Radar Communication System Recalibration using DNN-based Unscented Kalman Filter Modeling

In typical air traffic control (ATC) scenario where it is apparently challenging to deploy aircraft for special missions as reconnaissance and surveillance, a proposed ACD-UKF model becomes suitable especially where operational flexibility and without manual aircraft-turning are prioritized objectives. In this paper, novel deep neural network model (DNN) enhanced accurate continuous-discrete unscented Kalman filtering (ACDUKF) model for a radar's ordinary differential equation (ODE) solver system navigation tool is presented. The ODE solver system essentially works to control radar navigation parameters with-respect-to (w.r.t) global error control, monitoring metrics and tracking capabilities. With the proposed DNN scheme, limitations resulting from matrix factorization are addressed. A seven-dimensional (7-D) radar tracking drawback in constrained condition is mirrored, allowing the deployed aircraft to conduct supervised turns using the proposed ACD-UKF model. Performance evaluation was then conducted where real-time factors such as the system's outage thresholds, network sum-rate and yaw differences for the global navigation satellite system (GNSS) propelled aircraft radar tracker data-set, in stationary and in accelerating positions were trained and validated using the proposed DNN model.