What a Robot Is

Start with the cleanest working definition we can carry into everything that follows. A robot is a programmable machine that can sense its environment, compute a decision from what it senses, and act physically on the world - and can be reprogrammed to do a different task. Every word earns its place. *Machine* - it is physical, it moves real matter, unlike a piece of software. *Senses* - it takes in information about its surroundings, through cameras, touch, distance sensors, position encoders and more; it is not blind. *Computes* - between sensing and acting sits a decision, however simple, made by a program rather than by a human pulling a lever in real time. *Acts physically* - it does something in the world: moves, grips, places, drills, extrudes. And *programmable / reprogrammable* - its behaviour is set by a program you can change, which is what lets one machine do many jobs and what makes it worth the cost.

Those three verbs form a loop, not a line. The robot senses the current state of the world, its program computes what to do, it acts, the world changes as a result, and it senses again - a continuous sense-compute-act loop (sometimes called sense-plan-act). A car's cruise control is a baby version of the same loop: sense speed, compute throttle, act on the engine, sense again. A construction robot laying a brick runs a far richer version: sense where the last brick sits and where this one must go, compute the motion, place it, check the result, adjust for the next. The loop is the heartbeat of every robot in this course.

Notice what the definition does not require. It does not require a humanoid shape - most real robots look nothing like a person. It does not require intelligence in any grand sense - a robot can be quite "dumb" and still qualify. It does not require full independence - a robot can take direction and still be a robot. What it does require is that sensing and deciding happen through a program, on the machine's side of the line, rather than every motion being dictated live by a human hand. That single requirement is what we will use, again and again, to tell a real robot from the many machines and gadgets that borrow the name.

A robot is an assembly of four kinds of part, and once you can name them you can read almost any robot - including the building ones - at a glance. Actuators are the muscles: the motors (electric, hydraulic or pneumatic) that move the robot's joints and limbs. They convert electrical or fluid power into motion and force, and their strength, speed and precision largely set what the robot can physically do - a robot that must lift a concrete panel needs very different actuators from one that dispenses a thin bead of sealant.

Sensors are the senses. Some look outward at the world - cameras for vision, lidar and ultrasonic sensors for distance, force and torque sensors for touch and contact. Others look inward at the robot's own body - encoders on each joint report its exact angle, so the controller always knows the pose of the machine. Without sensors a robot is acting blind, running a motion open-loop and hoping the world matches its assumptions; with them it can close the loop and correct. On a construction site, where the world rarely matches assumptions, sensing is not a luxury but the whole game.

The controller is the brain: the computer, and the program running on it, that takes sensor data, decides what to do, and commands the actuators. This is where the robot's behaviour actually lives. Crucially, it is the part you reprogram - change the program and the same physical robot performs a different task, which is the economic magic of robotics and the reason a general-purpose arm can be repurposed across jobs. The end-effector is the hand: the tool at the end of the arm that does the actual work - a gripper, a welding torch, a drill, a spray head, or, for the printers later in this course, an extrusion nozzle. One robot body can wear many end-effectors, and the choice of end-effector often matters more to a construction task than the robot itself. Actuators to move, sensors to perceive, a controller to decide, an end-effector to act on the work: four parts, one machine. Keep them in mind and every robot arm, crawler and gantry in the rest of the course resolves into the same simple anatomy.

To do useful physical work a robot has to move, and degrees of freedom (DOF) is the vocabulary for how. A degree of freedom is one independent way the machine can move - one axis it can translate along or one axis it can rotate about. Your own arm has many: the shoulder, elbow and wrist together let your hand reach a huge range of positions and turn to almost any angle. The more degrees of freedom, the more freely the tool at the end can be placed and oriented, at the cost of more motors, more complexity and more to control.

The archetypal robot - and the workhorse of robotic construction and fabrication - is the robotic arm, usually with six degrees of freedom: six joints, each a motor that rotates about an axis. Why six? Because in three-dimensional space, positioning a tool fully takes six numbers - three for *where* it is (up-down, left-right, forward-back) and three for *how it is oriented* (the tilt and twist, often called roll, pitch and yaw). Six joints, arranged in series, let the arm put its end-effector at any reachable point AND at any angle - which is exactly what you need to weld a joint from the right direction, lay a brick flat, or point a nozzle along a curve. Some arms have seven joints, giving a "redundant" extra freedom to reach around obstacles; simpler machines have fewer.

Degrees of freedom also define the robot's working envelope - the volume of space its end-effector can reach. This matters enormously in construction: a six-axis arm is astonishingly dexterous but can only reach as far as it is long, which is why building-scale robots are so often either very large, mounted on rails or vehicles so the whole arm can move, or configured as gantries that carry a simpler tool over a wide area. We will meet all of these. For now, hold the idea: degrees of freedom describe how a robot can move, six is the number that unlocks full control of a tool in space, the six-axis arm is the icon of the field, and reach is always a hard physical limit you design around. These are illustrative principles, not machine specifications - the exact reach, payload and precision of any real robot come from its manufacturer's verified data.

The definition earns its keep at the edges, where robots are confused with two near-neighbours. The first is the plain machine. A washing machine, a concrete mixer, a power drill, a lift - all are machines that do physical work, some with motors and timers, but they repeat a fixed action and do not sense and decide their way through a changing world. A concrete mixer turns whether the mix is right or wrong; it has no senses, no judgement, no reprogrammable decision. It is automation, but not a robot. The presence of motors and even some automatic sequencing does not make a robot; the missing ingredient is the sensing-and-deciding loop under program control.

The second, and the one that does the most mischief in construction marketing, is the remote-controlled tool. A drone flown entirely by a pilot's sticks, a demolition machine steered by an operator with a handset, a camera on a rover someone drives - in each case a human is making the moment-to-moment decisions and the machine is simply relaying and executing them. That is teleoperation, and it is genuinely useful - it takes the human out of danger and extends reach - but a teleoperated machine is not, by itself, a robot in the full sense, because the deciding happens in the human, not in a program on the machine. The same hardware can cross the line: add a program that senses and decides for itself, even partly, and it begins to become a robot. This is exactly the spectrum Lesson 1.4 unpacks.

Why insist on the distinction? Because the border is where hype lives. "Autonomous" is the most over-sold word in construction robotics, stuck on machines that are really teleoperated or run a fixed routine in a rigged-up demo. A reader armed with the simple test - *does the machine sense its surroundings and decide for itself under a program it can be given, or is a human deciding every move, or is it just repeating a fixed action?* - can cut through most of it. None of this is a put-down of the simpler machines; teleoperation and fixed automation are valuable and everywhere in construction. It is about naming things accurately, so that when this course says "robot" you know precisely what is and is not being claimed - and so the binding questions of what any given machine can safely and reliably do are judged on evidence, not on the word stuck to it.