and a joint, and state clearly which of the two is responsible for a robot’s movement.)

2. (Give a reason why humanoid robots are built with aluminium, carbon fibre and plastic instead of steel, even though steel is stronger and cheaper.)

3. (Explain what an encoder does on an industrial robotic arm, and describe what problem the arm would face if its encoders were removed.)

4. (Differentiate between wheels and tracks on a mobile robot, giving one situation in which each is the better choice.)

5. (An engineer says that counting wheel turns is not enough for a mobile robot to know where it is. Explain why this is so, and state what SLAM does about it.)

6. (An encoder disc produces 600 pulses in one full turn. The controller counts 100 pulses while a joint moves. Work out the angle the joint has turned through, showing your steps.)

7. (Write a Python function named decide that takes a distance in centimetres and returns “reverse” if the distance is less than 10, “stop” if it is less than 25, and “go forward” otherwise. Then write the lines needed to print the action chosen for a distance of 8 cm.)

6.10 Case-Based Questions

1. (A school sets up a small robotic arm on a bench in its workshop. The arm has a gripper and is programmed to pick small components from a tray and place them into a jig. Students stand around the bench to watch it work.

   a. Name two safety features that should be fitted before the arm is switched on, and say what each one protects against. b. The arm has to be stopped in exactly the same position every time it reaches for the tray. Name the sensor that makes this possible and explain how it reports the position. c. The teacher wants to replace the gripper with a small screwdriver tool. State the precaution that must be taken before the tool is changed, and give the reason for it.)

2. (A farm outside Jaipur uses a mobile robot to spray a liquid between rows of crops. The soil is soft and often muddy after watering. The robot carries a tank and nozzles, and a sensor at the front reports the distance to whatever is directly ahead. On one run, the sensor reports these readings, one after another:

   Reading	Distance ahead
   1	120 cm
   2	70 cm
   3	40 cm
   4	15 cm

   a. Should this robot run on wheels or on tracks? Give two reasons for your choice. b. Using the rule from this chapter — stop below 20 cm, slow down below 50 cm, otherwise go forward — state the action the control system chooses for each of the four readings. c. The tank and nozzles are not part of the ten building blocks of a mobile robot. What is this addition called, and name one more sensor the farm robot would need that a warehouse robot would not.)