Automation and robotics offer tremendous potential for productivity gains. Automotive assembly, warehousing, and agritech are all benefiting from advanced automation, but beyond some early forays, automation and robotics in construction applications are noticeably scarce.

The first big robot success story was automation of repetitive, high-precision tasks in factories, where the environment is tightly constrained, problems can be quickly detected by continuous line QA processes, and the cost of errors is limited to wasted materials. Agricultural applications are different again: the environment is much less controlled, and the primary benefit of automation is to take on physically difficult tasks, rather than repeatability.

Construction sites are a different again: they are highly variable (even more so than a typical agricultural application), dirty, dusty, noisy and constantly changing. Construction materials are highly varied and inconsistent, and yet many tasks still require a high level of precision.

Below, we examine some of the key drivers and characteristics of successful construction automation systems.

“I know it when I see it”

Automation is only valuable if it reduces project costs or timescales, improves worker safety, or provides additional assurance of quality. Skilled tradespeople are constantly assessing the quality of their work, and adjusting to fix errors or work around obstacles. It’s easy to identify a poor paint job after the fact, but to be as good as a human an automated painting system must detect and fix problems while performing the task. Costs of fixing QA issues become greater the later they are discovered - once additional work has been built on a badly-done task output, reworking it becomes more difficult.

For example, doing a good paint job on a wall that’s not perfectly flat, or installing furnishings on a wall that’s not perfectly square, are highly interactive and skilled tasks that are difficult to codify. A human will always be the final arbiter of quality, and it is likely that an automated jobsite will require more intensive and frequent oversight than one staffed by human laborers.

General purpose or specialized?

Robots that augment human workers will generally be more successful than those that try to replace them

Humans are highly flexible general-purpose machines. We can turn our minds and bodies to a huge variety of tasks. When designing a robotic system, there’s a tension between the desire to make it flexible and widely-applicable (to target a wider market), and the costs of implementing that flexibility. For example, a robotic end effector with the full range of motion and sensing capability of a human hand is still unattainable.

So it’s inevitable that real-life robots end up being relatively specialized. The challenge for a system designer is finding the right niche and the appropriate level of flexibility to end up with a commercially viable product. Robots that augment human workers will generally be more successful than those that try to replace them - both in terms of value proposition, and how readily they are accepted by workers.

Automation bias and oversight

Overseeing an automated system that performs a complex task is difficult and mentally taxing. Paradoxically, this gets worse as the automated system performs better - maintaining concentration and effective monitoring is difficult when intervention is rarely needed. Automation bias - the tendency for human overseers to pay less attention to automated systems that normally work well - has been most heavily studied in life-safety applications such as aviation (for example Air France 447).

The consequences of inadequate oversight on a construction site are, of course, much less severe than in aviation or self-driving cars. But the economic costs of errors directly impact the value proposition of the automation solution, so it’s important for manufacturers and prospective customers to understand the cognitive biases that can impact it. Overseers will typically be senior tradespeople who have enough experience to spot errors as early as possible, but some retraining will inevitably be needed because the errors that machines make are of different types to those made by human workers. Being able to do the task well will remain necessary for overseers, but won’t be sufficient - additional training methodologies will be required to develop a workforce that can effectively supervise an automated jobsite.

Jobsite Variability

Even though many construction tasks (cutting, drilling, fastening, sealing, painting, plastering etc) can be boiled down to a set of consistent actions, the jobsite itself can be highly variable. Tools must either be hand-portable or otherwise easy to transport around a jobsite, which may typically include tight passageways, uneven surfaces, changes of level and other hazards. A physically large robot may need special consideration in planning the layout of the jobsite or ordering of tasks.

Horizontal Integration

By the time a skilled tradesperson rolls up and practices their craft, many other actions and decisions have occurred. Architectural plans are drawn, materials are specified, plans are drawn up and sequenced, to name just a few. Further out, suppliers have designed products, packaging and supply-chain logistics to suit the status quo. The existence of robotic construction systems will affect every part of this pipeline, introducing both opportunities and challenges.

A successful system will benefit from horizontally-integrated development where materials, packaging and movement are optimized for robotic manipulation and its unique needs. For example, as robots are able to more precisely place materials, prefabricated parts may become economically viable in more places, and more broadly design rules may change to take advantage of decreased tolerances from machine placement.

On the other hand, it may be worthwhile to constrain aspects of the design to allow automation to be used, for example by including passageways so that large robots can access internal features. We can draw parallels to the discipline of design for manufacturing (which imposes rules and considerations on the design of mechanical parts to accommodate automated production lines), and developments in agricultural technology (redesign of orchard layouts and even modification of plants themselves to optimize for robotic harvesting).

There are opportunities here as well - robotic tools can be used to reconfigure and disassemble as well as build. So designing to provide access for robotic construction may also make it easier for buildings to be repurposed or recycled well into the future.


The case for robotics and automation in construction is sound; the need for productivity gains, horizontal integration of construction processes, the pragmatism of modularization and the growing challenge of finding skilled labor collectively point to increased use of automation. We believe that successful innovations will have the following characteristics:

  1. They will augment the capabilities of skilled craftspeople rather than replacing them
  2. Be adaptable and capable of operating in varied environments
  3. Address tasks which are dangerous, tedious or physically demanding
  4. Co-evolve with changes to building design methodology, including greater horizontal integration and modularity of building components and materials

For more information please contact the authors, Saajan Chana and Bruce Ackman. Or speak to Saajan in person at the BuiltWords Machines Conference in Chicago June 6. 

Saajan Chana
Principal Embedded Systems Engineer

Saajan is a systems engineer with a background in embedded software and a wide experience of multi-disciplinary sensor development, ranging from low-cost consumer products to high-value industrial sensors.