INTELLIGENT CYBER-PHYSICAL SYSTEM FOR UAV NAVIGATION UNDER ENEMY ELECTRONIC WARFARE CONDITIONS

Abstract

In modern conditions of combat operations, in particular during the Russian-Ukrainian war, the role of unmanned aerial vehicles (UAVs) has become decisive. However, the massive use of electronic warfare (EW) by the enemy often makes it impossible to use standard satellite navigation systems (GNSS) due to signal suppression or substitution. This creates an urgent need to develop autonomous cyber-physical systems that are capable of ensuring accurate positioning of UAVs in the complete absence of external radio signals. The research is aimed at finding and implementing an effective software and hardware solution for relative positioning of UAVs. The main focus is on creating a system that combines data from inertial sensors and optical flow cameras to ensure reliable return of the device to the take-off point or maintaining position without using GPS. The paper proposes a conceptual model of the system based on the integration of data from an accelerometer, gyroscope, magnetometer, barometer, lidar, and optical flow camera. The extended Kalman filter (EKF) was used to filter and combine these data. A comparative analysis of three algorithms for calculating optical flow was conducted: sparse (Lucas-Kanade), dense (Farneback), and the modern neural network approach RAFT (Recurrent All-Pairs Field Transforms). To increase accuracy, a three-level filtering system and exponential smoothing of motion vectors were developed. Testing of algorithms in different scenarios (flight at an altitude of 25 m and movement at low altitude) showed that the sparse algorithm, despite its high speed, is the least stable and prone to chaotic errors. The dense algorithm demonstrated low efficiency under certain calculation parameters. The best results in terms of stability and accuracy of route reproduction were shown by the RAFT algorithm, especially its reduced model, which works 1.3 times faster than the full-size version. It was found that an intelligent cyber-physical system based on the RAFT algorithm and multi-sensor data fusion is the most promising for navigation under EW conditions. Despite the high requirements for computing resources, the use of specialized NPU or TPU modules allows achieving the required performance in real time.