Decentralized All-Source Track and Identification Fusion

"Multi-platform, decentralized tracking for air and ground surveillance has the potential to significantly outperform centralized tracking approaches. Although in principle the centralized architecture is optimal, its use is hampered by the need tointerface with legacy systems, by finite bandwidth constraints, network errors, etc. In addition, due to finite processing constraints, a hierarchical, decentralized architecture is known to outperform the single-stage, centralized approach to trackingwhen processing multiple data streams with widely varying update rates.Our approach to decentralized tracking leverages our experience in multi-source hierarchical tracking and data fusion. Our baseline algorithms include an efficient approach to multi-hypothesis tracking, cutting edge nonlinear filtering, mixeddetection-level and track-level processing, and a recursive approach to track fusion that supports real-time processing needs. New capabilities to be developed under this program include adaptive tracklet formation algorithms that reflect target dynamics,target interaction, network loading, and real time demands; a network-centric data association algorithm that is sensitive to transitions from unambiguous to ambiguous data association problems so as to defer key decision-making to downstream processes;effective tracklet processing algorithms that account for data correlation due to common process noise, sensor measurements, and conte