robust optimization

**Robust Optimization** is a **mathematical optimization framework that seeks solutions performing well under worst-case parameter uncertainty** — ensuring the solution remains feasible and near-optimal for all realizations within a defined uncertainty set, even when the worst case occurs. **How Robust Optimization Works** - **Uncertainty Set**: Define the range of uncertain parameters (e.g., CD variation ±2 nm, temperature ±3°C). - **Worst Case**: Optimize the objective for the worst-case parameter realization within the uncertainty set. - **Deterministic Reformulation**: Convert the uncertain problem into a deterministic (tractable) optimization problem. - **Trade-Off**: Robustness vs. optimality — more robustness typically means slightly worse average performance. **Why It Matters** - **Guaranteed Performance**: Unlike stochastic optimization, robust solutions guarantee performance for all scenarios in the uncertainty set. - **Process Windows**: Finds the center of the process window — maximizing the margin to specification limits. - **Risk-Averse**: Appropriate for high-consequence decisions where worst-case performance matters (yield loss, scrapped wafers). **Robust Optimization** is **designing for the worst day** — finding solutions that maintain performance even under the most adverse parameter combinations.