Wind retrieval from UAV-borne forward-detection lidar under limited field of view: ill-conditioning analysis and parameter optimization via least-squares framework

This study investigates ill-conditioned wind retrieval for UAV-borne forward-looking Doppler lidar when the usable azimuth sector is restricted by ground or hard-target blockage. A least-squares framework for dual-wedge conical scanning is developed, and the normal-matrix condition number is derived as κ ( M )∝Δ φ −4 . Based on the covariance propagation Cov (x^)= σ r 2 ( M ) −1 , the condition number directly governs the amplification of measurement noise into retrieval errors, with narrower sectors causing rapidly growing errors. The analysis attributes the amplified along-wind and vertical-wind errors in narrow sectors to loss of orthogonality among the azimuthal basis vectors, whereas the crosswind component remains comparatively robust. Simulations examine the continuous approximation, noise propagation, scanning-parameter selection, non-uniform wind effects, motion compensation, and comparison with GA-PSO. Three flight tests are conducted to validate the theoretical predictions and simulation results under realistic conditions. The experimental measurements confirm the predicted error growth trends with sector narrowing and demonstrate that motion compensation effectively improves retrieval consistency.