Multi-agent broadcast communication and self-screening system and method based on storage and calculation integrated basic unit
By establishing a multi-agent broadcast communication and self-selection system based on in-memory computing basic units, the system solves the problems of addressing overhead, high energy consumption, and poor heterogeneous hardware compatibility in agent communication under the traditional von Neumann architecture. It achieves low-complexity, high-efficiency information selection and collaboration, and is suitable for large-scale distributed agent clusters.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HANGZHOU MAGIC BYTE TECHNOLOGY CO LTD
- Filing Date
- 2026-02-25
- Publication Date
- 2026-05-26
AI Technical Summary
Traditional von Neumann architecture-based agent communication suffers from problems such as huge addressing overhead, high energy consumption for data transfer, rigid collaboration, and poor compatibility with heterogeneous hardware, making it difficult to achieve efficient collaboration of large-scale distributed agent clusters.
A multi-agent broadcast communication and self-selection system based on in-memory computing basic units is adopted. Through a semaphore-type communication mechanism, in-memory computing is used to achieve parallel matching and self-selection of signals, eliminating the complexity of traditional network protocols and realizing deep integration of storage and computing.
It reduces communication complexity and energy consumption, improves system scalability and compatibility with heterogeneous hardware, enables efficient information filtering and collaboration, and supports dynamic adaptation and optimization of large-scale intelligent agent clusters.
Smart Images

Figure CN122093201A_ABST