基于知识图谱关联的多智能体协同调用方法及系统
By using a knowledge graph-based multi-agent collaborative invocation method, which dynamically corrects association weights and generates invocation sequences, the problems of task matching accuracy and resource waste in multi-agent systems are solved, achieving efficient collaborative invocation and stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SICHUAN PROVINCIAL INSTITUTE OF ARTIFICIAL INTELLIGENCE
- Filing Date
- 2026-04-13
- Publication Date
- 2026-07-17
AI Technical Summary
Existing multi-agent collaborative invocation systems lack deep semantic deconstruction and multi-dimensional vector fusion capabilities when faced with unstructured natural language instructions. This results in insufficient accuracy in task distribution and node matching, static association weights failing to adapt to changes in task semantics, inaccurate redundancy removal mechanisms, and logical deduction severing the mapping between semantic priority and hardware state, leading to sluggish response and resource waste.
By using a knowledge graph-based approach, task feature vectors are obtained, the association weights of agent nodes are dynamically adjusted, and a collaborative call sequence is generated by combining semantic projection gradients and physical load states. Redundancy removal and conflict scheduling are then performed to achieve dynamic adaptation between semantics and physics.
It improves the response timeliness and resource utilization of the multi-agent collaborative invocation system in dynamic high-concurrency scenarios, accurately identifies complementary nodes, avoids false positives, and ensures system stability and execution efficiency.
Smart Images
Figure CN122021708B_ABST