The Robotic Arm

The 6-axis industrial robot was not invented for architecture. It was perfected over decades on automotive lines, where the same arms spot-weld, glue, handle and paint car bodies millions of times with brutal reliability. Names you will meet - KUKA, ABB, FANUC, Yaskawa - are industrial giants; Universal Robots (UR) popularised the smaller, safer collaborative arm. Architecture simply adopted this mature, cheap-because-mass-produced hardware and pointed it at new problems.

What architecture borrowed is specifically the articulated arm: a chain of rigid links connected by six rotary joints, mounted on a fixed base, ending in a tool-mounting flange (also called the end-of-arm plate). Count the joints and you have the '6-axis' name. Six is not arbitrary: three joints position the wrist anywhere in a volume, and three more orient the tool to any angle at that point. Fewer axes and you lose orientations; that full six-degrees-of-freedom capability is exactly what lets a robot approach a workpiece the way a craftsperson's hand would - from any side, at any tilt.

Compare the robot to the CNC machines of Module 3. A gantry CNC router is rigid, accurate and fast - but it is a box that moves a spindle in X, Y and Z, and it will only ever do that. Its geometry is its destiny. The robot trades some of that rigidity and absolute accuracy for something architecture values more: reconfigurability. Monday it mills foam; Tuesday, with a new end-effector and a new program, it stacks bricks; Wednesday it prints clay. No other machine flexes across the subtractive, additive, formative and assembly families with a tool swap.

That flexibility is why research labs like ETH Zurich (Gramazio Kohler), ICD Stuttgart and the Association for Robots in Architecture built a whole field on the standard industrial arm rather than commissioning custom machines. The honest trade-off: a robot is typically less accurate than a good CNC (repeatability often around a tenth of a millimetre, but positional accuracy across the whole envelope is looser), less stiff against cutting forces, and far harder to program. You choose a robot when the job needs its reach, its angles or its versatility - not when a dedicated machine would simply do it better. Put bluntly: if a laser or a 3-axis router can make your part, use the laser or the router; reach for the arm when the geometry, the orientations or the sheer range of processes leaves you no simpler option.

Every industrial arm is specified by three headline figures, and reading them is the first practical skill. Payload is the mass the arm can carry at full extension - including the end-effector itself, not just the workpiece. A 10 kg-payload arm sounds ample until your spindle-plus-mount weighs 7 kg, leaving 3 kg for the job. Architectural arms range from a few kilograms (a UR5, ~5 kg) up to heavy KUKA or ABB models carrying hundreds of kilograms.

Reach is how far the tool centre can get from the base - roughly the radius of a hemisphere. A common lab arm reaches around 2 to 2.7 metres; big arms exceed 3 metres, and mounting the robot on a linear track (a 7th axis) or a gantry extends that indefinitely along a wall or table. Repeatability is how precisely the arm returns to the same commanded point, typically +/- 0.03 to 0.1 mm - excellent, and often confused with accuracy, which is how close it gets to a new point it has never visited and is usually worse. A worked read: a 'KUKA KR 60, 2.03 m reach, 60 kg payload, +/- 0.06 mm repeatability' arm will hold a router steadily across a two-metre panel and come back to a drilled hole almost perfectly - but do not assume the first cut lands exactly where your model says without calibration.

Two more specs quietly matter. Stiffness - how much the arm deflects under load - is where robots trail CNC machines: push a spindle into hardwood and the arm flexes a little, so heavy cutting is done in light passes rather than one deep bite. And speed / duty cycle - the automotive heritage means these arms run reliably for years of continuous shifts, which is why a mass-produced industrial arm is cheaper and more dependable than any bespoke machine architecture could commission. Read all five together - payload, reach, repeatability, stiffness, duty - and you can tell at a glance whether an arm suits milling foam (easy), milling steel (no), or laying brick all day (yes).

An industrial robot is a heavy, fast, powerful machine that will move exactly where told with no awareness of a person in its path. A standard arm can swing a multi-kilogram tool through a two-metre arc faster than you can step back, and it does not stop because you are there. This is the single most important thing to internalise in this module: the hazard is not just the robot, it is the entire volume its tool can sweep - the reach envelope. Treat every point the tool can reach as live.

So you never run a bare industrial arm in a shared space. It lives in a cell: a guarded zone with physical fencing or light curtains, an interlocked gate that stops the robot if opened, and emergency stops within reach. Collaborative robots (cobots) like the UR series are force-limited and designed to work near people, which relaxes some guarding - but only after a proper risk assessment, and a cobot swinging a spinning router or a hot extruder is no longer safe to touch regardless of its rating. All robot work in this course assumes trained, supervised operation inside a properly guarded cell. When we make the arm fabricate in later lessons, this cell is the unspoken frame around every exercise.

The bare arm is not the end of the story - how you mount it multiplies what it can do. The most common extension is a linear track (a 7th axis): the whole robot rides on a rail, so an arm with a two-metre reach can work along a ten-metre timber beam or a whole wall of brick. Architecture loves this because buildings are bigger than any arm's envelope. Arms are also mounted inverted (hung from a gantry or ceiling) to work down onto a large table, mounted on gantries for wide-span coverage, or paired as multi-robot cells where two arms cooperate - one holding a part while the other works it. Each of these is still the same six-axis logic; you are just changing where the base sits and how far it can travel.

The articulated 6-axis arm is also not the only robot morphology, and knowing the family helps. SCARA robots (a few axes, fast, rigid, working mostly in a horizontal plane) dominate flat pick-and-place. Delta robots (spider-like, three arms to a shared platform) are blindingly fast but limited in reach and orientation - think high-speed sorting. Cable-suspended and mobile robots are the frontier for on-site construction, where a fixed base is impossible: a robot that drives around a site, or hangs from cables over a large area, to print or place at building scale. For architectural fabrication, though, the versatile 6-axis arm - optionally on a track - remains the default, precisely because its full orientation freedom suits the endless variety of building components. When you scope a job, 'which arm' is really two questions: which arm, and mounted how.