EVENT DOSSIER
To address the issue of evaluating running time as scale increases in task-parameterized applications, a study published in September 2026 proposed restructuring specific serial computations into forms suitable for parallel execution. The study conducted evaluation extensions at two levels: multitask optimization utilized an accumulation matrix representation for reconstructing recursive motion arm evaluations; sequential transition optimization used a hybrid matrix representation for reconstructing point-by-point B-spline trajectory evaluations, covering both trajectory and collision calculations. Experiments showed that both reconstructions maintained high numerical consistency with reference evaluations and significantly reduced running time, achieving end-to-end acceleration ratios of 256.72 times and 93.91 times, respectively. The study concluded that problem-specific restructuring is a practical approach for achieving scalable evolutionary transition optimization, and related implementation and experimental scripts have been released as open-source code.
针对任务参数化应用中问题评估随规模增长导致运行时间增加的问题，本研究将应用特定串行计算重构为适合并行执行的形式。研究将评估扩展分为两个层级：被评估任务数量及单个任务的工作负载。在多任务优化中，利用累积连杆方向的积累矩阵表示对矩阵递归运动臂评估进行重构；在顺序转移优化中，利用混合矩阵表示对逐点 B 样条轨迹评估进行重构，涵盖轨迹与碰撞计算。这两种重构均保持与参考评估的高度数值一致性，并将运行时间大幅降低，分别实现了 $256.72\times$ 和 $93.91\times$ 的端到端加速比。这些结果证明了问题侧重构是实现可扩展进化转移优化的实用途径。相关应用实现和实验脚本已作为开源代码发布以支持可重复性和重用。