Sequential simulation

This research area focuses on methods to accelerate sequential simulation, which traditionally suffers from significant sequential dependencies limiting parallelization and speedup. With the rise of high-performance and exascale computing systems, developing algorithms that reduce the inherent sequential bottleneck (sequential fraction ψ) and better overlap communication and computation becomes critical for scaling simulations to large problem sizes. Path-level parallelization and delayed difference schemes act on the sequence and timing of simulation events to relax data dependencies and enable more concurrent execution, thus improving scalability and efficiency.

This theme addresses the challenge of initial transient bias in steady-state simulations, where early simulation data can distort final statistical estimates. Traditionally, heuristic rules or statistical tests are employed to delete transient data, but this can be complex and uncertain, especially as computing costs decline and longer simulations become feasible. Research here investigates the reliability of transient deletion approaches within sequential simulation frameworks, coverage analysis methods of confidence intervals for output parameters, and proposes strategies to balance simulation cost, accuracy, and bias reduction, leveraging sequential analysis and confidence interval estimators.

This theme explores the deployment of sequential simulation beyond traditional analytic contexts, treating it as an immersive educational methodology and a quality improvement technique in healthcare and other socio-technical settings. Sequential simulation here emphasizes re-creation of longitudinal system behavior and interconnected stages, which supports experiential learning and system performance improvement. Research integrates multidisciplinary teams, anticipatory aspects, and the extended patient or process journey to better prepare practitioners and test system changes with minimal risk.