Recently, the topic of the use of combat robots is being discussed quite actively, and sometimes in a rather strange way. There is even an opinion that developments in this direction are meaningless, and “stupid iron” is in principle incapacitated: it will collapse into the first funnel, failing to overtake it, it will shoot the cats, failing to distinguish them from the person, management will be easily intercepted by hackers and electromagnetic weapon turn it into trash.
Meanwhile, the potential battlefield is already “saturated” with ATGWs, automatic grenade launchers and automatic cannons - and much more. As a result, the scale of losses in the "big" war between approximately equal opponents will be huge - even in the case of a short conflict. Using the same "mechanical soldiers" provides many advantages.
Firstly, it obviously reduces the expected loss of personnel. Secondly, the non-living force has a much greater stability and "maintainability" than the living. Third, cars do not know fear. Fourth, they can partly compensate for the lack of recruits. Finally, automata have a number of other options, about which below.
Nevertheless, fighting robots have long been something from the realm of fantasy - due to the state of hardware and software. While the actions of the robots were limited to a set of hard programs, even just autonomous movement in a heterogeneous terrestrial environment proved impossible. Similarly, classic (“Neumann”) computers were unable to provide effective pattern recognition.
However, in 1980, electronics began to shrink rapidly, and a bit earlier (in 1975), a multilevel neural network appeared in the world. As a result, it became possible to create “trained” automata capable of making at least elementary “independent” decisions (without which, for example, movement over rough terrain is impossible). Simultaneously, high-resolution surveillance equipment and digital communication lines appeared.
As a result, robotic machines began to appear one after another in the USA: Roboart I, Prowler, Demon and others. However, the first robots were extremely imperfect - for example, the AVL robot record driving serpentine became the worldwide sensation of 1985 of the year ... 1 km. Robotic "vision" and pattern recognition allowed, at most, to fix a suspicious silhouette.
Actually, skeptics appeal to these realities. However, since then, developments have been under way - especially after the reduced US military budget in 2000's has once again expanded to an epic scale.
The progress rates are most clearly seen from the results of the robots races organized by DARPA (teams receive a CD with a route map two hours before the start, external control of the robot is excluded). In 2004, the race in the Mojave Desert ended in complete failure: 7 machines from 15 could not leave the start at all, none reached the finish line, and the maximum achievement was reduced to shameful seven miles.
However, a year later, 4 machines from the 23-x traveled the entire 132-mile distance. 2007 competitions of the year were moved to a specially built campus, with the additional option of 30-ti ordinary machines - to create traffic. The robots had to overcome 90 km through the streets in 6 hours, while they were required to drive through many intersections and turns, drive into the parking lot and leave it, and perform a number of other maneuvers.
Results: 36 passed the qualifying round in the desert from 11, 6 got to the finish line, and 3 got into the car in the allotted time - and with a margin. In 2009, the speed of movement of robots in a “populated” urban environment has already reached 50 km / h - progress is evident.
Of course, the combat vehicle must, at least, effectively recognize the images. And if at the beginning of zero a simple recognition of the “incorrectly” written numbers was a very nontrivial task, then now the recognition of faces from an arbitrary angle and movement is already a completed step for advanced machines.
Now we are talking, for example, about reading very complex emotions. There are also robots capable of identifying themselves in a mirror, while not confusing their reflection with the reflection of a machine of the same type. In other words, the fall in the funnels and the shooting of cats are canceled.
These successes, in turn, are based on the key difference between modern neural networks and ordinary Neumann computers. "Neumann" need comprehensive programs, instructions, and, at most, can move from one "package of instructions" to another (adaptive robots). And the intellectual "neurons" task can be set in a general form, without detailed instructions.
The simplest case is: “go to such a point along such a route, and I don’t care how exactly you will deal with the obstacles you have encountered”; Possible cases and more difficult.
This, in turn, radically changes the functions of the operator. If earlier he had to simply remotely “steer” the robot in non-stop mode, then now he would only have to set tasks and exercise general control. In particularly difficult situations, he can give the car additional instructions. Equally, a robot, faced with an abnormal situation, can request instructions from the operator itself.
In this case, robots, compared with humans, are much better able to cope with routine actions. Thus, during the 2006 test, the SWORDS robot (Special Weapons Observation Reconnaissance Detection System - “weapon system specialized in detection, reconnaissance and observation”) fired from the distance of 1,5 km, and very accurately.
A trained soldier from a distance 300 m hits the target the size of a basketball - the robot hit the coin at the same distance (and 70 shots - without a single slip). Thus, for the first time, a huge advantage of robots was manifested in the performance of the simplest combat work that does not require “creativity”. The last person must provide - and as a result, a system emerges, potentially several times superior in efficiency to the usual “animated” fighter.
So, the presence of continuous communication with the operator for "intelligent" robots is not critical (at worst, the machine can always retreat on its own), although it is highly desirable.
At the same time, it is almost impossible to reliably score a noise-free military communication channel operating over a distance of 1-1,5 km. Further, the radio communication can be duplicated by control over an optical cable. In addition, there is FSO, it is also AOLS - laser communication. A device for laser signaling is equipped, for example, with a new American robot MAARS.
At the same time, neither fog nor smoke are an insurmountable obstacle for laser communication at a distance of 1,5-2 km - all these curtains are completely transparent for sufficiently powerful radiation of some frequencies. So even if any of the communication channels can be blocked, alternative channels will always remain. However, the duplication of communication systems is caused more by fear about mechanical damage to equipment than fear of interference.
Manufacturers and the military emphasize that only humans will give permission to open fire with a robot. But there is every reason to doubt this - such a control scheme will be obviously ineffective. In addition, some people already let it slip. According to one of the developers of the Korean "Intelligent Patrol-Security Robot," he "can independently detect suspicious moving objects, pursue them, and even open fire to kill."
The reaction of the frightened public forced the Korean military to abandon their statements, but hardly - from the development. So, in the 2020-ies, the Korean army should receive heavy combat robots with cannon armament capable of independently conducting combat operations, that is, completely autonomous. Thus, no one has canceled the independent use of weapons.
Now - about hacking. At first glance, you can drive into the exchange of data, hack the on-board computer of the robot and take control. However, the success of this event is extremely doubtful. To begin with, it is necessary to penetrate the “jumping” or narrow communication channel, which is not in itself easy. If it succeeds, the maximum that a hacker gets is a set of discrete signals (how and what to convert them is a separate question).
The data will inevitably be encrypted, and each robot can have its own, unique key that can be changed very quickly (which, by the way, sharply limits the time during which the robot will be under the control of a hacker). Finally, there is an archiver that compresses the data before being sent over the communication channel - and it is not known which compression method is used.
However, let us assume that all these problems are solved. But even then, full access to the robot control will not be - a priori, it is impossible to replace all of its software in the shortest possible time. The maximum that can be done is to transmit an indication of a new goal, an order to retreat, or a signal about self-destruction. However, in the first case, the robot will first clarify whether the specified target is on its list of “its own”.
If it is listed, all CCs will be written off as “spam”. With the second and third teams, the robot will evaluate the tactical situation and figure out whether the new order is not a fake. In the obvious case, he will request additional confirmation. At the same time, assuming a fake, the onboard computer uses a different cryptographic key, and, possibly, a different data format - then the hacker will be out.
In general, the "interceptions" have the right to exist - but they will be difficult and expensive, and their results - limited. War is not banking, the level of “hacker” complexity is much higher here.
Consider an EMI question. The media is full of tales of an electromagnetic bomb that semi-literate terrorists can collect for $ 400. However, electromagnetic ammunition (EMBP) for $ 400 does not operate within a radius of a kilometer, and those that operate in a decent radius are not worth $ 400.
Cheap MHP ammunition (“piezoelectric frequency generator”) has a range of literally a few meters. They can, for example, "stun" the active protection system of a particular tankbut no "mass defeat" can be achieved. Powerful UVI munitions (UVI - “shock wave impulse”) have a radius of action within 1000 of their own calibers (150 mm = 150 m), but they are extremely expensive.
As a result, robotics are likely to be pulled into a thunderstorm except that the VEGG are explosive magnetic frequency generators. They are relatively cheap (but, naturally, much more expensive than conventional ammunition), but they have a range that is several times smaller than the UWI. If the robot is somehow protected from EMP, the affected area is compressed even more. In general, UGCH will indeed be quite an effective means of dealing with robots - but nothing supernatural in this respect is foreseen.
At the same time, “iron” is completely indifferent to chemical and bacteriological weapons, and much more resistant to incendiary. So, in reality, most of the problems allegedly inherent in combat robots are either successfully solved or have already been solved. In fact, the trouble came from where it was not expected.
The first real "mechanical infantryman" of the Pentagon was the SWORDS robot ("Swords"). Mass crawler robot controlled by remote operator kilometer was 45 kg, speed 6-7 km / h, autonomy -. 8,5 hours machine arming 5,56-mm machine gun M249 or 7,62-mm machine gun M240, anti-materiel rifle, 40-mm grenade or a shotgun. In the future, the possibility of placing a laser with a power of 100 kW was considered.
In 2006, SWORDS successfully passed tests at the Army Research Center Picatinny Arsenal, flashing, in particular, with the aforementioned accuracy. In the middle of 2006, the robots entered the troops, and in 2007, three SWORDS were deployed in Iraq. The results of their use were not advertised, but turned out to be encouraging enough that the military would issue an Foster-Miller order for 80 robots and decide to send several heavier vehicles to Iraq.
But then the Pentagon revised its plans - and very swiftly and without intelligible explanations. For a start, the military and developers have long referred to abstract technical difficulties, but soon bad rumors were leaked to the press. According to them, “Swords” lost control and opened fire without an operator’s command. Kevin Fahey, a US Army official responsible for implementing these robots, confirmed that the robots had lost control and did not succumb to the commands of the remote operator, but did not open fire, but only moved erratically. No man was hurt.
However, the mood of Fatah himself did not fit in with this statement - in his opinion, it is necessary to spend another 10-20 years to finalize the robots, and only then adopt them. Apparently, the matter was not limited to chaotic movement - at least robots pointed weapons at their soldiers.
As a result of the incident, SWORDS software had to be completely replaced. However, the "rebel" quickly found an alternative. As early as May 2008, Foster-Miller announced the delivery of the first MAARS combat robot to the US Army. Its technical differences from its predecessor - a modular layout, speed, increased to 12 km / h and a more mobile machine gun installation. A less noticeable innovation is the three-story control system, which allows you to avoid fire on your own. Some of its features are suggestive.
In general, the security system looks like this. First, the operator can set the boundaries of the zones in which fire is allowed and prohibited. This is a natural precaution - only now it suggests the possibility of ... independent fire without an operator’s command in the permitted zones.
Secondly, at MAARS there is a device that, at any position of the vehicle, turns its barrel away from the positions of American soldiers, which already looks like a suspicious reinsurance. Thirdly, attention! - there is a system that does not allow the machine to shoot into its own remote control unit. Obviously, according to these security measures, you can reconstruct a contingency situation that ended the SWORDS career.
By the way, cases of going out of control of automated complexes are not uncommon. The bloodiest incident of this kind occurred in South Africa, when a failure in the computer of an automatic anti-aircraft gun led to the death of an 9 man.
However, the stakes are too high for anyone to afford to abandon the development of “terminators”. As a result, by the year 2014, the US Army will receive the 1700 MAARS and the unpredictable number of other combat vehicles.