Key Takeaways

  • Fleet dispatch is one of the longest-established “AI-adjacent” systems in mining — Modular Mining’s DISPATCH (now under Komatsu) and Wenco FMS (Hitachi) have run truck-shovel optimization for decades.
  • What’s new is the AI layer on top: Modular’s Adaptive Config is an AI-powered tuning add-on that adapts dispatch rules to changing conditions without new hardware.
  • Academic work (reinforcement learning for adaptive ore dispatch) is pushing beyond rule-based/heuristic dispatch toward learned policies, but this is still mostly research-stage outside the vendor add-ons.
  • The control room’s dispatch system maintains a live digital twin of the mine — trucks, shovels, haul roads — which is the substrate any AI dispatch layer optimizes against.

TL;DR

Modernizing dispatch with AI today mostly means turning on an AI add-on to your existing fleet management system — Modular Mining’s Adaptive Config is the clearest named example — rather than replacing DISPATCH or Wenco FMS outright.

How Do I Optimize Mine Dispatch With AI?

If you already run Modular Mining DISPATCH or Wenco FMS, the fastest path to AI-assisted dispatch is a vendor add-on rather than a rip-and-replace. Modular’s Adaptive Config plugs into an existing DISPATCH deployment with no additional hardware and continuously tunes dispatch parameters to current mine conditions — road quality, queue lengths, truck availability — rather than relying on a fixed rule set that needs manual re-tuning as conditions change. Both DISPATCH and Wenco’s FMS already maintain a real-time digital twin of trucks, shovels, and haul roads, which is what makes an AI optimization layer possible in the first place — the AI is scheduling against live state, not a static plan.

Beyond the vendor add-ons, there’s active academic research applying reinforcement learning directly to the truck-dispatch problem — treating it as a sequential decision process where an RL agent learns to route trucks to maximize throughput while respecting grade-blending and queueing constraints, an area explicitly framed as “adaptive ore dispatch” in recent engineering literature. This is genuinely promising but still mostly confined to published research and pilot studies rather than a product you can buy — if your dispatch tuning problem is complex enough to justify it, an RL-based approach is worth tracking, but plan for a custom modeling engagement rather than an off-the-shelf install.

For most operations, the practical starting point is: confirm your existing FMS vendor’s AI add-on availability and licensing, pilot it on a subset of the fleet against your current dispatch performance baseline, and treat custom RL dispatch as a longer-horizon R&D investment rather than this year’s project.

Try It With Geocluster

Working out whether a vendor AI add-on or a custom RL approach makes sense for your fleet — and what the real published evidence says about each — is exactly the kind of grounded technical research Geocluster can help you run before committing budget.