A Discrete Task Decomposition Method Guided by Knowledge Graph
Tianguo Jin, Dongliang Zhang, Xiaoqian Liu, Xinglong Chen, Qiao Lei, Wei Min Gao · 2025
As mission planning scenarios become increasingly complex and the coupling between tasks increases, traditional task decomposition methods are difficult to meet the real-time planning requirements of multi-constraint and highly dynamic scenarios. This paper proposes a discrete task decomposition method guided by knowledge graphs. First, for complex discrete tasks, task knowledge subgraphs are extracted from the domain knowledge graph, and the atomic task entities are mapped to low-dimensional vector space using TransR embedding technology to provide prior knowledge for weight calculation. Then, based on the mapping results and the timing relationship between atomic tasks, a timing constraint weighted graph is constructed. Finally, with the weighted graph as input, an improved spectral clustering decomposition algorithm is designed to cluster the atomic task nodes and realize the decomposition of discrete tasks. Under the guidance of the knowledge graph of the space station mission planning domain, the space station operation month events are decomposed into flight control events, verifying the effectiveness of the proposed method.