Dynamic Ranging-Error Compensation and Consistent Cooperative Localization in TDMA UWB UAV Swarms

Abstract

Unmanned aerial vehicle (UAV) swarms increasingly operate in GNSS-denied environments, where cooperative localization provides relative positioning by fusing onboard odometry with inter-agent range measurements, for which ultra-wideband (UWB) two-way ranging is a common infrastructure-free choice. Beyond accuracy, safe swarm autonomy needs a trustworthy measure of positioning uncertainty, since collision-avoidance and formation-keeping decisions derive their safety margins from the reported covariance. Under sustained agile flight, both are hard to achieve at once: motion within each time-division multiple access (TDMA) polling round induces a ranging bias well above the UWB noise floor, and reusing shared information across the network drives the reported covariance below the true error. To address these two problems jointly rather than in isolation, we propose a modular architecture coupling an online maximum-likelihood polynomial least-squares (MPLS) ranging front-end with a fading split covariance intersection (SCI) cooperative back-end through a per-link variance interface: online MPLS compensates the motion-induced bias and reports a calibrated, time-varying variance that fading SCI consumes as measurement noise while its continuous-time fading factor bounds the reused-information covariance. Monte Carlo simulation over anchored and anchor-free 16-node swarms shows the two effects to be empirically decoupled, with front-end ranging quality governing positioning accuracy and back-end correlation handling governing estimator consistency. The method attains sub-meter positioning accuracy in both settings and, without per-scenario tuning, keeps consistency—quantified by the average normalized estimation error squared (ANEES)—within a trusted band; comparably accurate extended Kalman filter and covariance-intersection baselines fall outside it, becoming overconfident and overconservative, respectively. It thus delivers the trustworthy uncertainty that safety-critical swarm decisions require in GNSS-denied flight.