Site Setup & Integration

A construction robot or a building-scale printer is, first of all, a large and heavy piece of plant, and before it can do anything clever it has to arrive and be assembled. A gantry printer the size of a small house ships in pieces - rails, frame, carriage, nozzle, control cabinet, pumps - on one or more trucks, and is craned off and bolted together on site over hours or days. A printer mounted on a robotic arm may be more compact, but it still needs transport, a stable mounting, and often a track or base to give it reach. Mobile and crawler machines drive or are carried in, but they are still weighty, awkward loads that need the right access.

This stage alone defeats more demonstrations than any software bug. Can a truck of that size reach the plot down a narrow Indian gully or a congested urban lane? Is there crane access, and hardstanding for the crane to stand on? Is there room to lay out and assemble a gantry that may be several metres longer than the wall it prints, because the machine needs to overrun the work at each end? Assembly itself is skilled work: rails must be set straight and level, the frame squared, the carriage trammed, cables and hoses routed and protected, the control system booted and its axes homed. A gantry that is a few millimetres out of level over its length will print a wall that is out of plumb, because the machine faithfully reproduces its own errors.

Then there is the base the machine stands on and prints onto. Printers generally need a flat, level, load-bearing surface - very often a conventional foundation or ground slab cast in advance by conventional means - both to carry the machine and to receive the first layer of the print. That slab is itself a conventional construction task with its own programme, curing time and tolerance, and it has to be finished and strong enough before the machine arrives. So the first honest lesson of deployment is that the automated part is wrapped in a thick layer of ordinary, heavy, logistics-driven construction work - and every bit of it must go right before a single bead is laid.

A machine works in its own coordinate system: everything it does is relative to its own origin, the point it treats as (0, 0, 0). The digital model knows where the wall sits relative to that origin. What neither the machine nor the model knows, until a human tells them, is where that origin actually is on the real plot - its position, its rotation, and its level relative to the building's survey datum and gridlines. Reconciling the two is called setting-out, and it is the single most important act of on-site calibration.

Conventional construction does this all the time: surveyors establish control points and a datum, then set out gridlines and levels so every trade builds to the same truth. An automated machine has to be tied into exactly that framework. The operator establishes where the machine origin is relative to the survey control, squares the machine's axes to the building grid, and checks level and plumb so that the model's 'up' really is up and its 'north' really points along the wall. A printed footprint that is rotated two degrees, or shifted five centimetres, or built on a base that slopes, is wrong in ways that ripple through every opening, every service penetration and every junction with conventional work - and a machine will reproduce that error perfectly, all day, without noticing.

Calibration goes beyond position. The machine's own geometry is checked and trued: the gantry levelled, the arm's reach and joints verified, the nozzle height above the base set, the extrusion and motion tuned to the actual material on the actual day. Many teams run a dry pass - tracing the toolpath with no material - to confirm the machine can reach the whole job without collisions and stays inside its working envelope. Sensors, where fitted, are referenced to the same datum. The discipline here is ordinary surveying and metrology wedded to an unforgiving machine: the robot removes human variability from the building of the wall, but it cannot remove the human judgement about where the wall belongs. That judgement, and the survey control behind it, stay firmly with qualified people.

Here is the reality the headlines hide: almost no project is built start to finish by one machine. The automated step - printing the walls, or a robot laying a course, or an arm placing components - is one operation inside a construction sequence that is overwhelmingly conventional and overwhelmingly human. A printed 'house' is, as this course keeps insisting, printed walls sitting on a conventional foundation, with conventional reinforcement strategy, floors, roof, windows, doors, services and finishes. Every one of those is a separate trade with its own timing, and the automated step has to slot in among them without breaking the flow.

That integration is a scheduling and coordination problem at least as hard as the technology. The foundation and base slab must be ready before the printer arrives. Any reinforcement, starter bars, cast-in fixings, conduits or service routes that must be placed before or during the print have to be planned into the process, because you cannot easily drill or chase a finished printed wall the way you would a block wall - penetrations are far better designed in. The machine needs its window on site: space, access and time when other trades are clear of its zone. And when the automated step finishes, the work hands back to conventional crews who build on top of, into and around what the machine made - which only works if the printed element meets the rest within the tolerances those trades expect (the subject of the next lesson).

So the designer and the site manager are really choreographing a relay. The machine is a fast, precise runner for one leg of the race, but it needs the baton handed to it cleanly and must hand it on cleanly in turn. In practice this often means the biggest savings the automation promises are eaten back by the coordination it demands, at least on early projects, and it means the automated step is only as good as the conventional work bracketing it. Treating the robot or printer as a drop-in replacement for a whole building is the classic category error; treating it as one specialised subcontractor to be sequenced with care is the competent stance.

Strip away the novelty and on-site automation is dominated by the same unglamorous logistics that dominate all construction - only with an expensive, inflexible machine in the middle that punishes every gap. Mobilisation (getting everything to site and ready) and demobilisation (breaking it down and taking it away) are real line items of cost and time, often charged whether or not the print goes well, which is part of why the economics only work above a certain size of job: you have to amortise all that setup over enough output to justify it.

Then there is everyone and everything the machine depends on. A trained operator and often a specialist crew from the manufacturer or a licensed partner have to be present - on early Indian and global projects the pool of people who can actually run these machines is small, and their travel and time are a genuine constraint. A reliable material supply chain has to be arranged: the special printable mortar or concrete is not ordinary ready-mix, and it may have to be batched on site or delivered in a narrow time window and used before it sets. Power, water, storage for materials, a clean and level working base, protection from sun and rain, and a plan for what happens when - not if - something interrupts the run: all of this is logistics, and all of it has to be solved before the clever part can even begin.

The honest summary is that deployment is where many demonstrations quietly differ from real projects. A demo happens in a controlled yard, with pre-batched material, clean power, flat ground, ideal weather and the maker's own experts hovering. A real site is messy, variable, weather-exposed, access-constrained and staffed by people meeting the machine for the first time. Closing that gap is not a matter of better software; it is a matter of planning the transport, the base, the setting-out, the supply, the crew and the sequence with the same seriousness you would give to commissioning any piece of heavy plant. Lesson 7.4 returns to these real-site constraints in depth. For now, carry this: on site, the machine is the easy part - it is everything around the machine that decides whether automation actually happens.