Multi-Agent Fusion Based on Extended Region Information Matching With Limited Fields-of-View

Multi-agent multi-target tracking systems can improve tracking performance in a wider range than the limited field-of-view (FoV) of a single agent via information fusion from cooperation between agents. When using FoV information for fusion, the estimated target state of the local agent is categorized into two distinct regions: exclusive and common FoVs. However, owing to the inherent uncertainty between the estimated target state and the actual state, the division of these two types of regions based on the estimated target state may not perfectly align with the real region, resulting in regional inconsistency. Additionally, the information fusion performance is affected by information loss owing to missing reports and detections. To address these challenges, this study proposes an information fusion algorithm based on the extended region information matching strategy. To address the regional inconsistencies, the proposed approach extends the real FoV and introduces a regional matching mechanism to detect and rectify abnormal information. Based on the extended region, a prediction observation vector generation method is applied, and the weight inheritance mechanism is matched to reduce the fusion performance degradation caused by missing reports and detections. Simulated multi-target tracking scenarios verify the effectiveness of the proposed algorithm in centralized and distributed networks with different clutter rates and detection probabilities.