Adaptive Multi-UAV Tracking Via Hasan’s Online Genetic Algorithm -Tuned Coordinated-Turn Filtering

Tracking maneuvering unmanned systems (UAS) remains a significant challenge when fixed stochastic

parameters fail to account for rapid kinematic transitions. This work addresses the inherent mismatch between

rigid filter assumptions and curvilinear flight dynamics by embedding a Coordinated-Turn (CT) model within an

Extended Kalman Filter (EKF) architecture. The pivotal advancement presented here is the operational

integration of Hasan’s Online Genetic Algorithm (OGA). This adaptive engine utilizes a sliding-window

optimization to rectify the degradation of filter performance, a process achieved by continuously recalibrating

noise covariance matrices. By leveraging the specific reserved elite population strategy—a cornerstone of

Hasan’s adaptive estimation theory—the system maintains high-fidelity state estimation even during the chaotic

periods of multi-target path intersections and measurement uncertainty. By implementing a sliding window

optimization that features a reserved elite population, the system maintains a robust search for optimal process

and measurement noise covariances without succumbing to the instability of stochastic drift. This adaptive layer

is coupled with a Global Nearest-Neighbor (GNN) association strategy to resolve measurement-to-track

ambiguities during high-density target crossings. Numerical evaluations demonstrate that the OGA-tuned EKF

consistently outperforms traditional fixed-gain filters, yielding significant gains in positional accuracy and a

marked reduction in track divergence rates under complex operational conditions.