Integration of Extended Kalman Filtering and Deep Reinforcement Learning for Autonomous UAV Navigation Under GPS Jamming

This paper presents a framework for autonomous unmanned aerial vehicle (UAV) navigation in environments subject to Global Positioning System (GPS) jamming. The framework conceptualizes interference as both a security threat and a source of actionable information. The proposed method utilizes the predictable effects of jamming on received power and path loss to infer environmental parameters critical for navigation. An extended Kalman Filter (EKF) estimates transmission power, received signal strength, and path loss exponent. These estimates are integrated into a Deep Reinforcement Learning (DRL) algorithm, which enables the UAV to optimize its trajectory by combining signal observations with adaptive navigation policies. This approach reduces the adverse effects of jamming and extracts useful parameters for navigation. Simulation results indicate that the EKF-DRL framework supports robust and accurate navigation under severe interference, enabling UAVs to complete missions reliably.