//OPEN SOURCE · MINING · ML
RouteOptimizer.
An open-source machine-learning model for mine fleet dispatch from Alio IT. A small network imitates the dispatcher. A* and a CBC shift plan still produce the proven plan. MIT license.
DISPATCH
A* search builds a truck movement plan over a shift. Travel time is the cost. A time limit stops a long solve and reports that status.
SHIFT PLAN
One mine-wide mixed-integer plan, solved with PuLP and the open-source CBC solver. No Gurobi or CPLEX license.
LEARNED POLICY
A small network trained to copy the solver’s choice among legal moves. It ranks actions. The solvers still produce the proven plan.
A shift, in motion
Two trucks leave the garage, load at the shovel, and dump at the crusher. The segment clock advances with them.
Truck 1 and truck 2 stay a half-shift apart. Segment 1 is highlighted while the first truck is still at the garage.

The same demo that runs on the Hugging Face model card: scenario file in, dispatch and shift plan out.
Find it
Demo Space, model weights, source, and scenarios. Install with pip install route-optimizer.
Hugging Face Space
Landing page with the CLI GIF, install commands, and a one-click Colab notebook.
Hugging Face model
Dispatch-policy checkpoint, scenario format, and install notes.
GitHub
pip install from git, Docker image, samples, and the Colab notebook.
Synthetic scenarios
Format-version-1 mine files for demos. Invented graphs, not haul history.
Run in Colab
Install the package in the browser, run A* dispatch and a CBC shift plan on tiny_feasible.
Canonical page: https://www.alioit.com/route-optimizer · Rights: Alfonso Bonillas / alioit.com