Multi-Robot & Cooperative

A single six-axis arm is astonishingly capable, but it is fundamentally one hand. Think of the everyday reality of building: a beam has to be held level while it is bolted; a printed strand needs something to land on; a long panel must be steadied at both ends as it is set. A person never does these alone - they ask for help, and a second pair of hands holds while the first fastens. A lone robot has no one to ask, so it falls back on jigs, clamps and temporary supports - physical fixtures that stand in for the missing second hand, and that must be built, positioned and later removed.

Multi-robot fabrication replaces those passive fixtures with active partners. Give the cell a second arm and it becomes the clamp - but an intelligent, repositionable, precise one. The classic division is one holds, one works: one arm grips and steadies a member exactly where the model wants it, while the second drives a screw, welds a joint, or extrudes against it. No temporary scaffolding to build and strip, no re-fixturing between steps, and the holding arm can move the part to present each joint conveniently to the working arm. It is the single most useful thing a second robot brings, and it maps directly onto how skilled human crews have always worked.

The deeper point is that a jig is a frozen decision while a second arm is a live one. A clamp or fixture holds a part in one pose; if the next step needs the part rotated, you build another fixture or stop and re-fix it by hand. A holding arm just moves - it can lift the member, tilt it to bring a joint into the working arm's best reach, hold it there while the fastener goes in, then rotate it for the next joint, all from the same instructions that drive the rest of the build. Fixtures also have to be designed and made for each new part, which is fine for mass production and painful for the one-off, differentiated work that digital fabrication exists to enable. Two arms give you re-fixturing for free, in software, on the fly - which is precisely why cooperation matters most for exactly the bespoke, non-repetitive structures this course keeps returning to.

Scale is the other reason to add arms. A member too long or heavy for one gripper can be carried by two arms at once, one at each end, positioned in perfect coordination - the same way two people carry a ladder. Groups of arms have assembled large spatial timber structures, each robot placing and holding members while others fasten, building frames far bigger than any single arm could reach across. Researchers at ETH Zurich, the ICD and elsewhere have shown multi-arm cells weaving timber lattices and cooperatively erecting structures that would otherwise need heavy temporary works.

Dividing the labour also brings speed and parallelism. Two arms building different regions of the same assembly roughly halve the time, provided their zones do not clash. And cooperation enables genuinely new sequences: one arm can hold a partly-built sub-assembly in the air at the exact pose needed while another adds to it, so gravity and awkward geometry stop dictating the build order. The recurring pattern across all of it is that the second (and third, and fourth) arm turns a rigid, gravity-bound sequence into a flexible one - parts can be supported, rotated and presented at will. As always at this frontier, the structures produced are novel and their performance is validated by testing and signed off by engineers; the robots guarantee precise, coordinated placement, not that the finished frame is sound.

Everything good about multiple arms comes with one serious cost: the instant two powerful robots share a workspace, they can hit each other, or the growing structure, or a fixture - at speed, with force. Coordination is no longer a nicety; it is a safety-critical necessity. There are two broad ways to organise it. Time-sharing keeps the arms out of each other's overlapping space by scheduling - arm A works its zone while arm B waits or works elsewhere, and they take turns in the shared region, so they are never in the same place at the same instant. It is simple and safe but leaves arms idle. True cooperation has the arms move in the shared space simultaneously, their motions continuously planned together so trajectories interleave without collision - faster and more fluid, but far harder to compute and to guarantee safe.

This is why multi-robot work lives even more deeply in software than single-arm work. The whole cell - every arm's kinematics, the shared workspace, the build sequence, the collision envelopes - is modelled and simulated together in frameworks like COMPAS or robot-aware Grasshopper tools before anything moves, and often monitored in real time during the build. Add people to the mix and the safety problem multiplies again: a human-plus-multi-robot team needs zones, speed limits and sensing engineered from the outset. Coordination, not raw capability, is the ceiling on how many arms you can usefully point at one structure.

So far most of this happens in the controlled world of the fabrication shop, where arms are fixed, fenced and the environment is known. The frontier - and the reason it matters for architecture - is taking cooperative robots onto the building site. Fixed arms have limited reach, so the leading idea is mobile fabrication: arms mounted on wheeled or tracked bases, or on cable-suspended and legged platforms, that drive to the work, fabricate a region, and move on. A team of such mobile units could, in principle, cover a whole building the way a crew of trades does, each robot handling a task and coordinating with the rest - a vision sometimes described as swarm or distributed construction.

The honest status is early. Sites are unstructured, weather-exposed, cluttered and full of people - exactly the conditions robots handle worst - so on-site multi-robot fabrication is largely research and pilot projects, not routine practice. The realistic near term is human-robot teams: a few mobile arms doing the heavy, repetitive, precise placement and holding, while skilled people supply judgement, dexterity and adaptation, all under carefully engineered safety. What ties this final lesson back to the whole module is the thread you have followed: an arm can cut, it can place, it can print in space, and - with partners - it can hold, carry and cooperate. That versatility, coordinated by computation and married to human skill, is what makes robotic fabrication the open frontier of digital making.

If you take one thing from this module into practice, let it be a way of asking questions. Faced with a hard piece of making, do not ask which machine; ask what the tool must do - reach an angle, remove bulk, place a part precisely, draw a member in space, hold something while it is fixed - and let that decide whether an arm helps and in what role. The robot is not a single answer but a versatile pair of hands whose value depends entirely on the job. Knowing when it is the right hands, and when a laser, a router or a human simply beats it, is the judgement this whole course has been building.