Autonomous Haul Trucks and Dozers: What Actually Works on Site Right Now

A mining operation in Western Australia ran the numbers last quarter: their autonomous haul truck fleet moved 2.1 million tons with a utilization rate of 89 percent. The same work, done with operators, would have required 16 additional truck-and-driver rotations. No overtime. No fatigue penalties at shift change. No drivers calling in on a Friday afternoon.

This is not a prototype story. This is not a press release from a vendor booth. This is what autonomous and semi-autonomous heavy equipment looks like when it actually works on a production schedule.

The critical word here is "actually." Autonomous dozers, haul trucks, and loaders have cleared two major hurdles in the past 24 months: they no longer crash every third day, and fleet operators have stopped treating them as science experiments. The technology is mature enough that a mine or large construction firm can run a mixed fleet of autonomous and traditional equipment on the same site without losing their minds. But maturity does not mean simple. It means knowing what the equipment will and will not do, and planning operations around those limits.

The Current State: What's Rolling

Two platforms dominate commercial autonomous haul truck deployments as of mid-2026: retrofit kits bolted onto existing Caterpillar and Komatsu trucks, and purpose-built autonomous models from vendors like Autonomous Solutions Inc. and Belaz's autonomous variants. The retrofit approach is cheaper upfront. A mining operation can convert existing equipment rather than buying new iron. The purpose-built approach offers better integration and typically better sensor redundancy, which matters when you are running a truck unsupervised in a pit.

On the dozer side, Caterpillar's D11 autonomous and Komatsu's D475 equivalents are seeing the most traction in North American quarries and large earthwork projects. These are not toys. A D11 weighs 845 horsepower and moves 100 cubic yards per pass. When it is autonomous, it doesn't stop at 3 PM because the operator is tired.

The reality of autonomous deployment on active job sites: most operations run a hybrid model. A fleet of 12 haul trucks might include 8 autonomous units and 4 operator-driven trucks. The autonomous trucks run scheduled routes: pit to crusher, crusher to stockpile, repeat. The manned trucks handle the unpredictable work: spotting ore samples, adjusting load routes when equipment fails, moving material to secondary piles when the primary fill is full. This is not a replacement strategy. It is an augmentation strategy, and it works because it acknowledges what robots are good at (repetition and consistency) and what humans still do better (problem-solving in chaos).

The Math That Moves Decisions

A large mining operation processed the ROI on a 15-truck autonomous haul fleet conversion. Capital cost to retrofit and integrate: $4.2 million. Annual operational savings: $1.8 million (labor, fuel efficiency, reduced downtime from collision). Payback period: 2.3 years. On a 10-year asset horizon, that is not speculative; that is cash in the bank.

The savings break down like this: labor accounts for roughly 60 percent of the savings. A haul truck operator in a major mining region costs approximately $95,000 to $120,000 fully loaded (salary, benefits, safety, training, retention). An autonomous truck running 20 hours per day, 350 days per year, eliminates that labor line. Fuel efficiency improvements account for another 25 percent. Autonomous systems optimize acceleration, gear selection, and haul routes in a way human drivers cannot match consistently. The remaining 15 percent comes from reduced wear and tear on brakes, transmissions, and tires because the AI does not make aggressive gear changes or hit the brakes hard at the grade transition.

But there is a hard cost that vendors will not advertise: integration and software licensing. Bolting a retrofit kit onto a truck is the easy part. Connecting that truck to your dispatch system, your mine management software, your safety monitoring platform, and your fleet telematics requires a capable systems integrator and 4 to 8 months of implementation. Budget $800,000 to $1.2 million for that work on a meaningful fleet.

A construction firm running autonomous dozers on a 40-acre grade job measured different metrics: dozers operated 16 hours per day (human operators tap out after 10), achieved grade tolerances within 2 centimeters (human operators typical at 4 to 6 centimeters), and completed the job 9 days ahead of schedule. The labor hours per cubic yard of cut dropped from 0.31 to 0.18. That is a 42 percent efficiency gain on a single performance metric.

Where the Tech Breaks Down

Autonomous equipment fails hardest when the job does not match the expected conditions. A haul truck programmed for a specific pit geometry cannot adapt when a slope fails and closes the primary route. It does not reroute itself intelligently. It stops and waits for a human dispatcher to reprogram it. A dozer with autonomous blade control can maintain grade, but if a water truck has just wet the haul road and the soil cohesion changes, the dozer's load estimations become unreliable. It drops loads and makes extra passes.

Sensor degradation is real. GPS-based positioning works until dust storms roll in. LiDAR performance drops in heavy rain. Cameras cannot see through fog. On a site where weather is unpredictable, autonomous equipment still needs a human backup. The truck does not just stop; it goes into a safe mode where it idles and waits. Every hour of idle time is revenue lost.

Equipment interference is another hard lesson. When you run a mixed fleet of autonomous and manned equipment in a tight haul road, collisions happen. Not at the frequency they did in 2023 and 2024, but they happen. Autonomous systems are conservative (they stop rather than risk a collision), but conservatism on a haul road means bottlenecks. A single autonomous truck stopping to ensure it does not hit a human-driven loader can cascade into a 30-minute delay across the entire haul fleet. Sites with tight geometry (narrow pits, constrained haul roads) have struggled with deployment. Sites with open geometry (large quarries, wide haul networks) see better utilization.

The Safety Question That Still Matters

Autonomous equipment has eliminated a specific class of accident: operator fatigue incidents and loss-of-control events. No autonomous haul truck has ever fallen asleep at the wheel. No autonomous dozer has misjudged a slope angle at 4 PM on a Friday. Those are real wins, and they are documented.

But autonomy introduces new failure modes. Sensor failures can cause equipment to behave unpredictably. Software bugs in path-planning algorithms have caused trucks to make wrong turns in low-visibility conditions. Cybersecurity is a growing concern; one firm reported an attempted breach of their haul truck dispatch system in early 2026. No data was exfiltrated, but the incident proved that autonomous fleets are networked targets.

Sites with strong safety cultures report fewer issues with autonomous equipment integration. Not because the autonomous trucks are safer (they are not inherently safer, just differently safe), but because those organizations treat autonomous equipment deployment like a major safety implementation. They do pre-deployment testing. They train operators and dispatchers on failure modes. They establish clear procedures for when autonomous equipment must be taken offline. Sites that treat autonomous deployment like a straightforward equipment upgrade tend to have more problems.

Implementation: The Right Way and the Regrettable Way

Firms deploying autonomous haul trucks or dozers successfully share common practices. They start small. A single autonomous truck in a mixed fleet, or a single autonomous dozer on a single push, rather than trying to convert an entire operation at once. They hire or develop in-house expertise in autonomous systems management; this is not something a dispatcher with 20 years of loader experience can just absorb. They establish redundancy. Every autonomous vehicle has a mechanical backup route and a human-override capability. They monitor performance metrics obsessively during the first 90 days of operation. Utilization, downtime, fuel consumption, grade accuracy, cycle times. If a metric deviates by more than 8 to 10 percent from expected, they investigate before moving forward.

Firms that have regretted their deployments typically made three mistakes: they bought the autonomous equipment before finalizing their integration plan, they expected the technology to work immediately without a learning period, or they cut corners on staff training and safety procedures to hit an aggressive deployment deadline.

One large construction company deployed 8 autonomous dozers across three simultaneous projects in late 2024. Two projects went well. On the third project, poor site coordination and inadequate communication between the autonomous fleet manager and the project superintendent led to multiple equipment conflicts and schedule delays. The dozers themselves performed as designed. The problem was organizational and procedural, not technical. That lesson is worth underlining: autonomous equipment is not a plug-and-play solution. It requires competent oversight and clear operational procedures.

What Comes Next

The trajectory is clear. By 2028, autonomous haul trucks will represent approximately 12 to 15 percent of haul capacity in large mining operations in developed economies. Autonomous dozers will be commonplace on any major grade or cut project where schedule is a primary constraint. The technology is not going to get dramatically better in the next two years; it is going to get more reliable and more integrated into existing fleet management systems.

What operators should be thinking about now: Do your operations have the infrastructure (GPS coverage, quality dispatch software, robust telematics) to support autonomous equipment? Are you willing to invest 6 to 9 months in implementation before you see the first dollar of savings? Do you have personnel who can manage and troubleshoot autonomous systems? If the answer to any of these is no, your timeline for deployment should be 2028 or later, not 2026. There is no penalty for waiting. The technology will be cheaper and easier to integrate in 24 months.

For operations that have the infrastructure and the expertise, autonomous heavy equipment is no longer a bet on the future. It is a decision about competitive position in the present. The mines and construction firms running autonomous fleets today have a measurable cost advantage, and that advantage will widen as the technology matures.