Manually managing hundreds of mining haul trucks by radio is getting harder: the number of events per shift has long exceeded what a person can handle. This case shows how an open-pit mine moved from manual dispatching to AI-driven traffic balancing without replacing its existing monitoring system, and achieved over RUB 300 million in confirmed impact in a single year.
The Starting Point at the Open-Pit Mine
The mine operates a fleet of 247+ haul trucks and over 25 excavators. Dispatchers balanced truck traffic manually over the radio. This approach ran into four key problems.
The Dispatcher’s Cognitive Limit
More than 10,000 deviations from the norm occur per shift. A dispatcher can physically process less than 0.1% of them. Meanwhile, conventional reporting captured only 10–20% of actual losses.
Unbalanced Truck Flow
Excavators alternated between standing idle with no trucks waiting and working with long truck queues. On average, a missing truck at an excavator meant one lost trip per shift. The dispatcher learned about a problem 10–15 minutes after it arose, when losses had already built up.
Losses at Shift Changes
Early departures and late returns accounted for up to 25%: vehicles left the line 30–40 minutes before the end of a shift and came back late.
Inefficient Refueling
Each refueling added an average of 15 minutes of unplanned downtime, since nobody controlled how trucks were spaced out in time.
Project Goal
Reduce excavator downtime and increase rock haulage with the existing fleet, without replacing the current monitoring system.
The Solution: AI Modules on Top of Existing Infrastructure
Interconnected AI modules were deployed at the mine on top of the existing fleet management system (FMS). Technically, the system can be deployed in 2 weeks.
Haul Truck Balancing Module
The module is built on more than 40 different AI models. The system:
- forecasts queues 30 minutes ahead;
- manages refueling breaks, spacing trucks out so that no more than one truck leaves the same unit at once;
- instantly reallocates the entire fleet when breakdowns occur.
Directives go straight to a tablet in the driver’s cab, and 95% of them are carried out without a call to the dispatcher. The system works without dispatcher involvement and adapts to changes in real time.
Multi-Agent Control System
This is a swarm of 30+ specialized AI agents. Each one oversees its own area: standard refueling time, loading cycles, shift changes, and other processes.
Every deviation goes through the cycle “event → warning → violation → directive” with three-level escalation: driver → dispatcher → shift supervisor. All deviations are quantified in rubles in real time.
Telematics as the Foundation for AI Algorithms
Every haul truck is fitted with a tracker that collects telematics data every 1–5 seconds: GPS, fuel level, and engine parameters. Without accurate real-time data, the neural-network ETA forecast and the balancer could not reliably predict queues.
Equipment used:
- GPS tracking devices;
- driver tablets;
- fuel level and pressure sensors;
- IIoT software.
AI Balancing Rollout Stages
- Prototype. Testing the balancer on one unit of 3–5 excavators.
- Industrial operation. Scaling to the entire fleet of 220-ton trucks and excavators, moving to an in-house interface, manually checking recommendations, and fixing bugs. Impact: RUB 12 million per month.
- Algorithm tuning. Experiments with balancing approaches, improved ETA forecast quality, and the first manual reports. Impact: RUB 25 million per month.
- Scaling. Expansion to 130-ton trucks, setting up directives on tablets, and experiments with automatic acceptance of recommendations. Impact: RUB 45 million per month.
- Quality improvement. Synchronization with equipment allocation, drivers working with directives for leaving the line, and a shift from manual refueling redirection to automated dispatching. Impact: RUB 80 million per month.
Project Results
The system is evaluated using two groups of metrics: operational (productivity, downtime, speed) and economic (rubles in real time).
Operational Metrics for 2019–2025
- +10% rock haulage with the same fleet and no additional trucks; haulage grew 2.2x over 5 years;
- +17% excavator productivity;
- −20% instances of a missing truck per trip during a shift (in automatic balancing mode);
- −30% truck time in loading queues;
- −15–25% excavator downtime;
- −60–70% dispatcher workload: instead of manually redirecting hundreds of trucks, dispatchers handle only non-standard situations;
- 95% of directives are carried out by drivers without calling the dispatcher.
Multi-Agent System Impact
- 100% of losses recorded in real time vs. less than 10% in conventional reporting;
- 15 minutes less unplanned downtime per refueling;
- 25% fewer early departures and late returns at shift changes;
- over RUB 20 million in economic impact in the first month of operation.
Economic Impact
In 2025 alone, the haul truck balancing module delivered over RUB 300 million in confirmed impact.
Conclusions
This case shows that moving from reactive dispatching to proactive, telematics-based management makes it possible to increase haulage without expanding the fleet. Algorithms predict deviations in advance, drivers receive precise directives right in the cab, and dispatchers focus on non-standard situations.
Contact: sales@galileosky.com