DISTRIBUTED AGENT-BASED SIMULATION FOR SMART CITY APPLICATION

以智慧城市为应用的基于智能体的分布式仿真

陈扬

11849250

硕士

计算机科学与技术

Georgios Theodoropoulos

2020-05-30

2020-07-08

哈尔滨工业大学

深圳

With the development of computer simulation technology and its application in other research fields, more and more researchers have begun to adopt computer simulations as essential research methods.In computer simulation technology, distributed discrete event simulation has been used more and more with the progress of distributed computing. This type of simulation has excellent performance for large-scale simulation applications.On the other hand, in recent years, in order to solve the problem of food shortage, urban agriculture has gradually become an emerging concept in the study of city planning. In urban agriculture, the food production and consumption chain no longer strictly follows the one-way structure from rural to urban but mixes urban with agricultural production to improve agricultural production efficiency through cutting-edge technologies.This research consists of two parts. The first is to design and implement a mechanism for dynamic computing resources management of distributed simulation engines for multi-agent systems. The second is to build a multi-agent model from the urban agriculture scenario and simulate it with the aforementioned mechanism. The main contributions in this thesis are listed as follows:Firstly, this thesis proposes a dynamic computing resource adjustment mechanism for the distributed discrete event simulation platform PDES-MAS. The mechanism is described in detail from the perspective of the underlying implementation. It monitors the computing resources in the system during simulation execution. When it finds that there are insufficient computing resources or redundant computing resources in the system due to changes in the simulation model's behavior, this mechanism can dynamically allocate or reclaim computing resources to achieve better simulation performance.Secondly, the dynamic computing resource adjustment mechanism is further analyzed, designed, and improved for the event message routing and range query functionalities in distributed discrete event simulation. This thesis studies the implementation details of event message routing and range query in the PDES-MAS simulation platform. By analyzing its assumptions about the computing cluster topology, this thesis clarifies the shortcomings of the current PDES-MAS simulation system in dynamic scalability, reconstructs its existing design and implements a dynamic computing resource adjustment mechanism that supports message routing and range query.Thirdly, a multi-agent model of urban agricultural systems in smart cities is proposed. By simplifying and abstracting the food supply chain in the smart city system, this thesis implements a smart city model on the PDES-MAS platform for the characteristics of a series of links from food production, transportation, and sales.Finally, this thesis completes some performance evaluations for the dynamic computing resource adjustment mechanism based on different simulation experiments. We use the performance benchmark model TileWorld in distributed discrete event simulation and the constructed smart city model to simulate under different experimental conditions. The performance of the dynamic computing resource adjustment mechanism is analyzed by the time required for simulation.Through the results of this study, we not only improve the performance of distributed discrete event simulation platform PDES-MAS in large-scale simulation applications by dynamic computing resource adjustment mechanism but also extend the application scope of computer simulation technology according to the modeling and simulation of practical problems. The design of message routing and load balancing in the dynamic computing resource adjustment mechanism can be applied to other parallel discrete event simulation systems. On the other hand, the smart city model established can also be used to build a more complex and realistic simulation model by adding more content and restrictions.

随着计算机仿真技术的发展和在其他学科中的应用，越来越多研究领域开始采用计算机仿真来作为重要的研究手段之一。在计算仿真技术中，分布式离散事件仿真随着分布式计算的进步得到了越来越多的应用。这类仿真方式对大规模的仿真应用具有较好的性能表现。另一方面，近年来，为了解决粮食短缺问题，在城市规划学研究中，城市农业逐渐成为受到人们关注的新兴概念。在城市农业中，食品生产消费链不再严格按照从农村到城市的单向结构，而是将城市与农业生产混合，通过前沿技术提高农业生产效率。本文的研究内容主要包含两个方面。首先是实现一种多智能体系统分布式仿真引擎的动态计算资源管理机制；其次是在城市农业系统情境下构建出一个智慧城市多智能体模型，并结合上述机制对其进行仿真。本研究的主要贡献列举如下：第一，对分布式离散事件仿真平台PDES-MAS设计了动态计算资源调整机制。我们从底层实现的角度对该机制进行了详细的描述。它在仿真执行过程中对系统中的计算资源进行监测。当发现因为仿真模型行为方式的变化而导致系统中出现计算资源不足，或者计算资源冗余的情况时，该机制能够动态地分配或回收计算资源。第二，针对分布式离散事件仿真中事件消息路由功能和范围查询功能对进行了动态计算资源调整机制进一步分析、设计和改进。我们首先研究了在PDES-MAS仿真平台中事件消息路由和范围查询的实现方式。我们通过分析其对计算集群拓扑结构的假设，阐明了当前PDES-MAS仿真系统在动态拓展性上的缺陷。我们针对其已有的设计进行重构，实现了动态计算资源调整机制能够支持消息路由和范围查询的支持。第三，设计了智慧城市中城市农业系统的多智能体模型。我们通过对智慧城市系统中的食品供应链进行简化和抽象，针对从食品生产，运输，销售等一系列环节的特点，于PDES-MAS平台上实现了智慧城市模型。第四，完成了对动态计算资源调整机制基于不同仿真模型的的性能测试。我们使用了分布式离散事件仿真中的性能基准模型TileWorld以及构建的智能城市模型，在不同的实验条件下分别进行了仿真。我们通过仿真所需时间作为性能的评判标准，分析了动态计算资源调整机制的性能表现。通过该项研究的成果，我们不仅通过动态计算资源调整机制提升了分布式离散事件仿真平台PDES-MAS在大规模仿真应用上的性能表现，也根据对实际问题的建模和仿真，扩展了计算机仿真技术的应用范围。本研究在动态计算资源调整机制消息转发和负载均衡中的设计可以应用在其他的并行离散事件仿真系统中。同时，我们建立的智能城市模型，也可以通过增加更多的内容和限制条件，被用于建立更为复杂和贴近现实需求的仿真模型。

parallel and discrete event simulation smart city load balancing multi-agent system urban agriculture

并行离散时间仿真 智慧城市 负载均衡 多智能体系统 城市农业

英语

联合培养

南方科技大学

Chen Y. DISTRIBUTED AGENT-BASED SIMULATION FOR SMART CITY APPLICATION[D]. 深圳. 哈尔滨工业大学,2020.