What techies keep getting wrong about industrial automation.

Futurama must have been amazing. No - not the TV series, but the model city it is named after, built for the 1939 New York’s World’s Fair by Norman Bel Geddes. It showed a bold vision for the future: thousands of tiny cars moving on wide highways - all without the need for a human driver. This future wasn’t too far away either - 1960 was the year envisioned for full self-driving cars by its creator.

77 years later, in 2016 Tesla announced “Full Self-Driving Hardware on All Teslas”. This was exciting. Chilling in the back of your car, watching a movie while being chauffeured to your destination sounded like a dream.

Today, in 2024, full self-driving is still mostly a dream. Outside of a few pilot projects in restricted areas, it is no more available than it was in 1939. And there are good reasons for that. For one, traffic is much more messy, chaotic, and unpredictable than it looks on the surface. And then, government regulators and legislators move at a snail’s pace when compared to technologists.

But there are some areas where neither problem exists. Areas, where traffic is highly structured and controlled. Where private companies are in charge that can adopt new technologies much quicker than governments. And where efficiency is paramount, meaning that every bit of machine autonomy could be a huge money saver.

Mines are such an area. Oil & Gas, large farms and facilities such as airports are another.

And for a while, it looked like these sites would become the very hotbeds of autonomous automation that technologists expected them to be. “Smart Mining” and the “Digital Oilfield” were all the rage. But today, after a lot of research, expensive technology purchases and innovation initiatives, operations still look largely the same as they did twenty years ago - with some marginal bits of automation around the edges.

Why is that? What is it that technologists keep getting wrong about autonomous automation? And how can we chart a realistic path forward? Let’s start with the obstacles that prevent large scale automation:

Human-Centered Design

Industrial sites are designed around humans in fundamental ways. And while one can add some degree of automation to these human centric designs, doing so leaves most of the potential efficiency gains on the table.

Now, that sounds very lofty and abstract, so here’s a concrete example: Take the iconic, giant haul trucks that operate on surface mines. The Caterpillar 797F is such a truck. It can transport a payload of 363 tons and costs around five million - which comes out to around 14k per ton of payload.

A regular 20t dump truck, on the other hand, costs around 100-150k which comes out to 5 - 7.5k per ton of payload. Maintenance costs are also much lower, as they run on standard truck tires and don’t require special equipment to repair.

So why do mines use a few giant trucks rather than a lot of small ones, even though they cost 2-3 times more per ton of payload? In a word: Labour costs. A giant truck is still driven by only one driver.

So, while various companies have set out to automate the giant dump trucks, having a continuous chain of smaller, self-driving trucks using standard parts and without the need for a driver cabin would be much more efficient. But, it would also require a shift from the human centric design choice towards a machine centric approach.

The same is true for many other aspects of mining. Shafts for autonomous vehicles can be much narrower since there’s no need for escape routes or ventilation. Truly automated subsurface operations can be performed with many small vehicles in winding tunnels that can approach the 3D underground structure of a seam, rather than the large, flat areas humans need to operate comfortably. Surface mining could run on an automated schedule with continued regional blasting and swarm-like vehicles automatically staying clear of the blast-area, rather than the large, safety related interruptions that come with humans handling explosives.

So, why not introduce at least some of these automation steps?

Some automation is often worse than no automation

Automating a subsection of an industrial process is often worse than not automating at all. Having a set of autonomous machines as part of a value chain often proves inflexible and complex, especially if things go off-script.

This is particularly true for more complex operations, such as just in time aggregate processing or on-site refining for Oil. Having a subset of the value chain automated frequently creates so much friction at its interfacing points with human workers that it offsets the efficiency gains.

Instead, automation needs to be applied to the whole production process at once - which is costly and creates interruptions.

Market Forces

But aren’t there strong market forces that drive automation and that pave the road towards autonomous operations?

Well…sort off - but, mostly, no.

Equipment Manufacturers Mining Equipment is extremely expensive and specialized. R&D cycles are long - and a proven track record of reliability is worth more to operators than having the latest in new technology. This has led to a high level of consolidation among equipment manufacturers - with only a few, giant companies now providing the bulk of supplies.

There is, however, a lot of competition between these manufacturers - but only for the initial purchase decision. Mining Equipment Manufacturers have mastered the art of vendor lock in and closed ecosystems to a degree that makes Apple look like an Open Source company. Their vehicles, hardware, software and protocols often form one impenetrable network of proprietary standards and locked down access.

As a result, buyers usually choose the manufacturer’s stack early on and are then locked in - with most revenue generated through upselling.

This has led to a complicated set of incentives for manufacturers. Yes, they need to innovate - but only as far as they can ensure that new products and features strengthen the network effects within their ecosystem and don’t tarnish the reputation of reliability.

This means, that anything too disruptive, such as fully autonomous systems come with a significant risk to the ecosystem at large.

Interoperability between contractor equipment These close ecosystems, however, present a constant headache for mine operators. Many modern mines are organized as fairly small operating companies with a revolving door of specialized contractors. Each contractor brings their own staff, equipment and software - all of which have to work together to some degree or another. Given the lack of open standards, this is a major obstacle to any more complex integration, such as truly integrated, automated systems would require.

Of course, there’s also still the government:

Government Legislation and Insurance Liability

Even though government legislation is not as directly involved with the day to day operations of an Oil Field or Mining Site as it is in urban traffic, there is absolutely no shortage of regulations, safety standards and environmental rules that need to be followed. This creates both a risk and an opportunity for automation. Sensors and Drones can address large swaths of the ever growing reporting requirements and not having workers on site significantly reduces safety risks.

But what happens when things go wrong? Legislation and liability around accidents involving autonomous vehicles are largely non-existent. If an autonomous truck runs over a worker - is the manufacturer to blame? The operating company? How do you price the insurance risk for this scenario?

It is this sort of legal murkiness that deters a lot of operators from adopting autonomous technologies.

And then, there are of course the societal aspects of autonomous machines:

Cultural Resistance and Tradition

“This is not how we do things here!” is a view you might expect to hear in the leather filled workshop of a cobbler or the dust covered studio of a stone mason - but not in the intrinsically high tech operation of an Oil Rig or a Mining Site. Yet habits and traditions are surprisingly strong in these industries. And trying to transition a highly skilled team of rugged heavy machinery operators into a much smaller squad of desk-based computer specialists is not going to make you popular.

And, of course, replacing the camaraderie felt with colleagues with sterile machine interactions will make for a demotivating workplace - to say the least.

And, of course, the unions will have something to say about this too:

Unions and Workforce Impact

Mining was among the very first industries to adopt organized labor practices. The United Mine Workers of America were founded in 1890, making it one of the earliest unions in history and unionization has remained very strong in Mining until the present day.
And there’s no sugarcoating it - automation makes human workers obsolete in the short term. Yes, job profiles change and workforce skillsets adapt over time, but in the short term there is a significant decline in workforce requirements as higher degrees of automation are introduced. Unions are fighting this tooth and nail, creating a significant barrier to the adoption of autonomous technologies.

The path forward

I believe we’re at a crossroads in world history. Growing population sizes, rise of living standards in China and India and the increasing electrification of everything have lead to a sharp rise in demand for commodities across the spectrum - from oil & gas to metals, fertilizer components and rare earths. Labour costs are at an all time high in the western world. Yet a trend towards de-globalisation makes buying cheap commodities from developing countries increasingly infeasible.

Automation and autonomy can be the solution to all of this. And there are two ways to get there:

Revolution: There might be a key event that reshapes industries. The introduction of highly efficient, fully autonomous industrial sites in countries with strong development incentives, such as China might function as a wakeup call for operators around the globe. I believe this might lead to a hard cut and a fundamental redesign of industrial processes with a focus on efficiency through automation and autonomy.

Evolution: More likely though is an incremental adoption of ever higher degrees of automation on industrial sites. This is met with all the challenges I’ve outlined above. But they can be overcome.

Hivekit - (the product this blog belongs to) - is trying to facilitate this. It combines Asset Tracking, Operations Management and Process Automation into one unified platform, providing insight, control, efficiency and safety across industries.

This enables human operators not just to control operations manually, but also to introduce automation in an incremental and mindful manner, designed to seamlessly integrate human and machine processes and ensure continuous productivity.

But whichever approach you end up taking - getting to automation might seem like a long way - until it isn’t. So it’s important to embark on this journey early on in a systematic and mindful way.