摘  要

工程进度是项目参与各方控制和决策的重点，工程建设过程中，纵向有契约关系的主体间以及横向无契约关系的平行主体间的进度冲突现象频发。造成各方进度冲突的原因除进度计划本身的复杂性外，很重要的一点是各方在进度控制上的利益冲突。各方都站在自身立场以利益最大化为目标进行决策，忽略了决策的相互影响，缺乏均衡优化的思想。而集中决策模式下的进度冲突协调，需要决策者收集所有相关信息，进行集中运算，求解工作量大，决策变量的个数在建模之后也就固定了，不适合工程环境的动态性；而且，由于忽略各方的自治性（信息私有性、利益独立性、决策自主性），很难在工程进度冲突协调这样的信息不完全和多主体分散决策的实际环境下得到应用。为此，本文从利益协调入手，运用博弈均衡等理论，针对总承包工程，构建进度冲突协调的量化模型，指导进度冲突各方的利益协调。同时，工程环境的动态性、信息不完全和多主体分散决策特点使得协商成为自然而重要的利益协调途径，为此，本文后半部分以前述的量化模型为基础，基于MAS协商协调理论，对进度冲突的MAS协商协调做了较为深入的探讨。从而形成了本文如下研究内容：

首先，明确了合作型项目组织中进度冲突协调的层次性，其中协商协调层和利益协调层是关键。考虑参与各方的自治性，将均衡优化作为协调的基本思路。指出纵向主体间进度冲突协调具有主从对策特点和决策的层次性，横向主体间进度冲突协调具有平等协商性。鉴于横向主体间进度冲突的主控协调模式的不足，引入自主协调模式作为有益补充，给出了进度冲突协调合作形成的三种机制及其互动关系。

在纵向，业主、总承包商和专业承包商是主从关系，可以在完善激励措施的基础上，实现双方进度目标的协调优化。这就涉及到上层如何选择激励对象并有效地分配激励和下层在何种情况，以多大的努力加速进度等问题。为此，建立以总承包商为主方、多专业承包商为从方的进度协调主从对策模型，利用逆向归纳法对模型进行求解，给出模型均衡解的形式化数学表达，对可能出现的各种情况进行了系统分析，并结合具体实例解释模型参数对各方均衡解的影响，得出各方协调优化原则，用于指导主从各方在进度协调中的决策。研究表明模型可以实现进度目标的协调优化和各方收益的帕累托改善。

在横向，专业承包商间是不存在契约关系的平行主体，出现进度冲突时需要各方的自主协调。以何种形式实现冲突的协调、各方如何决策、协调收益如何分配等是需要考虑的问题。引入多主体 TCTP（MaTCTP）来描述平行主体间进度冲突的自主协调。考虑任务所属主体的自治性，MaTCTP 是理性个体协作问题。借鉴资源交换协调的思路，将 MaTCTP 求解看作不同利益主体的任务间关于时间资源的交换协调，建立利益补偿机制解决 MaTCTP 中主体间的利益转移问题。在补偿机制的基础上建立基于 MaTCTP 的进度协调合作博弈模型，利用 Shapley 值法分配协调收益，并分析各方协调过程的决策行为。最后以实例阐述了 MaTCTP 建模和合作博弈模型的应用过程。

上述主从对策模型和合作博弈模型都不涉及理性个体达成合作的过程，需要构造适当的协商模型加以表现。同时，工程环境的动态性，使得达成的上述协调协议具有“契约不完全性”，基于协商的进度冲突协调更符合工程实际环境。为此，在个体理性，信息交换，MAS协商的基础上，建立进度冲突的MAS协商协调模型。

首先，利用Multi-Agent表达工程任务网络，建立进度冲突协商协调的MAS虚拟联盟，联盟中Agent扮演主控Agent（MAgent）和任务Agent（RAgent）两种角色，RAgent间是平行关系，MAgent与RAgent间是主从关系，彼此通过协商来协调进度冲突中的利益矛盾。给出RAgent间协商协调的整体目标，对RAgent间的交互进行建模，设计RAgent的效用计算模型和交互推理过程，使得RAgent在追求个体理性的同时实现整体目标。构建两阶段承诺协商协议支持RAgent间的信息交换以达成“互利”合作，并给出RAgent间协商协调的步骤，再以动态再调度为例阐述其应用。采用合同网刻画MAgent与RAgent间进度冲突的协商协调，MAgent与RAgent通过招投标建立初步的协调意向，然后就工期压缩和奖励金额等进行自动协商。给出了自动协商的偏好模型、效用函数和协商策略，并以第3章中实例的数据进行协商过程的仿真，结果表明能够达到彼此协调的目的。进度冲突的MAS协商协调与传统的集中决策模式下的冲突协调相比，更符合工程实际，易于理解和应用，而且Agent的构造灵活，模型具有更大的柔性。

关键词：工程项目；进度冲突；主从对策；合作博弈；TCTP；MAS；协商协调

Abstract

The process control is an importantdecisionmaking issue between stakeholders, in largescale engineeringconstruction project. The process conflicts happens frequently duringconstruction, between the owner、general contractor、subcontractors which are contractual relationship, and betweensubcontractors which are equal in status. Besides the complexity of processcontrol itself, the main cause of these problems rest with the conflict ofinterest behind process conflict. Every stakeholder makes his decisions from theirown perspectives, without considering the interactive impacts between eachother’s decision, the project management practices lack the equilibriumoptimization. The main process conflict coordination approaches establishprocess conflict coordination mathematic programming model, basing on theCentralized Decision Making, where the central decision-maker has to collectall information, and the model can hardly be changed once it is constructed. Ithas a lot of limitation in project practice, due to ignoring the autonomy ofthe stakeholders. Especially it is not easy to be applied in the project dynamic environment where information isincomplete and inaccurate.

With the purpose of improving thecooperative relationship in cooperative project organization, starting with theinterest coordination, by generally applying system engineering theory, gametheory and some other methods for system optimization and decision-makingmodeling, this dissertation focuses on process conflict coordination for generalcontractor engineering, and qualitatively analyze and quantitatively model to gainsome effective coordination tactics. At the same time, due to the dynamic environment、informationincompletion and decentralized decision, negotiation between stakeholdersbecomes an important approach to coordinate the conflict of interest behind processconflict. So, basing on the quantitative models abovementioned, thisdissertation studied the MAS coordination applied in project process conflictcoordination. The main contents and conclusions are as follows:

Firstly, the coordination problem ofprocess conflicts may be separated into three coordination layers: thecommunication layer, the negotiation layer and interest coordination layer, inwhich the latter two layers is core issue. Taking into account the self-governmentof the stakeholder, the equilibrium optimization as main approach is advocated.Analyze the principal agent decision relationship between the general contractorand the subcontractor and the peer-to-peer negotiation relationship between thesubcontractors, during the conflict coordination. Due to the limits of thecentralized coordination pattern, an integrative coordination pattern is putforward.

In vertical line, the project owner、generalcontractor and sub-contractor is principal agent relationship. Followingimportant issues need answered: from the  general contractor perspective, how toselectively inspire critical subcontractors to achieve his whole goal and howmuch inspiration density is optimal? From the subcontractor perspective, how muchthe compression time is appropriate, given the inspiration density from the generalcontractor.

Based on the game theory and bi-levelprogramming method, with the purpose of increasing the revenues, with thegeneral contractor as leader  and  subcontractors as followers, a Stackelbergmodel about dynamic interactive decision-making behaviors for process conflictcoordination  during construction  is established. The equilibrium solution andinteractive influencing between decision variables of general contractor and sub-contractorare  analyzed  and  inferred  theoretically. The research results  showed that themodel can help to actualize coordination  optimization  of project process andPareto improvement of revenue object for the stakeholders in contract.Furthermore, considering the owner’s revenue  of  project operation stage and thegeneral contractor’s opportunity cost etc, a two-stage Stackelberg model is putforward through an extended study.

In the level, subcontractors are equal inthe status. They form into dynamic alliance to deal with the process conflicts.Following issues need answered: What kind of cooperative agreement is to gain?How to make decision during the coordination? How to allocate the coordinationrevenues soundly?

Introduce multi-agent TCTP (MaTCTP) tocoordinate the process conflicts between the subcontractors. Considering theself interesting  and  independency  of  subcontractor, a compensation mechanism isdevised to attack the benefit transfer issue in MaTCTP.  Model the process of resolving MaTCTP asnegotiation coordination between activities of different subcontractors, inwhich each activity negotiates with each other about its activity time and thecompensation fee, basing its resource and cost information. A mathematicalmodel for process conflict coordination is given based on Cooperative Game Theory.The Shapley Value method is applied to the cooperative benefit sharing among thesubcontractors.

The coordination models abovementioned donot come down to the process of how the agreements were got, it is necessary toestablish the negotiation model to depict these processes. Moreover, thedynamic environment of engineering project lead to the “contractimperfectness”, negotiation between stakeholders naturally becomes an appropriateand important approach to resolve process conflicts. Basing on three assumptions,namely individual rationality, information exchange and MAS negotiation, themulti-agent negotiation coordination model for the process conflict isproposed.

Model the project network using themulti-agent methodology and establish the multi-agent virtual alliance for theprocess conflict coordination. In the multi-agent virtual alliance, agents canact the MAgent or RAgent roles, MAgent delegates the upper decision maker suchas general contractor et al, RAgent delegates the lower decision maker such as subcontractoret al. MAgent and RAgent are the principal agent relationship, the RAgents areparallel peer to peer relationship, they negotiate each other to resolve theprocess conflicts between themselves.

Firstly we discuss the negotiation betweenRAgents. Give the collective goal of all RAgents, model the interaction betweenRAgent. The RAgent’s individual utility function and the reason process aregiven, so that the RAgent can come into the collective goal in pursuit ofindividual rationality. Finally, the two stage promises negotiation protocoland the negotiation steps are designed to support interaction between RAgents,and a dynamic reschedule example during engineering project construction stageis given to illustrate the process of negotiation coordination for process conflict between subcontrators.

Basing on the contract net protocol andautomation negotiation model, MAgent and RAgent negotiate to resolve theconflict between them. They establish the preparatory negotiation intent throughthe contract net protocol, and then, both sides go into bargaining about theduration and the compensation fee. The preference model、utility functionand decision functions is given. Finally, simulation analysis is given on thebasis of the data given in third chapter. The MAS negotiation coordination ismore practical and easy to use. These works may have great importance onmulti-stakeholder project process conflict coordination and distributed dynamicschedule decision system and so on.

Keywords: Construction Project; ProcessConflict; Stackelberg Model; Game Theory; TCTP; MAS; Negotiation Coordination