The Multi-Agent Decision Fusion Method Based on Conflict Evidence Allocation Using Distance Proportions
Liming Sun, Jie Li, Xingwei Wang, Donghui Lv · 2023
In complex and dynamic environments, multi-agent decision systems can supplement each other's expertise and experience, thereby improving the accuracy and reliability of decision-making. The Dempster-Shafer (D-S) evidence theory can handle uncertainty and has been widely used in multi-agent decision fusion. However, conflicting evidence can lead to counterintuitive results in evidence synthesis. To address the issue of conflict evidence synthesis, this paper analyzes the basic properties and shortcomings of traditional D-S evidence synthesis and the confidence-weighted absorption method. A distance confidence-weighted absorption evidence synthesis method is proposed. This method first calculates the Basic Probability Assignment (BPA) after conflict evidence synthesis and then allocates conflicts to conflict focal elements based on the distance proportion between the BPAs of source conflicting evidence and synthesized evidence. Numerical examples demonstrate that the improved algorithm has a higher target identification rate, and the proposed method has a simple calculation process, wide applicability, and effectively resolves the problem of unreasonable conflict allocation among evidence.