Distributed multi-agent system architecture design and performance optimization for social simulation in universities

University social simulation faces four engineering pressures: strong community topology, periodic activities, bursty interactions and rule-neural-LLM heterogeneous agents. Existing distributed platforms assume spatial locality and a single execution channel, making it hard to balance scalability, synchronization efficiency and the compute-fidelity trade-off. This paper proposes a “partition-synchronize-execute” three-layer co-design. STS-HyperPart injects the campus schedule into a three-mode hypergraph Laplacian for rhythm-adaptive spectral coarsening partitioning, with the prior-fused Laplacian proved symmetric positive semidefinite. Rhythm-PIM & PhaseSync confines bursty rollback cascades to active community pairs through multi-scale rhythm decomposition, sigmoid-calibrated predictive interest management and phase-gated mixed synchronization. Causal-TABE dispatches the rule, neural and LLM tiers under a compute budget using a recursively defined causal-influence gradient on a temporally identified causal DAG, with a Grönwall-form consistency bound and a near-optimality guarantee against the exact multiple-choice knapsack optimum. Experiments on CampusX-50K, a condition-preserving extension of four public campus datasets whose distributional equivalence and higher-order structural fidelity are verified, with leave-one-campus cross-validation, show that at 50 K agents on 16 nodes the system improves EPS by 113.1% over FLAME GPU 2, reduces P99 latency by 54.8% and the cross-partition message rate by 23.0%, and attains a fidelity of 0.91 under a 50% budget, all confirmed by paired testing under Holm-Bonferroni correction at p <0.01. A combinatorial ablation over eight configurations shows marginal contributions amplified by 141 to 242%; cross-campus transfer retains 78.1% of the advantage; and a 70% schedule perturbation bounds degradation to a floor of 35% above the strongest baseline.

Authors

Institutions

Publication Details

Journal
Complex & Intelligent Systems
Published
2026-11-15
DOI
https://doi.org/10.1007/s40747-026-02562-3
Primary Topic
Simulation Techniques and Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Distributed multi-agent system architecture design and performance optimization for social simulation in universities

Xiangyu He
Complex & Intelligent Systems
Simulation Techniques and Applications
article

Distributed multi-agent system architecture design and performance optimization for social simulation in universities

Xiangyu He
article en

Abstract

University social simulation faces four engineering pressures: strong community topology, periodic activities, bursty interactions and rule-neural-LLM heterogeneous agents. Existing distributed platforms assume spatial locality and a single execution channel, making it hard to balance scalability, synchronization efficiency and the compute-fidelity trade-off. This paper proposes a “partition-synchronize-execute” three-layer co-design. STS-HyperPart injects the campus schedule into a three-mode hypergraph Laplacian for rhythm-adaptive spectral coarsening partitioning, with the prior-fused Laplacian proved symmetric positive semidefinite. Rhythm-PIM & PhaseSync confines bursty rollback cascades to active community pairs through multi-scale rhythm decomposition, sigmoid-calibrated predictive interest management and phase-gated mixed synchronization. Causal-TABE dispatches the rule, neural and LLM tiers under a compute budget using a recursively defined causal-influence gradient on a temporally identified causal DAG, with a Grönwall-form consistency bound and a near-optimality guarantee against the exact multiple-choice knapsack optimum. Experiments on CampusX-50K, a condition-preserving extension of four public campus datasets whose distributional equivalence and higher-order structural fidelity are verified, with leave-one-campus cross-validation, show that at 50 K agents on 16 nodes the system improves EPS by 113.1% over FLAME GPU 2, reduces P99 latency by 54.8% and the cross-partition message rate by 23.0%, and attains a fidelity of 0.91 under a 50% budget, all confirmed by paired testing under Holm-Bonferroni correction at p <0.01. A combinatorial ablation over eight configurations shows marginal contributions amplified by 141 to 242%; cross-campus transfer retains 78.1% of the advantage; and a 70% schedule perturbation bounds degradation to a floor of 35% above the strongest baseline.

Complex & Intelligent Systems
Heilongjiang University of Science and Technology (CN)
Simulation Techniques and Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon
Distributed multi-agent system architecture design and performance optimization for social simulation in universities — Xiangyu He · Complex & Intelligent Systems (2026) | TGRS Research Map | TGRS