Construction of multi-agent systems with speculative computation

目前对分布式问题求解的研究已经很多，但对不完全通信环境下的分布式问题求解的研究还很少。在本研究中，我们的目的是提供一个多智能体系统，解决上述问题。提出了一种推测计算的机制，并实现了一个原型系统。具体而言，我们提供以下内容。(1)We将已有的具有推测计算机制的主从式多智能体系统扩展为每个智能体都能进行推测计算的系统。(2)We将我们先前的可以处理是/否问题的推测性计算扩展到可以处理约束的机制。(3)We将逻辑程序设计中的推测计算推广到一阶逻辑中，并给出了一种基于结果投资问题的计算方法。(4)We提供了一个计算分析的结果发现问题。(5)We实现了一个推论发现问题定理证明器。(6)We形式化相关理论与投机计算，如默认问题解决，有序默认理论，溯因知识库更新。(7)We并进行了模拟实验，分析了推测计算的有效性。(8)We考虑推测计算在工业中的应用。

There are a lot of research on distributed problem solving, but there is very few research concering incomplete communication environment. In this research, we aim at providing a multi-agent system which solves the above problem. We provided a mechanism of speculative computation and made a prototype system. Specifically, we provide the following.(1)We extend our previous master-slave type multi-agent systems with speculative computation mechanism into systems where each agent can perform speculative computation.(2)We extend our previous speculative computation which can handle yes/no question into a mechanism which can handle constraints.(3)We extend our previous speculative computation in logic programming into first-order logic and provide a computational method using consequence funding problem.(4)We provide a computational analyses for consequence finding problem.(5)We implement a theorem prover of consequence finding problem.(6)We formalize related theory with speculative computation such as default problem solving, ordered default theory, abduction for knowledge base updates.(7)We make an simulation experiment to analyze speculative computation with consideration of effectiveness.(8)We consider an application of speculative computation in industry.