End-Effectors & Tooling

Every end-effector bolts to the robot's flange, a standardised circular face at the wrist (its pattern is set by an ISO mounting standard so tools are interchangeable across arms). Through or beside that flange run the tool's utilities: electrical power and signals, compressed air, coolant, filament or material feed. Designing an end-effector is as much about routing those services as about the tool itself.

The most fundamental end-effector is the gripper - the robot's hand, for pick-and-place and assembly. Grippers come in families: parallel-jaw grippers (two fingers that clamp, the workhorse), vacuum / suction grippers (a cup that grabs flat sheets, glass, panels - superb for handling large smooth material), and magnetic grippers for ferrous parts. Assembly robotics - stacking timber, laying brick, placing panels - lives on grippers. The craft is matching the gripper to the part: a suction cup that handles a glass sheet beautifully will drop a porous brick; a jaw sized for a brick will crush a fragile tile. Grip force, contact area and part fragility all have to be reasoned about before the arm ever moves. And a gripper does not only need to hold the part - it must hold it in a known position, because the robot assumes the grasped part sits exactly where the grip placed it; a component that shifts in the jaws mid-move throws off every placement that follows, which is why assembly grippers are often designed to seat a part against a hard stop rather than merely clamp it.

Beyond grippers are the process tools that make the robot a fabricator in each of our families. For subtractive work, a spindle - essentially a high-speed router or mill motor - lets the arm carve foam, timber, plaster or plastic from any angle, the robot's flexibility giving it access a 3-axis CNC cannot match (undercuts, sculpted surfaces, work approached from the side). For additive work, an extruder deposits material layer by layer: a paste extruder for clay or concrete (as WASP and construction printers use), a pellet or filament extruder for large-scale plastic, even spatial printing that draws in mid-air. For a fast formative-adjacent cut, a hot-wire (a heated taut wire) slices expanded foam (EPS/XPS) into ruled and curved surfaces in seconds - the standard way robots make foam moulds and formwork for cast concrete.

Others you will meet: welding torches (the arm's original factory job), knives and routers for sheet trimming, polishing and grinding heads for finishing, pick-up and placing heads for tiles or bricks, and printing/spraying heads. Each process tool brings its own settings from earlier modules - spindle speed and feed rate for milling, layer height and flow for extrusion, wire temperature and speed for hot-wire cutting - now delivered by an arm that can also tilt and orient the tool through the cut. Critically, each tool also brings its own hazard: a spinning spindle, a hot wire, a live torch, hot extruded material. The end-effector is usually the most dangerous part of the whole cell.

Because architecture keeps inventing new processes, a huge amount of robotic-fabrication research is really custom end-effector design. Labs like ETH Zurich's Gramazio Kohler and ICD Stuttgart routinely build bespoke tools: an end-effector that grips a timber beam AND drives a screw to assemble a frame; a head that winds carbon or glass fibre in space to build a shell; a tool that both extrudes and smooths. The end-effector is where a research idea becomes a real capability - and designing one is a genuine skill combining mechanics, pneumatics, electronics and control.

Many advanced end-effectors also sense, not just act. A tool can carry a force/torque sensor so it presses with controlled pressure (essential for polishing, or for assembly that must not jam), a camera or laser scanner to locate parts or inspect the result, or a probe to touch off a reference and correct for a workpiece that is not exactly where the model assumed. This sensing is what lets robotic fabrication cope with the messiness of real materials - a slightly warped board, a brick that is a millimetre oversize - rather than blindly executing a path. It is also the frontier where robotics meets the digital-twin and adaptive-fabrication ideas of Module 9.

The whole promise of a general-purpose arm is realised through tool changing. A manual change - unbolt one tool, bolt on the next - is fine for a workshop that does one process at a time. For a robot that mills, then places, then finishes within one job, an automatic tool changer (ATC) is used: a master coupling on the flange docks to a matching plate on each tool, passing power, air and signal through the connection. The robot drives to a tool stand, releases the current tool into its cradle, picks up the next, and continues - all under program control.

Here the previous lesson pays off: each tool carries its own TCP. The spindle's working point sits 210 mm beyond the flange; the extruder nozzle sits somewhere else entirely; the gripper's grasp centre somewhere else again. When the arm changes tools, the controller loads that tool's stored TCP so it always knows where the new working point is - otherwise every post-change move would be wrong. This is why TCP calibration is done carefully, per tool, and stored. Safety runs through all of it: tool changes happen with motion stopped and the cell secured, because a mis-docked tool - or a spindle that starts while being handled - is exactly the kind of event guarding and interlocks exist to prevent. Never change or handle an end-effector on a live, un-secured arm.

An end-effector is not just chosen for what it does - it has to be carriable, and this is where beginners get caught. The payload rating from Lesson 6.1 is really about more than mass: it is about the moment, the mass multiplied by how far its centre of gravity sits from the flange. A 3 kg spindle mounted close to the wrist is easy; the same 3 kg cantilevered 400 mm out loads the wrist joints far harder and eats into the effective payload. Long, heavy, off-axis tools also carry inertia - at speed, the arm must accelerate and stop that mass, and too much of it forces slower moves or risks overshoot and wobble that spoil precision. So a good end-effector is designed compact, light and balanced, with its centre of gravity close to the flange axis.

Then there are the utilities, the unglamorous half of the job. A spindle needs power and often coolant; a pneumatic gripper needs compressed air; an extruder needs a material feed and heating power; a sensor needs data and power. All of these arrive through a dress pack - the bundle of cables and hoses running along the arm - and that bundle must flex through every pose without snagging, stretching or getting caught in the tool's own path. Many a beautifully-designed end-effector fails on its first full move because a hose fouled the fixture or a cable ran out of slack at full reach. Designing the tool means designing where its services live at every point in the motion - which, again, is exactly the kind of thing you catch in simulation rather than on the floor.