The US Department of Defense has published a roadmap for the development of military robots. This document describes a vision of what unmanned aerial vehicles, ground, surface and underwater robots should become by 2042. Among the most priority areas for the development of such systems, the military named an open architecture and modularity of design, a high degree of automation to reduce the load on the operator and the formation of mixed units of people and robots.
The leading countries of the world today are actively developing various military robots: from reconnaissance vehicles to combat systems equipped with weapons. The military believes that such devices will allow soldiers to more efficiently and faster solve a wide variety of tasks. In addition, it is believed that robots will reduce losses during combat, more effectively conduct reconnaissance and strike at enemy targets. Finally, robotic systems will significantly reduce or completely avoid collateral damage when performing a particular combat mission.
The US military is already actively using robots in military operations. In particular, the US Air Force includes several types of unmanned aerial vehicles, which the military uses for reconnaissance and observation, as well as for striking reconnaissance enemy positions. The US Navy uses autonomous, unmanned underwater vehicles to search for mines, patrol the entrances to naval bases, provide communications with submarines and map the seabed.
According to the new "roadmap" of the Pentagon, when developing military robots, more attention should be paid to neural network technologies. It is assumed that robots equipped with artificial intelligence will be able to become companions of human soldiers on the ground and in the air. The US military believes that thanks to artificial intelligence, it will be possible to significantly increase the degree of autonomy of armed robots, which can not only detect and take on target tracking, but also prioritize them, reducing the burden on the operator.
The operator of highly autonomous combat systems will only have to make a decision about firing at targets detected by the robot. At the same time, the robots will be able to fire at targets at the operator's command and control their destruction. In general, a significant increase in the degree of autonomy of reconnaissance and combat robots will allow one human operator to simultaneously control several drones and ground systems. For comparison, today in the calculation of one, for example, an attack drone, there are at least two people: a pilot and an operator of sensors and weapons.
The Pentagon also believes that the development of military robots should focus on open architecture and modularity. Among other things, this will simplify the maintenance and modernization of robots, as well as simplify the integration of new combat vehicles into the Pentagon's unified information network, as well as tactical networks. The open architecture will allow third-party companies to create their own modules for robots, which will increase competition and give the military a wider choice of various additional systems for robotic technology.
Finally, an open architecture and modularity will allow for better communication between controlled and unmanned vehicles at the data exchange level. This, in turn, will significantly improve the efficiency of mixed robot and human units. The latter, in turn, in the organizational structure of the armed forces will be able to solve various combat tasks faster and more efficiently. In particular, due to the fact that robots will perform more risky actions. In general, the Pentagon believes that the first partner robots may appear in the troops between 2029 and 2042.
At the end of August 2018, the Defense Advanced Research Projects Agency (DARPA) of the US Department of Defense ordered the American company Raytheon to develop a neural network that could explain why it made this or that decision. These developments will be used in a large-scale project to create mixed combat units. The project received the designation XAI. It involves the creation of a neural network that is not inferior in speed to conventional neural networks, but is able to show the entire chain of analysis, on the basis of which it made a specific decision.
Promotional video:
Vasily Sychev