About this log
I’m Mustavi Ibne Masum, a graduate student at Lakehead University. More about me is on my profile page. This site is my daily research log for my thesis on open-pit mine production scheduling.
The project
Working title. Open-Pit Mine Production Scheduling under Geological Uncertainty. A comparison of MILP and LLM-guided simulation across blind, partial and full information.
Open-pit scheduling decides when each block is extracted and whether it goes to the mill as ore or to the waste dump. The objective is to maximize net present value (NPV) subject to slope precedence, mining capacity and processing capacity. The mine is modelled as a 3D grid of blocks.
The classic approach writes the whole schedule as a mixed integer linear program and solves it to optimality. That assumes every block grade is known in advance, it can’t easily react to unplanned events, and it gets slow as the mine grows.
In this thesis a simulator enforces all the physical and operational constraints, and an LLM agent only makes the scheduling decisions. The study compares three information settings.
| Setting | What the scheduler knows | Role |
|---|---|---|
| Blind | Surface grades only; deeper grades revealed as mining proceeds | Lower bound |
| Partial | Surface grades plus noisy estimates along exploratory drillholes, at a drilling cost | Main case |
| Full | True grade of every block | Upper bound |
How this log works
Each entry covers one working day, usually with what I tried, the results, and what’s next. Code lives in the project repository.