Multi-agent information loss rate optimization selection method
By combining expert experience base with trial and error, the selection of information loss rate in multi-agent systems is optimized, solving the problem of reliance on human experience, realizing the automation and dynamic adjustment of information loss rate, and improving the system's adaptability and training efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NORTHWESTERN POLYTECHNICAL UNIV
- Filing Date
- 2026-03-18
- Publication Date
- 2026-06-26
AI Technical Summary
In existing technologies, the selection of information loss rate in multi-agent systems relies on human experience, lacks systematicity and adaptability, and leads to blindness in the training process, making it impossible to obtain optimal performance in different task scenarios and dynamic environments.
This paper adopts a method that combines an expert experience base with a trial-and-error approach. By constructing an expert experience base, based on historical environmental metadata and performance evaluation records, a feasible range of information loss rate is derived, and a candidate set is generated within the feasible range. The performance index is evaluated using a trial-and-error approach, and the information loss rate is optimized by combining a comprehensive objective function. Finally, a compensation mechanism is introduced for dynamic adjustment.
It achieves automated, intelligent selection and dynamic adjustment of information loss rate, improves the adaptability and robustness of multi-agent systems in different tasks and environments, and enhances training efficiency and operational performance.
Smart Images

Figure QLYQS_1