Chapter 2: From Robots to Cobots
On a school trip to a small packaging unit near Jaipur, a class of students stops at two workstations. At the first one, a sealing machine presses down every few seconds. It does the same press, the same way, all day. If a carton is placed crookedly, the machine still presses. It has no way of noticing.
At the second station, a slim orange arm lifts cartons from a belt and stacks them on a pallet. A worker stands right beside it, turning each carton so the label faces out. When his hand comes too close, the arm slows down. When he steps back, it speeds up again. Nobody has fenced the arm off. Nobody presses a button to tell it what to do next.
Both are made of metal, motors and wires. Both run on electricity. Yet one of them keeps checking its surroundings and changing what it does, and the other one never does.
See Fig. 2.0-1 for the two workstations side by side.
Diagram
Fig. 2.0-1 – two workstations drawn side by side. Left: a fixed sealing machine pressing down on a carton on a belt, with a warning barrier around it and no person nearby; label it “presses every few seconds, always the same way”. Right: a jointed robotic arm lifting a carton onto a pallet with a worker standing beside it and no barrier; label it “senses the worker, slows down when he is close”
This chapter will answer each of these questions as you go through it.
2.1 What Is a Machine?
Try opening a tin of paint with your fingers. You will not manage it. Slip a screwdriver under the lid, press down on the handle, and the lid lifts easily. Nothing about your hand changed. What changed is that you put your effort into a tool that turned a small push at one end into a big lift at the other.
That is the whole reason machines exist. Some jobs need more force than a person can produce, some need more speed, some need the same action repeated a thousand times without getting tired. A machine is what we build when human effort alone is not enough.
Two words there are worth pinning down. The input of a machine is whatever goes into it — the energy you supply, the material you feed in, or the push you give. The output is what comes out — the movement, the force, the finished item. A hand pump takes your push as input and gives out water.
What a machine is made of
A machine is not one solid lump. It is a set of parts joined together so that each one passes on movement or force to the next. These parts include gears, levers, pulleys, belts, shafts, springs and motors.
Think of a bicycle, shown in Fig. 2.1-1. You press the pedal, the crank turns, the chain carries that turning to the rear sprocket, and the sprocket turns the wheel. Every part has one job, and the chain of parts together does something no single part could do.
Diagram
Fig. 2.1-1 – a side view of a bicycle drive with the pedal, crank, chain, front and rear sprockets and rear wheel labelled; an arrow at the pedal marked “input: your push” and an arrow at the rear wheel marked “output: the cycle moves”
Machines also need an energy source. A bicycle runs on the force of your legs. A ceiling fan and a mixer grinder run on electricity. A tractor and a bus run on fuel. Whatever the source, the machine’s job is to convert that energy into useful mechanical work.
From simple to complex
Machines cover an enormous range. At one end sit simple tools with very few parts — scissors, a bottle opener, a pulley on a well, a wheelbarrow. At the other end sit systems with thousands of parts working together, such as a car engine, a printing press or a lift in a building.
You will find machines wherever work has to be done at scale: in factories, on farms, in hospitals, on construction sites and in transport. A power loom weaving cloth, a threshing machine separating grain, an X-ray machine taking an image — all of these are machines.
But notice one thing about every machine described so far. Each one does exactly what its design allows, in exactly the same way, every single time. Press the scissors and they cut, whether you are cutting paper or your own sleeve. The sealing machine from the chapter opener presses down whether the carton beneath it is straight or crooked. A machine has no way of checking what is in front of it and no way of changing its mind. That limitation is exactly where the next idea begins.
2.2 What Is a Robot?
A sealing machine cannot tell a straight carton from a crooked one. To fix that, we would have to give the machine three new abilities: a way of finding out what is in front of it, a way of working out what should be done about it, and a way of acting on that answer. Add those three abilities to a machine and you have a robot.
So a robot is still a machine. It has a body, motors, joints and an energy source, exactly like the machines in the previous section. What sets it apart is the loop it runs: sense, work out, act, and then sense again.
Sensing has to be continuous
A robot does not check its surroundings once and then stop looking. It keeps checking, many times a second, all the while it is working. This matters more than it first appears.
Picture a robot that carries trays along a corridor in a hospital. It has to keep watching for people stepping into its path. It has to keep its speed steady so the tray does not tip. And it has to keep doing the one task it was set — deliver the tray to Ward 4 — and nothing else. Take away the continuous checking and it becomes an ordinary trolley on wheels that will happily drive into the first person who crosses it.
Programmable means it can be retaught
The second key word in the definition is programmable. A program is a set of instructions written by a person and stored in the robot, telling it what to do in each situation it may meet.
This is a genuine difference from a machine. A power loom is built to weave; you cannot make it stitch buttons instead. But the same robotic arm that tightens screws on a car door today can be given a new program tomorrow and spend the week applying glue. The metal does not change. Only the instructions do.
How much is left to the robot
Robots differ in how much of the task they handle by themselves.
An autonomous robot carries out its whole task without a person guiding each step. A robotic vacuum cleaner is switched on once and then covers the floor on its own, turning when it meets a wall.
A semi-autonomous robot handles part of the task itself and takes the rest from a human operator. A surgical robot is a good example: the surgeon guides the instrument, while the robot steadies the movement and holds the position exactly.
Notice what a robot is not doing in either case. It is not thinking about the job. It is comparing the readings it has collected against the rules stored in its program, and picking the action those rules point to.
Robots come in many shapes
There is no single robot shape, as Fig. 2.2-1 shows. A robot may be a jointed arm bolted to a factory floor, a boxy machine on wheels, a drone with four propellers, or a two-legged figure built to walk. The shape follows the task, not the other way round.
Diagram
Fig. 2.2-1 – three robots drawn side by side and labelled: a jointed industrial arm with a gripper fixed to a base, a small wheeled robot with a flat top carrying a box, and a four-propeller drone in flight; caption line under each naming its usual task — welding, carrying goods, surveying land
2.3 How a Robot Works
Knowing that a robot senses, works out and acts is a start. But if you were asked to build one, that sentence would not take you very far. You would want to know what happens first, what happens next, and where each part of the robot fits into that order.
Every robot, whatever its shape, runs through the same sequence of stages. The stages are always in the same order, and the last one leads back to the first, so the sequence runs over and over for as long as the robot is working.
It begins with the task — the one job the robot has been built and programmed to do. Stacking cartons is a task. Sweeping a floor is a task. Welding a car door at the same three points every time is a task. Everything else the robot does exists only to get that task done.
Fig. 2.3-1 shows the full sequence as a block diagram.
Diagram
Fig. 2.3-1 – a block diagram drawn as a closed loop. Blocks in order, joined by arrows: “Task” → “Sensing” → “Processing” → “Deciding” → “Actuator movement” → “Interaction with the object” → “Feedback”. A return arrow runs from “Feedback” back to “Processing”. Label the return arrow “what actually happened”
The four sections that follow take these stages in turn: how a robot senses, how it processes what it has sensed and picks an action, how it moves and handles objects, and how it checks its own work afterwards.
2.3.1 Sensing the Surroundings
Shut your eyes and try to walk across your classroom. You will get a few steps before you slow right down, because you no longer know where the desks are. A robot with no way of reading its surroundings is in exactly that position, except that it does not slow down.
So the first stage of the loop is to collect facts about the world. The part that does this is a sensor.
The word “converts” is doing the real work in that sentence. A robot’s electronics can only handle electrical signals. Heat, distance and sound are not electrical signals, so something has to stand in between and translate. That is the sensor’s whole job.
The main kinds a robot carries
Different tasks need different facts, so robots carry different sensors. Some common ones are listed below, each with the quantity it detects.
- Camera — light, giving a picture of what is ahead.
- Microphone — sound, including spoken commands.
- Ultrasonic sensor — the distance to an object in front.
- LIDAR (Light Detection and Ranging) — distances all around, built up into a map of the space.
- Touch or tactile sensor — contact, and how hard the contact is.
- Temperature sensor — how hot or cold a surface or the surrounding air is.
- Angle sensor — how far a joint has turned.
- Inertial sensor — how fast the robot is speeding up, slowing down or tilting.
A delivery robot in a hospital corridor might use several of these at once, as Fig. 2.3-2 shows.
Diagram
Fig. 2.3-2 – a wheeled delivery robot with a flat carrying tray, drawn from the side with labelled call-outs to its sensors: camera at the top front, ultrasonic sensor on the front panel, LIDAR unit on the upper body, touch sensor strip along the front bumper, angle sensor at a wheel axle, inertial sensor inside the body
2.3.2 Processing and Deciding
A sensor hands over a signal, and a signal on its own means nothing. An ultrasonic sensor does not report “a wall”. It reports a length of time. Something has to turn that time into a distance, compare the distance against what the task requires, and pick what the robot should do next. That something is the controller.
In a small robot the controller is usually a microcontroller — a single chip with a processor, memory and connection pins, built to be wired straight to sensors and motors. Larger robots use a full onboard computer. The size differs; the job does not.
First, turn the reading into data
Data here simply means the sensor’s information written as numbers the program can work with. Turning a raw signal into such a number is the first half of processing, and it often needs a small calculation.
Take the ultrasonic sensor from the previous section. It gives the time for the sound burst to travel out and come back. Sound travels through air at a steady speed, so the controller works out
\[\text{distance} = \frac{v \times t}{2}\]
where \(v\) is the speed of sound in air, \(t\) is the time for the full round trip, and the division by 2 is there because the sound covered the gap twice — once going, once returning.
Then, apply the rules
Now the controller has a number it can use. The second half of processing is comparing that number against the rules the programmer wrote, and picking the action those rules point to. The set of rules, written as an exact list of steps, is called an algorithm.
Suppose our delivery robot has been told to stop if anything comes within 30 cm of it. Fig. 2.3-3 shows how that rule is applied.
Diagram
Fig. 2.3-3 – a flowchart. Start → box “read echo time from ultrasonic sensor” → box “calculate distance” → diamond “distance less than 30 cm?” → “Yes” branch to box “stop the wheels” → “No” branch to box “keep moving at set speed”; both branches join and loop back with an arrow to the first box
Not all robots follow fixed rules
A robot like the one above follows instructions written out in advance. Every situation it can handle was thought of by a programmer beforehand.
More advanced robots work differently. Instead of being given every rule, they are given large amounts of past data and a method for finding patterns in it. That is what machine learning means, and it is why such robots can cope with situations no programmer listed. You will study how this is done later in the book.
Either way, the robot is not thinking about the problem. It is comparing numbers against stored rules, or against patterns worked out from stored data, and picking whichever action that comparison points to.
2.3.3 Acting and Interacting with Objects
The controller has picked an action. Nothing has happened yet. An instruction inside a chip is only an electrical signal, and a carton on a belt is not going to move because a chip would like it to. Something has to turn that signal into real movement. That part is the actuator.
You can think of a sensor and an actuator as opposites. A sensor takes something physical and turns it into a signal. An actuator takes a signal and turns it back into something physical.
The actuators on a robot include motors that spin its wheels, servo motors that turn its joints to a set angle, and grippers that open and close to hold an object. A robot’s limbs, wheels and gripper are all driven by actuators.
The actuators decide what the robot can do to an object
A robot rarely just moves. It usually does something to an object — picks it up, sets it down, turns it over, joins it to another part, or paints it. This is the interaction stage, and what the robot can manage here depends entirely on the actuators it was built with.
An arm with a simple two-finger gripper can pick up a carton, as in Fig. 2.3-4. It cannot pick up a loose handful of grain. A robot with a suction cup can lift a flat glass sheet, but not a rough brick. Change the actuator and you change what the robot can handle.
Diagram
Fig. 2.3-4 – a jointed robotic arm lifting a carton off a belt. Label the actuators: a motor at the base for turning, servo motors at the shoulder, elbow and wrist joints, and a two-finger gripper closing on the carton. Show a small arrow at each joint indicating its direction of movement
Precision matters here too. Placing a carton on a pallet allows a centimetre or two of error. Fitting a component onto a circuit board does not. The finer the placement needed, the more accurate the actuators and the joint sensors have to be.
2.3.4 Feedback, Learning and Adaptation
When you fill a glass from a tap, you do not set the tap once and look away. You watch the water level, and as it nears the rim you close the tap. You are checking the result of your own action while it is still happening, and correcting as you go.
A robot has to do the same thing, for the same reason. Sending an instruction is no guarantee that the instruction was carried out. A wheel may slip on a wet floor. A carton may be heavier than expected and the arm may fall short. So after acting, the robot senses again and reports the result back to the controller. That report is called feedback.
The difference between the two is the error, and it is the error that the controller works on:
\[\text{error} = \text{wanted value} - \text{actual value}\]
As long as the error is not zero, the controller keeps sending correcting instructions. When it reaches zero, the robot has done what it was told. Fig. 2.3-5 shows this arrangement.
Diagram
Fig. 2.3-5 – a closed-loop block diagram. “Wanted angle” arrow into a small circle marked with a minus sign; from the circle an arrow labelled “error” into a “Controller” block; then into a “Motor” block; then into a “Joint” block; an arrow from the joint labelled “actual angle” goes to an “Angle sensor” block and from there back into the circle as the feedback path
Learning from many attempts
Feedback fixes one action while it is being carried out. Some robots go a step further and use what they have collected over many attempts to work out a better way of doing the whole task.
A robot moving goods across a warehouse can record how long each route took on every trip it made. Given enough of those records, the program can identify which route has been fastest and use that one in future. Nobody rewrote its instructions. The change came out of its own stored data, using the machine learning methods mentioned earlier.
That completes the loop. Sensing gathers the facts, the controller turns them into a decision, the actuators carry it out, and feedback reports what actually happened so the next pass can be better. This whole loop is what an ordinary machine does not have, and it is the real answer to the question the chapter opened with.
2.4 Telling a Machine and a Robot Apart
Every robot is a machine. Not every machine is a robot. So the question is never “is it a machine or a robot?” as though the two were separate families. The real question is whether this particular machine has the extra abilities that make it a robot.
There are three checks, and a robot passes all three.
The three checks
First, does it keep sensing? A robot senses its surroundings continuously while it works, not once at the start. A washing machine’s timer counts down whatever is inside the drum. A robot keeps looking.
Second, does it hold the conditions its task needs — a steady speed, a set position, a fixed pressure — and correct itself when they drift? A robot uses feedback to hold them. A machine simply runs and whatever happens, happens.
Third, is it doing a task it was programmed to do, and can that program be changed? A robot’s task lives in its instructions. A machine’s task lives in its metal.
A ceiling fan fails all three. It cannot sense the room, it does not correct its own speed if the voltage drops, and it will never do anything except spin. An automatic door passes the first check — it senses a person approaching — but it does not correct itself against a target, and its behaviour cannot be reprogrammed into anything else. It sits closer to a machine.
Table 2.1 sets the two side by side.
2.4.1 Machine or Robot? Reading the Picture
The three checks work well when you can watch something running. But often all you have is a picture — in a book, on a screen, or in a factory brochure. A photograph cannot show you sensing happening or feedback being applied. So you have to read the clues that a picture does show.
What to look for
- Sensors on the body. A lens, a pair of round ultrasonic openings, a spinning drum on top for LIDAR, a thin strip along a bumper. If a device is carrying sensors, it is collecting facts about its surroundings for some reason.
- Separate joints. Count the places where the structure bends. Several joints, each with its own motor housing, means each one can be moved independently, which a fixed machine has no need of.
- A controller and its wiring. A small board inside the body, or a thick bundle of cables running from a cabinet to the moving parts, points to instructions being sent rather than a mechanism simply turning.
- A changeable tool at the end. A gripper, a suction cup or a bolt-on tool head suggests the device is meant for more than one job. A permanently fitted blade or press suggests one job only.
- What is missing. A device that is only a handle, a pivot and a working edge has nowhere to put a sensor or a controller. That settles it.
Fig. 2.4-1 shows six devices to read this way.
Diagram
Fig. 2.4-1 – six line drawings in a grid, each labelled: (a) a hand-operated water pump with lever and spout; (b) a jointed arm on a base with a two-finger gripper, cables running to a control cabinet; (c) a ceiling fan with three blades and a regulator switch on the wall; (d) a small round floor-cleaning robot with a bumper strip and a camera on top; (e) a bicycle with pedals, chain and wheels; (f) a four-propeller drone with a camera underneath
Work through two of them.
Take (b), the jointed arm. It has four bends, each with a motor housing. It has a gripper at the end that can be unbolted and swapped. It has cables running to a cabinet, which is where its controller sits. That is sensing, deciding and acting, all visible. It is a robot.
Now take (a), the hand-operated pump. It has one pivot, no motor, no wiring, and nothing anywhere on it that could collect information. Its task is fixed by its shape. Whatever you do to it, it will only ever lift water. It is a machine.
2.5 The Progress from Robots to Cobots
The first robots to be used in real production were industrial robots — large, powerful arms built for factory work. They were a huge success at what they were designed for: welding, painting, lifting engine blocks, doing the same movement to the same accuracy for years on end.
They were also dangerous to stand next to. A big industrial arm swings fast and carries enough force to injure anyone in its path, and it has no way of noticing a person there. So the accepted way of using one was to put it inside a fenced cell, with the gate wired so that opening it cut the power. People and robots worked in separate spaces, as Fig. 2.5-1 shows.
Diagram
Fig. 2.5-1 – a horizontal three-panel strip showing the shift over time. Panel 1: a fixed automatic machine stamping parts, captioned “does one job, cannot be retaught”. Panel 2: a large industrial arm inside a wire safety cage with a warning sign on the gate and a worker standing outside the cage, captioned “can be reprogrammed, but must be kept away from people”. Panel 3: a small arm on a workbench with a worker at the same bench and no cage, captioned “safe to share a bench with a person”. An arrow runs left to right beneath the panels marked “time”
What the fence cost
Keeping people out solved the safety problem, but it created three others.
Space. A cell needs floor area for the fence, the gate and a clear zone inside. In a small workshop that space is often more valuable than the robot.
Cost and skill. These arms were expensive to buy, and programming one meant writing instructions in a specialised language. A factory needed a trained robot programmer on the payroll. A workshop with eight employees could not justify either.
Lost teamwork. Some jobs are best done by a person and a robot together. Fitting a wiring loom into a dashboard needs human judgement about which clip goes where, and it also needs something to hold the loom steady at an awkward angle for several minutes. Behind a fence, that partnership is impossible. Either the robot does the whole job or the person does.
The next step
So a second design direction opened up. Instead of asking how to keep people away from a robot, engineers asked how to build a robot that people could safely stand next to.
That meant lighter arms, so there was less force to control. It meant sensors that detect a person or a touch and slow or stop the arm at once. It meant rounded casings without sharp corners or pinch points. And it meant a way of teaching the robot a task that did not need a programming specialist.
A robot built on those lines is not a smaller version of an industrial robot. It is a different design, aimed at a different situation — the shared bench rather than the fenced cell. That is what the next section is about.
2.6 What Is a Cobot?
A robot designed to share a workbench with a person, rather than to be fenced away from one, is called a collaborative robot (cobot). The name says what it is for: collaborating, which means working together with somebody rather than working instead of them.
The important words there are “shared workspace”. An industrial robot has a working area and people have a separate one, with a fence in between. A cobot’s working area and the person’s working area overlap, and both of them use the overlap at the same time. Fig. 2.6-1 shows the difference.
Diagram
Fig. 2.6-1 – two pairs of circles. Left pair: two circles side by side, not touching, labelled “person’s working area” and “industrial robot’s working area”, with a fence drawn in the gap between them. Right pair: two circles that overlap, labelled “person’s working area” and “cobot’s working area”, with the overlapping region shaded and labelled “shared workspace”
Safe, interactive, easy to use
Three design ideas make that overlap possible.
The cobot is built to be safe. It is lighter than an industrial arm, it moves at moderate speed, its casing is rounded, and it carries sensors that watch for a person coming close or for the arm meeting unexpected resistance.
It is built to be interactive. It is meant to be approached, touched and worked with, so a person can hand it a part, take one from it, or move its arm out of the way.
It is built to be easy to use. A cobot is usually taught by leading its arm by hand through the movements of the job while it records the positions, or by tapping through a few steps on a tablet. No specialist programming language is needed.
Because a person and a cobot can work in the same place at the same time, each can do the part of the job it is better at. The person brings judgement and handles the awkward, varying steps. The cobot brings steadiness and repeats the tiring ones without complaint. That combination is what makes cobots useful, and it is why they are usually reported to raise output and to make the work less exhausting for the people doing it.
2.6.1 Humans in the Loop
Look again at the working loop from earlier in this chapter: sense, process, decide, act, feed the result back. In an ordinary industrial robot that loop is closed entirely inside the machine. A person sets it running and then stays out of the way.
In a cobot, a person sits inside the loop while it is running. This is what is meant by a human in the loop — an arrangement where a person supplies part of the sensing, part of the judgement, or part of the action, on every cycle of the task rather than only at the start.
There are three places where the person enters.
Setting the task by showing it
Most cobots are taught by demonstration. The operator takes hold of the arm, leads it by hand through the movements the job needs, and presses a button at each position to record it. The cobot stores those positions and plays the sequence back. Fig. 2.6-2 shows this being done.
Diagram
Fig. 2.6-2 – a worker holding a cobot arm just behind its wrist and guiding it towards a tray of parts on the bench. A small teaching pendant or tablet is shown in the worker’s other hand with a “record position” button marked. Dotted outlines behind the arm show two earlier positions already recorded
Here the person is doing what a programmer would otherwise do, but with their hands instead of with code. Nobody has to know a robot programming language to set up the job.
Watching and stopping
The person is also the final check. A cobot has an emergency stop button within reach, and a worker who sees something going wrong — a part slipping, a tool loose — stops the arm at once. The robot’s own sensors handle the situations it was built to detect. The person handles the ones nobody predicted.
2.6.2 Working as a Multi Robot System
One cobot on a bench is the simplest arrangement, and it is not the usual one. Most real jobs have several stages, and it makes little sense to build one arm that can do all of them badly when you can put two or three simpler arms in a row, each doing one stage well.
When several robots work together on one job, sharing the space and passing work between them, the arrangement is called a multi robot system. A cobot is designed to fit into such a system, because working with others — people or robots — is the point of it.
Think of a school canteen counter at lunchtime. One person serves rice, the next serves dal, the next hands over a spoon. Nobody does the whole meal. The tray moves along and each person adds one thing at the point where they stand.
What has to be arranged
Splitting a job across robots brings its own problems, and they all come down to coordination.
Each robot has to know when its turn has come. A part arriving at the wrong moment gets dropped or double-handled. Usually a sensor at each station reports that a part is in position, and only then does that arm begin.
They have to keep out of each other’s way. If two arms can reach the same patch of bench, their movements have to be timed so that only one is in that patch at a time.
And they have to be told about each other. The stations are linked, so that a stoppage at one is known to the rest and the line does not keep feeding parts into a station that has halted.
Diagram
Fig. 2.6-3 – a straight workbench seen from above with three stations in a row, joined by a short belt. Station 1: a cobot arm picking a part from a bin and setting it on the belt. Station 2: a second cobot arm fitting a small component onto the part. Station 3: a worker checking the finished part and packing it. Arrows along the belt show the direction of travel; a dashed line joins all three stations, labelled “stations linked so each knows when the previous one is ready”
Fig. 2.6-3 shows a three-station line of this kind, with two cobots and a person in it. Note that the person is one of the stations, not a supervisor standing behind them.
The slowest station sets the pace
There is a useful consequence of splitting a job this way. However fast the other stations are, the line as a whole can only turn out work as fast as its slowest station.
2.7 Key Characteristics of Cobots
You have now met a cobot from several directions — how it stays safe, how a person works inside its loop, how it fits into a line with others. It is worth gathering the qualities that follow from all of this into one list, because these six are what people mean when they say a robot is a cobot. Fig. 2.7-1 shows them together.
Diagram
Fig. 2.7-1 – a cobot arm drawn small in the centre of the page with six labelled boxes arranged around it, each joined to the centre by a line: “Safety features”, “Easy to teach”, “Adaptability”, “Shared workspace”, “Versatility”, “Affordability”. Keep the boxes evenly spaced around the arm
Safety features. This is the first quality because everything else depends on it. Force sensing at the joints, a rounded casing with no pinch points, moderate speed, a vision system watching the area, and an emergency stop within reach. Together these are what allow the fence to come down.
Easy to teach. A cobot is set up by leading its arm through the job or by tapping through steps on a tablet. That matters for a small unit that has no robot programmer and cannot afford to hire one. The person who knows the job is the person who sets it up.
Adaptability. Because the task is a recorded sequence rather than fixed metal, a cobot can be switched to a different job quickly. A small pharmacy packing unit might have a cobot filling strips into cartons in the morning and, after a short reteaching, placing leaflets into a different carton size in the afternoon.
Shared workspace. A cobot is built to have a person inside its reach while it runs, so the two can hand work back and forth and stand at the same bench. This is the quality that gives cobots their name.
Versatility. The same basic arm turns up in very different places — assembling parts in a workshop, packing boxes in a food unit, checking finished pieces under a camera, handling samples in a laboratory. Change the tool on the end and the job changes with it.
Affordability. A cobot costs less to buy than a large industrial arm, and it needs no cage, no separate cell and no specialist programmer. Those savings are often larger than the price difference, and they are what brings a robot within reach of a business with a dozen employees rather than a thousand.
Where these qualities run out
None of this comes free. To be safe near people, a cobot is light and slow, so it carries a modest load and takes longer over each cycle than a fenced industrial arm would. For work that is heavy, fast or repeated in enormous volumes, that trade is a poor one.
Being easy to teach has a limit too. Hand guiding suits a job made of a series of positions. A task that needs the robot to respond differently to many varied situations still has to be programmed properly.
2.8 Cobots Compared with Traditional Industrial Robots
Both are robots. Both sense, process, act and use feedback. Both can be reprogrammed. The difference lies in the question each was designed to answer.
A traditional industrial robot answers: how do we get this heavy job done as fast and as accurately as possible? A cobot answers: how do we get this job done with a person standing right there? Almost every other difference follows from those two questions. Table 2.2 sets them out.
Choosing between them
The comparison is not a scoreboard with a winner. It is a set of trade-offs, and the right choice depends on the job.
If a workshop makes eight different products in small batches, with a worker involved at each stage, a cobot fits. The job changes often, the volumes are small, and the person is needed anyway.
If a plant welds the same car body panel ten thousand times a month, an industrial robot fits. Nothing changes, nobody needs to be inside the cell, and the speed and force are exactly what the job demands.
Many factories run both. The heavy, unchanging work goes behind a fence; the varied work that needs a person goes to a cobot on an open bench.
2.9 Why Cobots Matter
A new design only matters if it does something the old one could not. Cobots are slower than industrial robots and they carry less. So it is fair to ask what they actually change.
They put the two kinds of strength together
People and robots are good at opposite things. A person can look at an odd-shaped part and work out at once which way it should go, cope with a batch that arrives slightly different from the last one, and notice that something is wrong before it becomes a fault. A robot can hold a load steady without shaking, repeat a movement identically for hours, and work to the same accuracy at the end of a shift as at the start. Fig. 2.9-1 sets these side by side.
Diagram
Fig. 2.9-1 – two columns headed “What a person does well” and “What a cobot does well”. Under the first: judging an unfamiliar situation, handling variation, noticing something is wrong, deciding what to do next. Under the second: holding steady for long periods, repeating a movement identically, keeping the same accuracy all shift, lifting without tiring. A brace joins both columns to a single box at the bottom labelled “one job, done better than either could alone”
Keep them apart, behind a fence, and you have to choose one set of strengths for the whole job. Let them share a bench and you can give each step to whichever of the two is better at it.
They bring automation within reach of small units
This is the change that has mattered most in practice. A fenced cell with a specialist programmer was only ever affordable to a large factory. A cobot needs no cell, little floor space and no programmer, so a workshop with a handful of employees can use one.
That opens automation to the kind of unit that makes up most of Indian manufacturing — a small engineering shop, a food packing unit, a garment workroom.
They change tiring work rather than removing the worker
The tasks cobots take over tend to be the ones that wear people down: lifting the same weight hundreds of times, holding a part still at an awkward angle, making an identical movement all day. Handing those to a machine while the person keeps the judgement is a real improvement in the working day, not just in the output figures.
2.10 Summary
A machine is a device that takes in energy and uses it to do a useful task, making work easier, faster or more accurate than a person could manage alone. Its parts — gears, levers, pulleys, shafts, motors — pass movement and force from one to the next.
A machine does exactly what its design allows, every time. It cannot check what is in front of it and cannot change what it does.
A robot is a programmable machine that senses its surroundings, works out what to do from its instructions, and acts on those surroundings. Every robot is a machine; not every machine is a robot.
A robot’s sensing is continuous, its task is held in a program that can be changed, and it holds the conditions its task needs by correcting itself.
An autonomous robot carries out its whole task on its own. A semi-autonomous robot does part of the task and takes the rest from a human operator.
Every robot runs the same working loop: task, sensing, processing, deciding, actuator movement, interaction with the object, feedback, and back again.
A sensor detects a physical quantity — light, sound, distance, heat, touch, position — and converts it into an electrical signal. Common ones include cameras, microphones, ultrasonic sensors, LIDAR, touch sensors, temperature sensors, angle sensors and inertial sensors.
The controller reads the sensor signals, turns them into data, applies the rules in the program, and sends out the next instruction. In a small robot it is a microcontroller.
An algorithm is the set of rules written as an exact list of steps. Simple robots follow rules written in advance; more advanced ones use machine learning to find patterns in stored data.
An actuator converts a signal from the controller into physical movement. Motors, servo motors, wheels and grippers are actuators, and what a robot can do to an object depends on which actuators it has.
Feedback is information about what an action actually achieved, sent back so the controller can compare result with target and correct the error.
Early industrial robots were fast, powerful and unable to sense people, so they had to be kept inside fenced cells. That cost floor space, needed specialist programming, and made human–robot teamwork impossible.
A cobot, short for collaborative robot, is designed to work safely alongside people in a shared workspace. Its safety comes from being light and slow, from force sensing at the joints, from rounded casing with no pinch points, and from an emergency stop within reach.
A cobot works with a human in the loop: the person teaches it by hand guiding, hands work to and from it, judges the awkward steps, and can stop it at any moment.
In a multi robot system, several robots share one job, each handling a stage and coordinating with the rest. The slowest station sets the pace of the whole line.
The key characteristics of cobots are safety features, easy teaching, adaptability, a shared workspace, versatility and affordability.
Cobots and industrial robots suit different work. Cobots fit varied, small-batch jobs done alongside people; industrial robots fit heavy, fast, large-volume work done behind a fence.
Cobots matter because they combine what people do well with what robots do well, bring automation within reach of small units, and take over the tiring, repetitive parts of a job rather than the parts that need judgement.
2.11 Key Terms
Actuator — the part of a robot that converts an electrical signal from the controller into physical movement.
Algorithm — a set of rules written out as an exact list of steps for solving a problem or carrying out a task.
Autonomous — describes a robot that carries out its whole task on its own, without a person guiding each step.
Collaborative robot (cobot) — a robot designed to work safely alongside people in a shared workspace, sharing a task with them instead of being fenced away.
Controller — the computing part of a robot, which reads sensor signals, turns them into data, applies the rules in the program and sends out the next instruction.
Data — information written as numbers that a program can work with.
Feedback — information about what an action actually achieved, sent back to the controller so it can compare the result with the target and correct the difference.
Human in the loop — an arrangement in which a person is one of the steps of a robot’s working cycle, supplying part of the judgement or action on every pass rather than only at the start.
Industrial robot — a large, powerful robot arm built for high-speed, heavy factory work, normally operated inside a fenced cell away from people.
Input — whatever goes into a machine or robot: the energy supplied, the material fed in, or the readings collected.
Machine — a device that takes in energy and uses it to do a useful task, making the work easier, faster or more accurate than a person could manage alone.
Machine learning — a way of getting a program to work out patterns from large amounts of stored data, instead of being given every rule in advance.
Microcontroller — a single chip holding a processor, memory and connection pins, used as the controller of a small robot and wired directly to its sensors and actuators.
Multi robot system — a set of robots working together on one task, each handling a stage and coordinating with the others so the work passes correctly from one to the next.
Output — what comes out of a machine or robot: the movement, the force or the finished item.
Program — the set of instructions, written by a person and stored in the robot, that tells it what to do in each situation it may meet.
Programmable — able to be given a new set of instructions, so the same hardware can be made to do a different task.
Robot — a programmable machine that senses its surroundings, works out what to do from its instructions, and acts on its surroundings to carry out a task.
Semi-autonomous — describes a robot that handles part of its task on its own and takes the rest from a human operator.
Sensor — a device that detects a physical quantity such as light, sound, distance, heat, touch or position and converts it into an electrical signal the robot’s electronics can read.
Task — the one job a robot has been built and programmed to do, which every other stage of its working loop serves.
2.12 Objective Questions
Which one of these is a machine but not a robot?
- A floor-cleaning robot
- A pair of scissors
- A drone with a camera
- A robotic arm with a gripper
Fill in the blank: The part of a robot that converts an electrical signal from the controller into physical movement is called an ______.
State whether true or false: A cobot has to be kept inside a fenced cell while it is working.
Fill in the blank: Information about what an action actually achieved, sent back to the controller so that it can correct the difference, is called ______.
How is a cobot usually taught a new task?
- By writing instructions in a specialised programming language
- By replacing its controller board
- By leading its arm through the positions by hand and recording them
- By changing the gears inside its joints
In a line where three stations each handle one stage of the same job, the number of finished items per hour is set by
- the fastest station
- the first station in the line
- the slowest station
- the total number of robots in the line
Match each part in Column A with what it does, in Column B.
| Column A | Column B |
|---|---|
| a. Sensor | i. Converts a signal into physical movement |
| b. Controller | ii. Reports what the action actually achieved |
| c. Actuator | iii. Converts a physical quantity into an electrical signal |
| d. Feedback | iv. Applies the rules in the program and picks the next action |
2.13 Concept Questions
Define a robot. State the one ability it has that an ordinary machine does not.
Give two differences between a machine and a robot. Support each difference with one example.
Why must a robot keep sensing all the while it works, instead of checking its surroundings once when it starts?
An ultrasonic sensor on a robot reports a round-trip time of 0.002 s for its sound burst. Taking the speed of sound in air as 343 m/s, calculate the distance between the robot and the object in front of it. Show each step of your working, and give the answer in centimetres.
A cobot works on an open bench with no fence around it. Name three design features that make this safe, and explain briefly how each one reduces the risk to the person beside it.
Explain what is meant by a robotic system with a human in the loop. Describe two points at which a person enters a cobot’s working cycle.
What would happen to a robot’s work if its feedback stage were removed but everything else were left in place? Explain your answer using an example of a task going slightly wrong.
2.14 Case-Based Questions
- A small unit in Jaipur assembles geometry boxes. The work is split across three stations on one bench, and a part moves from each station to the next as soon as it is finished.
Station 1: a cobot places the empty box on the belt — 10 s per box Station 2: a worker sets the instruments into the box — 8 s per box Station 3: a cobot closes the lid and seals the box — 25 s per box
- How many finished boxes will this line turn out in one hour if it runs without a break? Show your working.
- The owner wants a higher output and plans to buy a faster cobot for Station 1. Explain whether this will help.
- Station 2 is handled by a person rather than a cobot. Suggest one reason why that stage has been left to a worker.
- A workshop makes six different kinds of metal brackets in small batches. The batches change every few days. Each bracket has to be held steady while a worker checks the drilled holes and stamps a serial number on it. The workshop has floor space for one bench and no trained robot programmer.
- Would a cobot or a traditional industrial robot suit this workshop better? Give two reasons drawn from the workshop’s situation.
- Describe how the work would be shared between the robot and the worker at that bench.
- The bracket has sharp edges after drilling. State one safety point the workshop must still attend to, even though the arm is a cobot.