Summary

A simulation consultancy worked with Pathmind to apply reinforcement learning to intelligently re-route haul trucks at an open-pit mining site. The mining optimization project discovered a solution that resulted in a 19% increase in ore preparation.

Challenge

Open-pit mines need to transport ore to the ore preparation plants (OPPs) before they are sent to ore processing plants. Haul trucks move between the mines and the OPPs. Those trucks and their diesel fuel are some of the mine’s largest expenses. OPPs are often placed in-pit to minimize elevation changes and distances traveled for the trucks. Even with OPPs close by, the haul trucks need to be routed intelligently between the mines and the OPPs. 

Open-pit mining optimization requires more than simple heuristics such as “shortest queue,” which simply sends trucks to the facility with the least wait time. True optimization requires the coordinated action of all trucks working together, and aware of one another’s movement, like a team on a field. 

Why Pathmind Was Needed

Pathmind was approached by a simulation consultancy supporting a mining company that needed to increase the efficiency of its haul truck routing. 

The miner wanted to: 

  • increase the amount of ore processed
  • proactively deal with equipment failures
  • minimize the number of trucks and fuel required to operate at capacity

Commercially available solvers, optimizers based on linear programming, do not adapt well to multi-agent scenarios, such as when a mine has to make decisions for many trucks simultaneously. 

Pathmind was able to train a reinforcement learning algorithm that coordinated the action of all the haul trucks at the site in an agent-based simulation model, testing various strategies to lower the time required to deliver ore to the OPPs. 

Outcomes

For this mining optimization project to be successful and for the company to consider replacing its existing technology, it needed to see at least double-digit improvements over its current method.  

Pathmind’s web application was able to train a solution that produced a 19% lift in the amount of ore processed by the OPPs in simulation, by routing the haul trucks more intelligently and coordinating their movements. Based on those results, the site has a new strategy to process more ore, more efficiently. 

For more information on how to coordinate haul trucks more efficiently and using Pathmind for other mining optimization projects, contact us.