电子学报
電子學報
전자학보
ACTA ELECTRONICA SINICA
2014年
12期
2448-2456
,共9页
分布式仿真%群体仿真%任务划分%图形处理单元
分佈式倣真%群體倣真%任務劃分%圖形處理單元
분포식방진%군체방진%임무화분%도형처리단원
distributed simulation%crowd simulation%task partition%graphic processing unit (GPU)
本文通过引入环境结构因素,提出了一种适用于多层次复杂环境的自适应任务划分算法。自动读取场景模型并通过理解转换为连通邻接区域集,然后对区域进行快速粗粒度划分,有效提高划分性能;自然消除了被障碍隔离的相邻区域个体间的感知计算,大大减少了节点间通信量,使之更适合于大规模群体仿真应用。实验结果表明该算法的划分代价和执行性能均较优。文中设计了一种适合该划分算法的分布式仿真模型,基于该模型的分布式系统对室内多层楼宇或室外场景大规模群体仿真均具有较高仿真性能,相同规模群体的仿真性能与仿真节点数成线性关系表明系统具有良好的可扩展性。
本文通過引入環境結構因素,提齣瞭一種適用于多層次複雜環境的自適應任務劃分算法。自動讀取場景模型併通過理解轉換為連通鄰接區域集,然後對區域進行快速粗粒度劃分,有效提高劃分性能;自然消除瞭被障礙隔離的相鄰區域箇體間的感知計算,大大減少瞭節點間通信量,使之更適閤于大規模群體倣真應用。實驗結果錶明該算法的劃分代價和執行性能均較優。文中設計瞭一種適閤該劃分算法的分佈式倣真模型,基于該模型的分佈式繫統對室內多層樓宇或室外場景大規模群體倣真均具有較高倣真性能,相同規模群體的倣真性能與倣真節點數成線性關繫錶明繫統具有良好的可擴展性。
본문통과인입배경결구인소,제출료일충괄용우다층차복잡배경적자괄응임무화분산법。자동독취장경모형병통과리해전환위련통린접구역집,연후대구역진행쾌속조립도화분,유효제고화분성능;자연소제료피장애격리적상린구역개체간적감지계산,대대감소료절점간통신량,사지경괄합우대규모군체방진응용。실험결과표명해산법적화분대개화집행성능균교우。문중설계료일충괄합해화분산법적분포식방진모형,기우해모형적분포식계통대실내다층루우혹실외장경대규모군체방진균구유교고방진성능,상동규모군체적방진성능여방진절점수성선성관계표명계통구유량호적가확전성。
The paper proposed an adaptive partition method considering environmental structural factors,which is applicable for multilayered complex environment .Simulation scene is automatically extracted to walkable adjacent areas,and a coarse granular-ity partition based on regions is applied to get shorter execution time .The inter-individual perceptual computing of any two individu-als separated by obstacle between two adjacent regions is negligible,so it efficiently reduces the inter-node communication cost and makes the algorithm more suitable for large scale crowd simulation .The results show the proposed algorithm gets lower cost and higher performance .An efficient distributed simulation model is designed for the partition method,and a distributed system based on the model gets higher simulation performance on both inner door and out door scene .The performance of system with the same crowd size linearly increases with the increase of compute nodes,which proves high scalability of the system .