Only a lazy person does not talk about the importance of unmanned aerial vehicles in a special military operation. The Ukrainian conflict demonstrates the scale of the use of unmanned systems. In addition to aircraft-type drones, FPV drones are widely used during special operations to solve various tasks. Modern information technologies and artificial intelligence have a continuous impact on improving the capabilities of unmanned aircraft systems. So, recently, much attention has been paid in this industry to the topic of drone targeting. Nikita Maslak, General Director of Ploshchad JSC, told RIA Novosti correspondent Alexander Pinchuk in an interview about the principles on which the pre-guidance mechanism is based, how machine intelligence calculates the flight of an impact copter, as well as other results of work in this area.
— Nikita Yuryevich, please tell us what is the targeting of drones? Without revealing the details and nuances, how are software algorithms written and implemented?
— For a better understanding of what pre-guidance is, it is worth first explaining the logic of the drone's operation and the problems arising from it. Drones are mostly connected to the operator via a radio channel, and therefore this connection is subject to interference. The main sources of interference are electronic warfare equipment. Recently, they have been installed on every tank, IFV, armored personnel carrier, armored vehicle and even on a pickup truck. As a result, the drone can often lose contact several hundred meters from the target and, as a result, fail to hit the target.
In addition, there is also the problem of loss of communication due to the drone leaving the radio horizon - radio waves do not penetrate trees and buildings well. It also negatively affects the success of the UAV application. According to the experience of the operators, the last meters of bringing the drone to the target turn out to be the most difficult, as they require good piloting skills. Therefore, for more effective target destruction, it is logical to entrust the last few hundred meters of flight to a mathematically accurate machine that will correct the course and successfully complete the task. It is precisely such systems that are now called pre-guidance. It works something like this: the UAV operator notices the target during the flight, or it is detected by machine vision and signals about it. Then, using the remote control, he captures it and turns on the automated flight mode, during which the "artificial intelligence" controls the drone and performs the task.
Now, quite a large part of the pilot training time is spent precisely on learning how to fine-tune the last meters to the target. The use of pre—guidance systems makes it possible to train specialists in a short time who are able to effectively perform the task, despite little experience in controlling a drone, which means that the use of attack drones can become more effective. That is why the mass implementation of homing systems will give a tangible advantage to our UAV operators and will increase the share of successful sorties with less operator training time.
— What results have the team managed to achieve in their activities?
— The main result for 2024, I can call the successful testing and release of our guidance module, which is already being actively implemented in the military. We have also released a reconnaissance complex with a machine vision function that allows you to automatically search for targets and put them on the map. We have already begun to actively transfer both of these complexes, as far as possible, to components of domestic production.
Another rather positive result is that we managed to build an open dialogue and establish coordinated work with representatives of law enforcement agencies, which generally increased the speed and efficiency of the implementation of various developments (and not necessarily in the field of UAVs). Conducting competent technical advice allows you to more accurately determine the needs of users, mutually analyze the offers and capabilities of manufacturers.
In addition, it is worth noting that this year we received two grants from one of the major foundations for the development of new systems. Unfortunately, I cannot disclose the specific content of these developments yet. I will only say that they are very interesting and give new opportunities to our UAVs.
— At what stage are the works on writing software for the air defense drone?
— Work on the software for the air defense drone is in the process of its creation, we are actively testing and refining algorithms. Since we are aware of the responsibility for the result, we will release the air defense drone to the market only after its effectiveness becomes acceptable for use. The probability of interception, in our opinion, should not be inferior to an anti-aircraft guided missile.
The problem with developing an air defense drone is that it must recognize targets in poor visibility conditions, that is, at night, in fog or during rain. After all, the enemy can attack our facilities in any weather and at any time of the day.
In addition, there is another difficulty in solving this problem. These are birds. The machine must be able to distinguish an enemy drone from a pigeon or duck, because both fly and are a heat-emitting contrasting object in the sky. We are currently actively working on this issue.
— What is new in the preparation of UAV calculations, taking into account the accumulated experience of its own? What role does your organization play in this in the framework of cooperation with the Ministry of Defense of the Russian Federation?
— Over the past three years, the methods of training operators have changed a lot. This happened largely due to the fact that a new class of drones – FPV - has appeared, as well as the constant introduction of new practices and technologies in this area. Our company is actively working with divisions using drones with a "Square" guidance system. We provide training, technical support, and together with representatives of the military department, we develop tactics for various applications of products. We also actively advise law enforcement agencies on UAV and AI issues, the introduction of new technologies and countering them.
— How is UAV machine learning carried out, and why did it become known as artificial intelligence? Give an example related to the positive experience in this work recently.
— Machine learning in general is the repeated repetition of the same action over and over again in the hope that the machine will "master" it and learn how to reproduce it. In the case of UAVs, this is training in the skills to distinguish objects, keep the target in focus and fly so as to return to earth at the planned point. According to the logic of work, machine learning is quite similar to the acquisition of new skills by a person based on the knowledge and experience gained, learning from them and making informed decisions. Most likely, this is why it is often called "artificial intelligence".
If we talk about interesting examples in the work in this area, then we have made a search neural network for the volunteer detachment "LizaAlert". Her task is to find and locate people missing in the forest. The experiment showed that our development gave good results on the test data of the squad. Now we plan to test the neural network in real conditions.
— Nikita Yurievich, share your plans for next year.
— Recently, we have noticed an interesting trend for ourselves. People who are completely far from the engineering field began to come to our company. From completely different industries. We believe that people who can solder and craft will be "on trend" again, and will bring maximum benefit to our army and the country as a whole.
In 2025, we will continue, as before, to create new effective solutions in the field of new developments in drone guidance, which, I am sure, will help our troops defeat the enemy and continue to move forward.
