Robust optimization package

Setup

1. new virtual environment python3 -m venv env
2. activate virtual environment: source env/bin/activate (windows: env\Script\activate)
3. install requirements: pip install requirements.txt

• To solve non-linear convex model please install ipopt in your computer first (for non-linear uncertainty set)

Explanations

• Robust optimization is a methodology handles uncertainty without distributional assumptions
• You can choose your own uncertainty set (must be convex), supported sets are listed below:
  • Box: robustoptimization.components.uncertaintyset.box
  • Ball: robustoptimization.components.uncertaintyset.ball
  • others: refer to the reference ./refs and inherit robustoptimization.components.uncertaintyset.UncertaintySet on your own
• Metrics for RO solution qualities are defined in ./robustoptimization/utils/metrics.py, including:
  • mean_value_of_robustization: The mean objective value improvement of RO compared to deterministic optimization
  • improvement_of_std: The objective std improvement of RO compared to deterministic optimization
  • robust_rate: Proportion of solutions feasible using RO but infeasible using deterministic optimization

Examples

• supply chain network model
  • mathematical formulation provided at ./scn.md
  • code definition (OOP wrapper) provided at supplychainnetworkmodel.py
  • entry point: python3 main.py --scn
• machine scheduling model
  • mathematical formulation provided at ./scheduling.md
  • code definition (OOP wrapper) provided at ./schedulingmodel.py
  • entry point: python3 main.py --sch

Future works

• only variable >= 0 supported
• check dim consistency
• check variable and parameter naming
• math simplification
• box robustness parameter
• simplify parameter object to uncertain param
• sympy https://stackoverflow.com/questions/30225348/how-to-rearrange-sympy-expressions-containing-a-relational-operator
• bind uncertainty set with constraint as class
• small cases and big ones to demonstrate package util.