ParaOpt-Agent: A Knowledge-Driven Decision Agent for Simulation Model Parameter Optimization
Abstract
Simulation model parameter optimization requires selecting optimizers from task knowledge and translating those decisions into execution configurations that satisfy interface, authorization, and budget requirements. We propose ParaOpt-Agent, a knowledge-driven pre-execution agent that combines task and optimizer descriptions, historical cases, and large language model scoring. Optimizers are selected by equally combining model ranking with historical winner-coverage ranking, and deterministic checks then produce call configurations. The portfolio contains 10 optimizers, and the combined ranking uses 224 historical records. Across 2 batches totaling 64 new instances, the post hoc pooled mean utility difference relative to a same-information history rule is +0.001909; the individual batch differences are +0.013045 and −0.009227. On a further 320 new tasks, the Agent produces 319 valid actions and meets a five-percentage-point non-inferiority criterion against a model without history, the same-information rule, and an ExtraTrees selector. A same-input control ablation retains acceptance of all 16 legal inputs and reduces optimizer-entry arrivals among 528 constructed faults from 448 to zero. Public-observation cases using the 20-task history selector demonstrate parameter search and temporal validation. This study separately quantifies the selection utility of combined knowledge, action validity, and the contribution of execution checks, providing empirical evidence for optimizer decisions in simulation modeling.
