Construction needs to change: safety thinking with automation and robots

Automation and advanced technology are reshaping construction.

Yet to unlock the huge productivity shift they promise, we need to first rethink safety. It’s another example where prioritising resilience – the systemic ability to anticipate and adapt to known and unknown disruptions while maintaining core functionality – is crucial.

Safety in construction has always been its Achilles’ heel. Despite years of steady improvements, we’ve recently reached a plateau where injuries and fatalities are still worryingly and tragically commonplace.

In the UK, 2023-2024 saw 51 fatalities in this sector – an uptick on the five yearly average – while non-fatal injuries consistently number in the tens of thousands each year.

Delve a little deeper into those figures and one particular fact stands out – contact with earthmoving equipment is often a significant or leading cause of fatalities on construction sites.

So it might seem to be a relief that automation is revolutionising earthwork operations in construction sites as part of a wider digital transformation. Yet although this digital transition might feel inherently safer, caution is warranted.

Why have construction operational environments been so resistant to existing safety measures?

The answer lies in the highly dynamic, unstructured and transitory nature of these complex systems, filled with countless interactions of processes, people and machinery. Add in robotics and automated machinery, and the complexity magnifies rather than diminishes.

With considerable ambiguity in how these technological advancements can proactively address safety risks in such a dynamic environment without human cognitive oversight and intervention, we now have a new field of safety concerns.

So as seismic shifts reshape construction, working out how to shape the operational safety of automated systems has become a priority. To bring that to life, we’ll need new and urgent safety evaluation approaches able to provide a system of systems perspective and account for the interconnectedness, complexities and uncertainties that a construction operational environment might experience over its functional lifetime. That work has already started, and it’s called resilience engineering.

The old normal

Normative safety management has been the industry standard for some time. Safety I, as it’s now known, looks at what could go wrong during the operation and hence focuses heavily on rules, standards and quantitative risk assessment to prevent unwanted outcomes.

These are exemplified by the procedures developed by the Occupational Safety & Health Administration (OSHA) in the United States. But despite widespread adoption for years, these have had little impact on the number of fatal and non-fatal injuries.

We still need Safety I, its rules and its standards. But it’s simply not capable of tracking the fluid interdependencies and unpredictable threats endemic to complex systems. The shift to automated systems in construction only amplifies this – smart sites are intricate cyber-physical systems whose automation paradoxically introduces novel uncertainties as humans cede control and decision-making to software.

Fully grasping these “unknown unknowns” is impossible using classic risk modelling. For highly complex safety critical IT and human agent-based systems, a new safety paradigm is necessary.

Embracing resilience

Enter resilience engineering – or Safety II. Its proactive ethos views safety not as hazard avoidance but as a capacity to anticipate and adapt to disruptions while maintaining systemic functionality.

It accepts that safety is an ongoing, ever-changing response to sets of uncertainties arising from dynamic, interconnected and highly complex webs of possible known and unknown threats.

Whereas Safety I monitors compliance against predetermined performance, Safety II monitors resilience – evaluating real-time agility to flex and reconfigure in response to unanticipated shocks.

It thereby aims to futureproof operations against blind spots and chronic stresses, and critically, its emphasis is on guarding system productivity as well as anticipating and avoiding safety risk.

The resilience dividend

In the case of automated earthworks systems, the perception of safety risks is critical. It will require full sensor coverage, automated safety monitoring and object detection technologies to add cognitive ability to the system

Those capabilities will need to detect and localise both above-the-ground and under-the-ground objects to be effective in an appropriate time window. Though numerous studies are testing innovations that might do precisely this, there’s still a dearth of reliability.

It’s important we keep examining this issue. A single serious injury can cost over $1m, including stoppages, liabilities and investigations. Any utility strikes will not only damage equipment and harm workers, but they may trigger comprehensive, time-consuming and costly re-surveying.

Moreover, as we shift towards greater automation – a question of if rather than when – a lack of systemic agility will dilute the potential of transformation.

If automated systems prove neither reliable or safe then confidence will plummet and adoption will stall. Automated excavation is an invaluable long-term solution to meeting the demand for greater productivity in the construction industry and, in turn, is a foundation for how we’ll sustainably grow to meet the challenges of net zero and climate resilience. Those wider benefits are at risk if these issues aren’t thought through.

Machines at work

What’s needed right now are methods and techniques to evaluate the effectiveness of coordination strategies that optimise both safety and efficiency by scrutinising the influence of object detection technologies on the intricate numerous interactions within complex systems, including the interactions between separate technologies.

We need to conduct real-time testing of reactive capabilities to ascertain the degree of adaptability to manifold hazards. And developments in simulation and AI can help to create scalable, flexible models that might be able to tackle this huge task.

Through generating interactions between a wide range of site conditions, requirements, layouts and parameters, it will be possible to effectively simulate effects of individual actions on fleet-level coordination, evaluate reactive capacity to hazards and the operational efficiency of pre-determined plans and strategies, and also predict emergent patterns and behaviours that dynamically form and disform during operation.

That, in turn, will allow us to increase system-level awareness, determine weak nodes and failure points, and ultimately prevent unsafe situations.

Leading questions on the future of safety automation in construction

Will digitalisation and automation eventually lead to safer systems? We hope so, but it’s vital we continue to inquire to ensure a comprehensive answer to that question, and to understand the essential prerequisites for relying on automated equipment to independently manage safety risks in these constantly evolving environments.

This article was originally published in PBC Today.