活動

Simulation can model the complexity and uncertainty of modern systems. This capability complements the inherent limitation of traditional optimization. The use of simulation under optimization for decision making is growing in popularity. We will present several applications including missile defense agency problem, power grid, and semiconductor scheduling problems. The main challenge for simulation-based approaches is its computational efficiency. Optimal Computing Budget Allocation (OCBA) initially developed by the speaker can dramatically enhance simulation efficiency. Its idea is to maximize the overall computational efficiency for finding an optimal decision. To further reduce the time to reach a good decision, we propose a concept of borrowing time from the future and develop a two-phase framework: i) Pre-event look-ahead simulation-driven learning: Before a decision point, we generate look ahead data by smartly simulating some future functioning scenarios, and discover the distribution of the optimal policy from simulated decisions; ii) Post-event fast-time decision: At the decision point, our innovative synthesizer efficiently utilizes look-ahead simulation learning and additional minimum new simulations to quickly offer optimal actions. This new framework enables fast-time simulation-based decision making.