基于ASP的多智能体分布式协同规划关键技术研究

In distributed network environment, the effective coordination of multiple autonomous agents plays an important role in constructing the intelligent systems. In this project, we study the collaborative planning approach of multi-agents with non-monotonic knowledge based on Answer Set Programming. In view of the open characteristics of network environment, the semantic interpretation and reasoning mechanism of ASP knowledge oriented to open world should be defined to provide basic theoretical support for distributed collaboration. For the lack of mature authorization mechnism of multi-agent system with non-monotonic knowledge, an ASP knowledge authorization mechanism is proposed based on security constraint and demand-driven model, which can improve the accuracy and security of knowledge authorization. Due to the possibility of incompatibility between different agents’ knowledge, the consistency judgment theory and conflict resolution method of distributed knowledge base are studied to ensure the correctness of information interaction between knowledge bases. Aiming at agent which own incompleteness of knowledge and under a dynamic of the distributed environment, an incremental collaborative planning method is proposed under incomplete information. The research proposed in this project will have important guiding significance and application value for the collaborative planning of multi-agents with non-monotonic knowledge in distributed environment.

分布式网络环境下，多个自治智能体的有效协同对于构建智能化系统至关重要。项目研究基于非单调ASP知识库的多智能体协同规划方法。针对网络环境的开放性特征，研究面向开放世界ASP知识的语义解释和推理机制，为分布式协同提供基础理论支撑。针对具有非单调知识的不同智能体间缺乏成熟知识授权机制的问题，研究基于安全约束和需求驱动的ASP知识授权方法，从而提升知识授权的精准度和安全性。由于不同智能体之间可能存在知识不兼容的情况，研究分布式知识库的一致性判断理论和冲突消解方法，保证知识库信息交互的正确性。针对智能体知识的不完全性和分布式环境的动态性，研究不完全信息下的增量式协同规划方法。所提出的研究内容对于分布式环境下具有非单调知识多智能体的协同规划具有重要的指导意义和应用价值。

随着Internet、移动互联网和物联网的大规模推广，分布式系统成为最主流的软件应用形式。多智能体是一种适用于分布式环境的人工智能技术，通过多个相互协作的自治主体组成一个完整的系统，不但可显著提高系统处理复杂问题的能力，还可采用合理的协作机制来提高整体系统的智能水平。多智能体技术已被应用到了多个领域，如智能电网、应急事务处理和灾难救援，具有广泛的网络应用价值。目前面向分布式环境的多智能体系统已成为学术界和企业界共同关注的热点。本项目研究分布式计算环境下，尤其是资源受限边缘计算环境下，多智能体协同规划过程中的关键问题，为智能化网络应用的开发提供理论基础和技术支撑。通过项目的开展，课题组在基于缺省语义的ASP推理规则定义与自动生成方法、面向智能体的认知计算模型与服务提供方法、多智能体系统任务规划与协调方法、资源受限环境下多智能体服务自主协同调度方法等方面获得一些成果。在国际知名期刊以及国际会议上发表和录用论文11篇，其中SCI源刊8篇，EI检索国际会议论文3篇；获得发明专利6项，申请发明专利3项。培养硕士研究生8名，均已获得硕士学位。

内容获取失败，请点击重试