On-Site Task Robots

Start with the pattern that holds across the whole family, because it is the thing the marketing hides. An on-site task robot is a single-task specialist. It does not build; it performs one repetitive operation - set a brick, tie a bar, drill a hole, mark a line, trowel a slab - very well and very tirelessly, and nothing else. It is, in effect, an industrial robot dragged out of the structured factory and pointed at a small, temporarily prepared zone of a messy site.

That transplant is exactly why the human loop never disappears. Think of the robot as the fast middle of a process that humans bracket on both sides. Before it can start, people prepare: level and clear the ground, set out the work, position and calibrate the machine, and load it. While it runs, people supervise: watch for the thing it cannot handle, correct drift, clear a jam, reset it when the situation changes. After its burst of repetitive work, people finish: the corners, the junctions, the awkward edges, the one-off conditions the robot was never built for. The robot owns the repetitive middle; humans own the variable ends.

This reframes what 'a robot does the job' actually means. A bricklaying robot may place the long, straight, repetitive runs of a wall quickly - and then a skilled mason still does every corner, every opening reveal, every service penetration and every tie-in to other work. Count the whole loop, not the robot's burst, and the labour saved is real but modest, and it is saved on the most repetitive, least-skilled part of the task.

Two consequences follow, and they run through the rest of this lesson. First, maturity varies enormously by task, because some jobs are naturally more repetitive and bounded (good for a machine) and some are irreducibly variable (bad for one). Second, the economic case is narrow and local. In a high-wage, labour-scarce market the saved hours may justify an expensive, finicky machine; in India, where site labour is abundant and comparatively cheap, the same machine has to win on speed, consistency, quality, safety or working where people cannot, rather than on simply replacing wages. Hold that lens as we look at each family in turn.

The most famous on-site task robot is the bricklaying or masonry robot. In the usual form a robotic arm, often mounted on a wheeled or tracked base or a small gantry, works from a digital layout: it picks a brick or block from a feed, applies mortar or adhesive (buttering the unit or laying a bed), and places it to a surveyed position, repeating the cycle. Vision and sensors help it find the work and check placement. On long, straight, repetitive courses it can be fast and consistent, and it spares human backs a punishing, repetitive lift. But it is fed by people, it needs the wall set out and the base positioned, and it hands off every corner, junction, opening and tie-in to a mason. Some systems are better understood as pace-setting aids working alongside a crew than as replacements for it. Maturity: real and piloted, including on actual projects, but still a specialist tool, not a standard site fixture.

Rebar-tying is a different and quietly promising case. Tying the thousands of wire knots at the intersections of a reinforcing mat or cage is repetitive, ergonomically brutal (endless stooping), and fairly bounded - which is exactly the profile a machine likes. Two forms exist: a handheld tying tool (barely a robot, but a real automation of the knot that is already widely used) and mobile tying robots that roll or crawl across a laid grid of bar and tie the crossing points they detect. The bar still has to be laid out and positioned by people, and the robot handles flat, regular mats far better than congested, three-dimensional cages. Maturity: emerging, with the handheld tool genuinely mainstream and the autonomous crawler earlier. Crucially, none of this touches the engineering that matters: the size, grade, spacing, lap and cover of the reinforcement, and whether the result is structurally adequate, remain the structural engineer's decision under the governing code - the robot only ties what it is told to tie.

Some of the most useful on-site robots do the least dramatic jobs. Layout and marking robots are a strong example and arguably the most quietly successful of the family. A small rover drives the finished slab and prints the building's setting-out directly from the BIM model - wall lines, opening positions, penetration points, fixing locations - full size, onto the floor. The task is a near-perfect fit for automation: a flat, structured surface, a purely digital source of truth, and a job (accurate setting-out) where machine precision beats a tape and chalk line and where human error is costly downstream. It removes drudgery, speeds the work and improves accuracy, and it is genuinely deploying on real sites. Notice the common thread with the factory: the robot thrives because the slab is, briefly, a structured environment.

Drilling robots - especially ceiling or overhead drilling rigs for fixings and services - attack a task that is slow, overhead (so hard on the body) and repetitive across a grid of points. A mobile base positions under each marked point and a drilling head bores to depth, working from the model's coordinates. Good fit, real products, still a specialist. Welding on site is harder. Robotic welding is utterly mature in the factory, but site steel is variable, positions are awkward, access is poor and conditions are uncontrolled, so on-site welding robots are a niche - used where there is enough repetitive, accessible welded work to justify bringing a controlled cell to the site, and otherwise left to skilled welders. The pattern repeats: the more the task can be made structured, repetitive and model-driven, the more a robot can take it; the more it stays variable, awkward and one-off, the more it stays with people. And across all of it, weld quality, fixing adequacy and structural connection remain matters for qualified engineers and inspection, never assumed from the fact that a machine did the motion.

Two more families round out the picture, and they sit at opposite ends of the readiness scale. Concrete finishing robots - powered screeding and power-trowelling machines that level and finish a freshly poured slab - are a comfortable fit. A slab is large, flat and repetitive; finishing it is tiring, time-sensitive work; and ride-on or remote trowelling machines are already common, with more autonomous versions emerging that drive a programmed path across the pour. This is one of the more mature on-site cases precisely because, for a little while, the wet slab behaves like a structured floor the machine can sweep.

Demolition robots are the other genuinely established family, and their driver is not productivity but safety. A compact, remote-controlled (teleoperated) machine carrying a breaker, crusher or shear goes into places that are dangerous for people - unstable structures, confined spaces, high heat, dust, or heights - and an operator runs it from a safe distance. Here the value is obvious and the case is made: the machine is expendable and the worker is out of harm's way. These are widely used, if more teleoperated tools than autonomous robots.

So what is the honest verdict on on-site task robots as a whole? They are real and advancing, useful in specific niches, and nowhere near a general replacement for site trades. The tasks that work best are the ones that can be made structured, repetitive and model-driven - layout, finishing, rebar-tying, overhead drilling - and the ones that succeed on safety grounds, like teleoperated demolition. The glamorous one, the bricklaying robot, is real but remains a piloted specialist that still needs a crew around it. Almost none work alone; almost all sit inside the human loop of prepare-feed-supervise-finish. And none of them changes the binding facts: the structural adequacy of what is built, the reinforcement, the welds and connections, the code compliance and the machine and site safety all remain with qualified engineers, inspection and the governing codes. The robot does the motion; people and the codes still own the building.