Simulation Optimization of Spatiotemporal Dynamics in 3D Geometries.

Area of Science:

• Engineering
• Healthcare Systems
• Computational Science

Background:

• Spatiotemporal dynamic processes are common in engineering and healthcare.
• Traditional sequential decision-making methods struggle with complex 3D geometries.
• Simulation modeling is crucial for evaluating decisions in dynamic systems.

Purpose of the Study:

• To develop a novel simulation optimization framework for sequential optimization of 3D dynamic systems.
• To address limitations of traditional methods in complex geometric environments.
• To enable effective decision-making for optimizing spatiotemporal dynamics.

Main Methods:

• Proposed a coherence measure for functional simulation outputs to assess decision effectiveness.
• Developed a Gaussian Process (GP) model with a Hausdorff distance-based kernel to estimate coherence.
• Devised a Normal-Gamma GP Monte Carlo Tree Search (NG-GP-MCTS) algorithm for sequential optimization.

Main Results:

• Successfully implemented NG-GP-MCTS for optimizing ablation paths in cardiac simulations.
• Demonstrated the algorithm's effectiveness in a 3D atrial geometry.
• Achieved promising results in designing sequential procedures for spatiotemporal dynamics optimization.

Conclusions:

• The novel NG-GP-MCTS framework is highly effective for optimizing spatiotemporal dynamics in complex 3D systems.
• The approach shows significant promise for applications in engineering and healthcare, such as cardiac ablation.
• Simulation optimization provides a powerful tool for complex sequential decision-making problems.