The Research On Theory And Method In Parallel Designing Environment Based On Cooperating

Parallel processing has been an extremely effective method to solve many large-scale science and engineering problems. In parallel processing, task partition, processor allocation, communication between processors, and task synchronization should be taken into account. A user not only must master base theories about parallel processing but also must possess high practice skills if he wants to deal with parallel processing, thus leading to parallel processing being difficult to be widely applied. Researching and designing a parallel developing environment that is simple to use and powerful in function is a crucial research field in parallel processing currently.Taking complex electromechanical system as a coalescent, in this thesis, requirements for system designing method that multiple processes interaction proposes are analyzed, parallel designing theory and method based on cooperating is studied, a visible parallel program developing software package and a multi-agent system integrated development environment based on cooperating are designed and realized.Aimed at the characteristic that a great number of dynamic information interaction processes are involved in the parallel cooperation computing, a visible parallel program-developing package based on coupling problem is designed and realized in the thesis. In the package, a cooperation task type is set up, direct and indirect communication modes are supplied, so that parallel tasks can cooperate for many times. Active message reporting and tow-tier task scheduling mechanism are brought forward to reduce bottleneck problem. Automatic producing of cooperating communication statement and automatic modifying of source programs are realized. By enveloping the parallel software, transparency is got in parallel programming and in parallel cooperation solving.Understanding the relations between multiple agents is very important in developing a multi-agent system. Two concepts, multi-agent system architecture and multi-agent system architecture prototype, are brought forward in this thesis, and the method how to describe relations between agents using the concepts is explored. The structure model of multi-agent system integrated development environment is designed; the functions and designing methods of distributed manager and messenger that are designed for agent management and communication are mainly introduced. According to the point that cooperation between distributed agents is erratically, understanding and expressing cooperate message exactly are necessary to agent, so a controlled natural language communicating function is designed and realized. A comprehensive word library for given field is created. Some techniques to enhance the lexical analysis performance, such as replacement algorithm to determine which word in memory should be replaced and the method how to dynamically refresh commonly used word library, are brought forward. A visible multi-agent system integrated development environment is realized, and a set of application programming interfaces for agents are offered. By applying these interfaces the user's workload consumed inagent communication and management is reduced, the efficiency of developing a multi-agent system is effectively advanced, at the same time, robustness of the system is improved.In this thesis, by two practical examples that are coupled problem parallel cooperated solving and overall coupled optimize designing in complex electromechanical system respectively, the performance of parallel program developing software package and multi-agent system integrated development environment is validated.At the end of this thesis, research productions in every chapter are summarized, and the further work that next should be done is prospected.