Three years ago, the narrative around autonomous haul trucks was almost entirely optimistic. The technology was proven at a handful of Pilbara iron ore sites, costs were projected to drop, and the industry consensus was that human-operated haulage would be a relic within a decade.

The reality in 2026 is more complicated than that. Autonomous haulage has delivered genuine results, but it’s also exposed problems that nobody talked about in 2023. Here’s what we’ve actually learned.

The Pilbara Success Story Is Real - With Caveats

Let’s start with what’s working. In the Pilbara, autonomous haulage has been a legitimate success. Rio Tinto now operates more than 200 autonomous Komatsu 930E trucks across its iron ore network. BHP’s South Flank operation runs a mixed fleet of Cat 794AC autonomous trucks alongside traditional operations. Fortescue has expanded its autonomous fleet at Christmas Creek and Cloudbreak.

The numbers back it up. Rio Tinto’s autonomous trucks have logged over 100 million kilometres with zero lost-time injuries from truck-related incidents. Productivity gains of 15-20% compared to manned operations are well documented, primarily because autonomous trucks don’t take crib breaks, don’t slow down at shift change, and maintain consistent speed on haul roads.

But these are large-scale, open-pit iron ore operations with relatively simple pit geometries, predictable weather, and the budget to invest hundreds of millions in supporting infrastructure. The Pilbara is essentially the best-case scenario for autonomous haulage.

Underground and Multi-Fleet: Still Struggling

Take the technology underground or into more complex surface operations, and the picture changes fast.

Underground autonomous haulage has made progress, but connectivity remains a persistent headache. GPS doesn’t work underground, so vehicles rely on a combination of LiDAR, radar, and network-based positioning. Signal degradation in wet conditions, dust interference with sensors, and the challenge of mapping constantly changing headings all create reliability issues that don’t exist on the surface.

Multi-fleet interoperability is another sore point. Most mine sites operate equipment from multiple OEMs - Komatsu loaders feeding Cat trucks, for example. Each manufacturer’s autonomous system uses proprietary protocols and management software. Getting a Komatsu autonomous truck to coordinate efficiently with a Liebherr excavator equipped by a different autonomy vendor is still an exercise in workarounds rather than genuine integration.

The International Council on Mining and Metals has been pushing for interoperability standards, but adoption is slow. OEMs have limited commercial incentive to make their systems play nicely with competitors.

The Maintenance Surprise

Here’s what caught most operators off guard: maintenance costs didn’t drop the way everyone expected.

The trucks themselves are more reliable in terms of collision avoidance and reduced operator error. But the sensor arrays, communication systems, and computing hardware that make autonomy possible have introduced new failure modes. LiDAR units get sandblasted in the Pilbara. Radar modules fail in extreme heat cycles. Edge computing hardware needs replacing on schedules that don’t align with traditional component change-out intervals.

According to analysis from the CSIRO’s Data61 group, predictive maintenance systems have helped offset some of these costs. Operations using AI-driven maintenance scheduling for autonomous fleets have reduced unplanned downtime by roughly 25% compared to time-based maintenance programs. But the total cost of keeping the autonomy stack running is higher than original projections suggested.

Some operators are now working with AI consultants Brisbane to build custom fleet management dashboards that integrate autonomy data with traditional maintenance planning systems. The goal is a single view of fleet health that captures both mechanical condition and sensor system status - something that most off-the-shelf solutions still don’t do well.

The Operator Retraining Question

This is the part of the story that deserves more attention. Autonomous trucks don’t eliminate operators - they change what operators do.

At most autonomous sites, former haul truck drivers have transitioned into remote monitoring roles, sitting in air-conditioned control rooms watching multiple trucks on screens. Others have moved into maintenance, field support, or traffic management positions. The FIFO workforce hasn’t shrunk as dramatically as unions feared, but the skills required have fundamentally shifted.

The challenge is that retraining a haul truck operator to troubleshoot sensor calibration issues or interpret autonomy system fault codes isn’t a two-week course. Sites that invested early in structured retraining programs have smoother operations. Sites that treated it as an afterthought are dealing with higher turnover and gaps in technical capability.

So Where Does This Leave Us?

Autonomous haulage works. In the right conditions, it works very well. But the gap between Pilbara iron ore and the broader mining industry is wider than vendors initially acknowledged.

The next three years will be defined by how well the technology adapts to messier environments - underground coal, mid-tier gold operations, and multi-commodity sites where ore grade variability makes route planning more complex.

The technology is not the bottleneck anymore. Integration, workforce transition, and honest cost accounting are. Operators who treat autonomous haulage as a systems problem rather than a truck problem will get the most out of it. Those waiting for a plug-and-play solution will be waiting a long time.