AI boosts mining trucks efficiency with 55% shorter queues

[Redator]

Artificial intelligence can slash truck waiting times in open-pit mining by as much as 55%, while increasing productivity and fuel efficiency, according to a new study by GEM Mining Consulting.

The research tested Reinforcement Learning (RL) as an alternative to conventional rule-based truck dispatch. Across multiple scenarios, the AI system delivered stronger results: 23–33% more tonnage moved, 19–24% faster cycle times, 6–21% better fuel efficiency, and up to 48% less downtime.

Truck allocation remains one of mining’s toughest challenges, with loading and hauling making up 50%–60% of costs. Traditional dispatch systems depend on fixed rules that often break down when conditions shift, such as equipment failures or unscheduled maintenance. As operations expand, delays and inefficiencies compound, reducing output.

To test RL’s potential, GEM built a discrete event simulation in Python, modelling an eight-hour shift with 30 trucks, four shovels, multiple pits, and destinations including plants and waste dumps. The simulation incorporated real-world complications such as refuelling, maintenance, queues, and random breakdowns.

Real-time tweaks

Researchers then applied Proximal Policy Optimization (PPO), an RL technique that learns from experience and adjusts in real time. Unlike static heuristics, the AI continuously evaluated variables such as truck locations, queue lengths, and fuel use to maximize efficiency.

In repeated trials, RL outperformed conventional dispatch. Even with equipment failures, the AI kept deliveries steady with shorter queues and faster cycles, while the rule-based system faltered, widening the productivity gap.

“The study demonstrates that Reinforcement Learning not only increases productivity but also significantly improves operational efficiency and environmental sustainability,” Sebastián Faúndez, Head of Analytics at GEM Mining Consulting, said.

“In a sector where uncertainty and variability are inevitable, AI offers a robust and adaptable alternative that outperforms traditional truck dispatching approaches.”