LETI: LETI Research Will Increase the Efficiency of Drone Recognition

The mathematical model developed by employees of Prognoz Research Institute of LETI will help solve the problem of radar recognition of low-visibility UAVs.

Today, unmanned aerial vehicles (UAVs or drones) actively enter all spheres of life. They solve applied tasks: mapping, cargo delivery, reconnaissance. However, in the hands of intruders, drones are an effective tool for destructive operations. Radars are used to detect drones, but airspace control is complicated by the small size of UAVs: the signal reflected from a drone can be missed or confused with the signal from a bird flying by (drones fly at the same height with a comparable speed). That is why radar system developers around the world are now searching for methods to accurately detect and recognize drones.

Employees of the Research Institute of Forecasting and Monitoring Systems for Emergencies (Prognoz Research Institute) of LETI have developed a mathematical model that allows the simulation of signals reflected from drones. The regularities revealed by the modeling were confirmed in experimental studies. Researchers measured signals reflected from drones in an anechoic chamber of the Collective Use Center of LETI, which minimizes the influence of interference from extraneous signals.

“Analysis of the simulation results and the experimental studies showed dependences of the structure of the reflected signal spectrum on such radar parameters as the bistatic angle, the frequency of the probing signal, the polarization of the receiving and transmitting antennas, and the material of the drone propellers. The results obtained can be used for radar design when selecting the parameters necessary for effective detection of drones.”

Ekaterina Plotnitskaya, Engineer at Prognoz Research Institute of LETI
The researchers have found that the ability to detect drones using radar systems depends on many parameters, including the materials from which propellers are made. For example, the signal reflected from rotating carbon propellers is clearly visible over a large range of probing frequencies, while detecting drones with plastic blades requires frequencies in the centimeter range and higher. In addition, the researchers found a dependence of the spectral characteristics of the reflected signal on the position of the drone relative to the radar receiver and transmitter.

The results were published in the proceedings of Signal Processing Symposium 2021 (SPSympo 2021).

“We can use the developed model to create a set of model data needed both for preliminary selection of parameters and evaluation of the characteristics of the developed radar and for debugging drone recognition algorithms,” explains Ekaterina Plotnitskaya.

In the future, the scientists plan to use simulation data from the developed model to train neural networks to automate drone recognition.