Answers are grounded in this paper and cite their sources.
Demonstrates superior estimation performance in missile guidance using a deep learning filter, suggesting improved stability under measurement noise uncertainty.
This study proposes a deep neural network-based adaptive line-of-sight (LOS) rate filter for homing guidance under time-varying measurement noise. The Kalman filtering problem for LOS rate estimation is complicated due to the fact that the measurement noise statistics—comprising thermal and glint noise—vary dynamically with the missile-target relative range. To address this issue, the problem of estimating the unknown noise parameters was formulated using measurement residuals as a function approximation problem and then solved using a feed-forward neural network (FFNN). Furthermore, an algebraic transformation was employed to derive an analytic filter gain in order to alleviate the computational burden of solving the differential Riccati equation. The proposed filter integrates the FFNN-based noise parameter estimates into the analytic gain calculation, thereby achieving adaptiveness to the time- varying noise environment without compromising the stability of the Kalman filter structure. Simulation results for a missile-target engagement scenario confirmed that the proposed filter provided superior estimation performance and robustness against measurement noise uncertainty compared to conventional methods.
No takes yet. Share an insight, caveat, or question.
Jeong et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: