Study On Intelligence Decision-making Technics For Multi-Agent Systems In JSSP

Globalization and information development is changing themanufacturing environment, and in turn profoundly influencing a wide rangeof areas including traditional management, control mode, organizationstructure and decision-making rules. As an important part of the job shopmanufacturing system, Job Shop Schedule affects the agility and intelligenceof the whole enterprise. Its control architecture essentially determines theenterprise's production efficiency and competition ability in the market. So itattracts great attention from both the academia and industry. Therefore, thestudy of Job Shop Schedule is of great theoretical and practical importance.In real job shop scheduling scenarios, resource restriction coexists withprocess restriction, which makes job shop scheduling problems(JSSP)NP-hard. Meanwhile, the changeability of market demand together with theuncertainty in a real job shop adds to the complexity of JSSP. As a result,there is no effective and widely applicable job shop scheduling methods.This paper puts forward a MAS(Multi-Agent System)-GASA(GeneticAlgorithm and Simulated Annealing), which is based on GPGP(GeneralizedPartial Global Planning) in the job shop scheduling model. Our schemecombines the advantages of static GASA and dynamic MAS, taking intoconsideration the characteristics of JSSP. Based on the model, weimplemented an effective and widely applicable prototype system fordynamic JSSP. The main research work is described as follow:(1)The paper introduces the background, the categories and thedevelopment of job shop scheduling. Meanwhile, the promising aspects ofGA and MAS for solving NP-Hard problems are highlighted.(2)A GASA hybrid genetic algorithm is proposed for job shopscheduling in this paper. It is using GA (Genetic Algorithm) excellent wholesearch ability and simulated annealing which is efficient to avoid getting intopart minimum. It is capable of generating alternative schedule after uncertaindisturbance takes place on a job shop. An optimum or second optimumsolution is obtained by using multiple crossover and mutation operations.Simulation results based on some flow shop scheduling benchmarks showthat the GASA is feasible, efficient and superior to the simple GA and SA(Simulated Annealing).(3)CNP(Contract Net Protocol) is widely used in manufacturingsystems based on MAS to solve scheduling and controlling problems.However, it has several defects in solving real problems. We gave an outlineof GPGP, a kind of coordination structure for MAS and TAEMS (TaskAnalysis, Environment Modeling, and Simulation) language in this paper.GPGP is a domain -independent coordination structure, which provides a setof coordination mechanisms that can be dynamically and flexibly applied indifferent task environment by various requirements. TAEMS is a powerfultask description language that provides bases for GPGP applications.To overcome CNP's defects in Job Shop Scheduling System based onMAS, a new method was presented dealing with task distributing andscheduling based on GPGP coordination mechanisms and TAEMS. Atask/capability environment model and the process of scheduling part wereput forward that can provide the basic structure of knowledge belief based onscheduling. In MM-MES, a manufacturing execution system based on MAS,they were implemented to research distributed manufacturing system'sscheduling and coordination.(4)The paper introduces MAS mechanism base on GPGP to constructa dynamic job shop scheduling system, with GASA providing basic support.MAS mechanism decomposes a continuous and dynamic job-shop schedulingproblem into series of JSSP so that the predefined GASA can work out theschedule. Specially, we put emphasis on deadlock problems since deadlock isa big obstacle to GASA-based on JSSP solutions. Considering the importanceof deadlock, special adaptations based on MAS are made constructed in thepaper. The new solution much more quickly and can find the most suitableroute for an operation. Finally, the results from simulation shows theMAS-GASA, which is based on GPGP in scheduling system is promising forpractical job shop scheduling.(5)We apply intelligent algorithms GASA and MAS technique toproduction scheduling by summarizing the past research achievements. Wealso introduce the structure and functionality of production planning andscheduling software system acknowledged by the project. By analyzing thesystem functions, we suppose the whole structure of the software system anddiscuss the system structure from two different views of function andworkflow. At the same time, we explain the database, software designing andother pivotal technologies from software engineering. At last, we gave aprocess for job shop scheduling.