A Learning Agent For A Multi-Agent System For Project Scheduling In Construction
A Learning Agent For A Multi-Agent System For Project Scheduling In Construction
复制标题
DOI：
10.7148/2016-0011
复制
发表时间：
2016-06
期刊：
影响因子：
--
通讯作者：
F. Wenzler;W. Günthner
中科院分区：
文献类型：
--
作者：
F. Wenzler;W. Günthner
创建的项目计划的质量对于实现建设项目至关重要。这对规划者来说是一个巨大的挑战，因为有许多限制因素需要考虑。所要解决的问题被称为多模式资源受限项目调度问题（MRCPSP）。本文提出了一种多智能体的方法，其中的资源和流程表示为协作代理。自治进程和资源代理在一个中央黑板上注册自己，在那里协商活动的资源分配。作为对现有工作的扩展，集成了一个学习代理来改进所创建的解决方案。一个离散事件模拟实现了模型，它是从运筹学领域的标准化项目计划进行评估。
The quality of the project plan created is essential for realizing a construction project. This is a big challenge for planners, because there are many constraints to be considered. The problem to be solved is known as the multi-mode resource-constrained project scheduling problem (MRCPSP). This paper presents a multi-agent approach in which resources and processes are represented as collaborative agents. Autonomous process and resource agents register themselves on a central blackboard where resource allocation to activities is negotiated. As expansion to prior works, a learning agent is integrated to improve the solutions created. A discrete-event simulation implements the model and it is evaluated with standardized project plans from the field of operations research.