Conflict Data Fusion in a Multi-Agent System Premised on the Base Basic Probability Assignment and Evidence Distance.

Area of Science:

• Artificial Intelligence
• Information Fusion
• Decision Support Systems

Background:

• Multi-agent information fusion (MAIF) systems enhance problem-solving in complex environments by enabling agent cooperation.
• Dempster-Shafer (D-S) evidence theory is crucial for multi-source data fusion but struggles with highly conflicting data.
• Traditional D-S combination rules can yield counterintuitive results when evidence is contradictory.

Purpose of the Study:

• To propose a novel conflict data fusion method for MAIF systems.
• To address the limitations of traditional Dempster combination rules with conflicting evidence.
• To improve the accuracy of identification processes within MAIF systems.

Main Methods:

• A new method based on base basic probability assignment (bBPA) and evidence distance is developed.
• Initial belief degrees for each agent are constructed using new bBPA and reconstructed BPA.
• Evidence reliability is modified by calculating information volume via evidence distance.

Main Results:

• The proposed method effectively modifies evidence reliability based on information volume.
• Numerical examples demonstrate the method's effectiveness in handling conflicting data.
• The approach leads to more reasonable evidence fusion and improved identification accuracy.

Conclusions:

• The developed conflict data fusion method enhances MAIF system performance.
• The technique provides a more robust approach to handling conflicting evidence in D-S theory.
• This work contributes to more accurate and reliable information fusion in multi-agent systems.