The United States is already testing the ATLAS system: the tank itself will show the gunner who to shoot at

72
Source: sturgeonshouse.ipbhost.com

In general, a tank is such a contraption that needs automation very much, and the more it is, the higher its efficiency. For example, automatic gun loader, automatic fire control system, computerized engine control - all these elements can hardly be overestimated. They really help.

This is all understandable, and has not been considered novelty for a long time. The question here is different: what if the tank also independently searches for targets for firing and even prompts the gunner which projectile should be loaded into the cannon for this? Fiction? Looks like no.



The Americans have such a system not only on paper or in angry threats. She is already being rolled on tanks Abrams. It is called ATLAS, which is an abbreviation for Advanced Targeting and Lethality Aided System, or translated into Russian - "advanced targeting and lethality enhancement system."

And why is it needed at all?


Of course, the presence of a thermal imaging sight and a commander's multi-channel panoramic observation device is already a mandatory standard for any modern tank. These gizmos significantly increase the combat effectiveness of the vehicle, and there are no questions about this. But there is one catch here: no matter what modern sighting and observation devices and the fire control system as a whole are, no one has canceled the human factor.

Take the tank commander as an example. Despite the fact that the gunner should also be engaged in the search and identification of targets, it is on this crew member that the lion's share of the load falls in this process - that's what the all-round view is given to him. But in tense combat conditions, the commander has very little time for a thoughtful examination of the terrain, since, in addition, he needs to maintain contact with the commanding unit, navigate the terrain, correctly assess the situation, make certain decisions on further actions of the crew, and so on. So even a cool "panorama" with a good thermal imager will not completely eliminate errors - something "blurred eye" may miss.

The gunner has fewer responsibilities, but he is not immune from mistakes. Even experienced tankers, especially at long ranges, sometimes misidentify targets and don't even notice them. Objects through the electronic vision of a thermal imager do not always contrast strongly with the background and clearly show their outlines. You may not understand what is in front of you. Situations are rare, but they do happen.

Source: dzen.ru
Source: dzen.ru

In addition, after detecting and identifying the target, you need to perform a number of operations to prepare the shot: measure the range, select the type of projectile, set the appropriate ballistics, and so on. All this takes time, albeit a relatively short one.

Hence the conclusion that an electronic system capable of automatically searching for targets, correctly identifying them, measuring the range to them and choosing the right type of projectile would still be useful. And the crew will unload, and reduce the time to defeat, and get rid of mistakes to the maximum.

This system under the abbreviation ATLAS has been working in the USA for more than two years. Previously, they only talked about her, but now she is lit up on tanks.

Development of a system for searching and identifying targets


According to the plans, ATLAS should become a fully automated software and hardware system that will be installed on tanks and other armored vehicles with minimal alterations. Its capabilities, of course, in the future will make it possible to minimize the work of gunners and commanders in searching for and identifying targets, as well as to facilitate and speed up the preparation of a shot due to automatic range measurement, selection of projectiles for hitting the enemy and making corrections for firing in real time. All this, of course, due to neural networks.

Roughly speaking, this thing will continuously inspect the battlefield and, having noticed the enemy, it will identify it itself and issue recommendations to the gunner. In a very exaggerated version, it will look like this: hey, gunner, I found the target and determined it - this is a tank, so I already measured the range to it, load the sub-caliber projectile and just press the trigger.

The initiator of the work, represented by the Aberdeen Proving Ground, in his requirements also indicates that a tank with ATLAS will have time to hit three targets in the time spent on one target in manual mode.

The program, judging by the abundance of areas of work, the scale is far from just research. They were divided into five main groups, and they were allowed to deal with not only specialized companies that ate a dog in service, but also companies completely far from the “military” in whose portfolios of orders the US Army had never appeared before. In fact, this is one of the rare cases when the potential of hardened and giant defense corporations seemed insufficient. However, the US Army Weapons Center and the C5ISR Center are listed as the main performers. How many contractors they involved is not specified.

ATLAS tests on the Griffin I tank. The photo was taken from a modified M113 armored personnel carrier, from where the vehicle was remotely controlled. Source: c4isrnet.com
ATLAS tests on the Griffin I tank. The photo was taken from a modified M113 armored personnel carrier, from where the vehicle was remotely controlled. Source: c4isrnet.com

So what are the directions?

First of all, these are works with a digital image. As part of these activities, target tracking systems (automatic target tracking), their complete identification up to determining the model of a specific piece of equipment in the sight, passive ranging, mapping and accelerated decision-making by the on-board computer are being worked out.

Equally important is the work with machine learning. The "brains" of ATLAS teach algorithms for determining targets, deal with signatures by which the system will identify the enemy, and also make the product resistant to visual interference and adapt the software of the complex for use on existing processors used in military equipment.

Signature definition of enemy armored vehicles on the battlefield. The capabilities of the system allow you to identify the model of the combat vehicle in the sight. Source: covar.com
Signature definition of enemy armored vehicles on the battlefield. The capabilities of the system allow you to identify the model of the combat vehicle in the sight. Source: covar.com

The remaining three are fire control automation, integration into combat vehicles and the development of all kinds of sensors, from thermal imaging surveillance devices with sensitivity in different wavelength ranges to laser rangefinders and lidars.

It is worth noting that ATLAS will still be controlled by a person - a gunner and / or tank commander. Therefore, to clarify all the ergonomic points, the developers have already attracted several dozen tankers and other military specialties, who tested the controls of the system and gave their recommendations on how to refine it in order to make it more convenient to use.

Refinement of the ergonomic component of ATLAS. Source: nationaldefensemagazine.org
Refinement of the ergonomic component of ATLAS. Source: nationaldefensemagazine.org

Not only on paper


Work on this product went far beyond scientific research. Prototypes are already being actively tested, presented to the general public and even provided with some comments on their characteristics.

Initially, the experiments were carried out on the platform of an experimental Griffin I tank from General Dynamics. Its modified turret was fitted with a 50mm autoloader cannon and controlled by wire from a nearby M113 armored personnel carrier.

ATLAS tests on the Griffin I tank. Initially, two electronic observation and aiming modules were used, but the integration of the system into the tank's FCS makes it possible to get by with one. Source: c4isrnet.com
ATLAS tests on the Griffin I tank. Initially, two electronic observation and aiming modules were used, but the integration of the system into the tank's FCS makes it possible to get by with one. Source: c4isrnet.com

Now the turn has come to the Abrams, the pictures of which were recently published in the West.

Tank "Abrams" with the ATLAS system for testing in order to refine the product. Source: sturgeonshouse.ipbhost.comt.com
Tank "Abrams" with the ATLAS system for testing in order to refine the product. Source: sturgeonshouse.ipbhost.comt.com

The "eyes" of the system is an electronic module, somewhat reminiscent of a commander's panoramic sight, mounted on the roof of the tank in a mobile platform. This platform provides not only a 360-degree view, but also can rotate the module vertically for a more complete observation of the area. Cameras installed on all projections of the tank can also be used additionally, complementing the situational awareness of the crew and the ATLAS computer system.

In the module itself, based on rather scarce information, there are thermal imaging cameras susceptible to infrared waves of various lengths, a lidar and a laser rangefinder.

"Examining" the surrounding space on the battlefield in manual or automatic mode, the ATLAS thermal imager isolates objects that contrast with it from a general and generally homogeneous background. Accordingly, the slightest flashes of infrared radiation are immediately processed by "artificial intelligence" (neural network), which compares the contours of an object lit up in the sight with signatures sewn into memory. That is, having reference images of equipment and people, he determines what is in the sight - some kind of tank or infantry squad of the enemy, or even a pack of dogs running around.

If the computer nevertheless realized that there was a dangerous target in the lens of the thermal imaging camera, it informs the crew about this and remembers its coordinates in azimuth (the angle by which the electronic observation module was turned at the time the enemy was detected). Thus, ATLAS can store data on 3-4 targets, as they say, online.

In order to select a specific target from the list, the gunner (or commander) just needs to press the corresponding image on the touch screen. The system will immediately measure the range to it, turn the turret and issue a recommendation on the use of a particular projectile. For example, if a tank is caught, a message will appear stating that it is necessary to load a sub-caliber projectile, and if manpower, then a cumulative fragmentation or multifunctional high-explosive fragmentation projectile.

In general, a good help for the crew, which in the heat of battle is quite difficult to keep track of the situation in conditions of limited situational awareness.

Conclusions


Most publications about ATLAS in the Western media, and ours too, are based on the main slogan: destroy three times as many targets in the time it takes the crew to defeat only one. It sounds about the same as advertising some Internet provider tariff in the style of "speed up to 500 Mbps", where the main word is "to”, which means it could be less.

Yes, the speed of preparing a shot increases, but the main thing here is to reduce the neuro-physical load on the crew in terms of monitoring the battlefield, as well as reducing the number of errors in target identification. Of course, the neural network can have “bugs” and false positives, but they are successfully corrected by running new scenarios and signatures.

Of course, this is not a combat artificial intelligence. This is just an assistant to tankers, which expands the capabilities and improves the efficiency of the tank as a whole. And, apparently, the Americans are quite serious about the fact that they will be serially equipped with tanks and other combat vehicles. As they say, we'll see.
72 comments
Information
Dear reader, to leave comments on the publication, you must sign in.
  1. +6
    21 November 2022 05: 53
    The war of robots is getting closer and closer ... although the infantry is as it was, and is
    1. +4
      21 November 2022 06: 21
      Robots work hard, not a person ... why the hell is he needed in a tank ... it’s better to transfer him to UAV operators and let him press the remote control buttons there giving permission to shoot.
      I wonder if Atlas will be able to distinguish an inflated model of a tank from a real one.
      1. +1
        31 December 2022 20: 27
        Maybe. Inflator does not move. And cold.
        1. 0
          14 February 2023 09: 08
          Inflated has the same temperature as the real one.
    2. -6
      21 November 2022 06: 24
      Quote from Buyan
      The war of robots is getting closer and closer ... although the infantry is as it was, and is

      Well, the robots are good, but the Akhly have already escaped from captivity without drones.

      Ukrainian reconnaissance marines recently returned to the territory controlled by Ukraine, not on exchange, but after escaping from Russian captivity. This was announced by the deputy of the Rada Igor Kopytin in the social network.
      The command of the naval forces (Navy) and the marines decided to plan an operation to try to get the fighters out of captivity. As a result, it was possible to organize the escape of the scouts, and then transport them to the territory controlled by Ukraine.
      Azovstal defender Artyom Dyblenko, who also recently returned from captivity, took part in the operation to free the marines.


      link:
      https://www.pravda.ru/news/world/1772019-francija_peredala_zrk_ukraine/
      1. +4
        21 November 2022 07: 22
        It is logical ... electronics can "see" at the same time to the sides ... which surpasses human vision ...
        1. 0
          31 December 2022 20: 28
          But not the brain of a pelovek either. Namely, the brain "sees". The eye only supplies "data".
      2. +2
        21 November 2022 08: 23
        Lies, probably. We have military police over there so ̶e̶f̶f̶e̶k̶t̶n̶o̶ effectively "detains" the accused under Article 333 - it is doubtful that some miserable dill scouts could escape from such brave guys.
        1. 0
          22 November 2022 23: 35
          Quote: Al Manah
          Lies, probably. We have military police over there so ̶e̶f̶f̶e̶k̶t̶n̶o̶ effectively "detains" the accused under Article 333 - it is doubtful,

          And that 108 Bandera sadists were released on exchange, 11 of them were officers, is this also not true?
          1. 0
            23 November 2022 07: 53
            And what, is sarcasm really, really invisible? hi
      3. +3
        21 November 2022 09: 28
        Are you sure that this is not a fairy tale for the insane? Everything happens and fled from German captivity.
      4. The comment was deleted.
  2. 0
    21 November 2022 06: 00
    Cool! But some experts predicted the "decline" of tanks in modern warfare. One would like to believe that ours do not stand still and "Armata" will eventually become massive and one of the best tanks in the world.
    1. The comment was deleted.
      1. +1
        21 November 2022 06: 23
        Personally, I prefer materialism

        you'd better learn how to write articles on the merits, and not arrange another butthurt here on the topic "oh, the unfortunate mobilized."
      2. -1
        21 November 2022 07: 29
        Well, according to the bots, the claim is not in the mind. The radiant Americans with the British rubber shoe covers are also quite in use.
      3. +1
        21 November 2022 08: 20
        Personally, I would also prefer felt boots in winter. And he is any duffel bag - a sample of the millennium before last.
      4. -3
        21 November 2022 09: 07
        Did you serve at all? Or are you generally from Ukrainians, talking such crap?
      5. +7
        21 November 2022 10: 50
        I understand that this is off topic, but sometimes you need to respond to provocations. Circled in red boots, not even cast-offs at all, these are EVA warm boots that are very popular now.
      6. +1
        21 November 2022 13: 22
        Quote: Region-25.rus
        Many saw a video from the training ground ......

        Interestingly, and what are the claims to the sapper shovel (it is one of the underlined ones)? or should a fighter carry a tractor with him?
      7. 0
        21 November 2022 19: 59
        Well, specifically for the photo, I have a question - what is wrong with the sapper shovel. Digging is a standard, and even REQUIRED practice. But the rest is really no questions. A duffel bag of the 40s and felt boots / boots of the same years, this is already .... strange.

        Where exactly is the photo from? Definitely not from the front, our Armata will not be sent there. And if from a training point, then why train soldiers for work with Armata, if we use T-62, T-72 and T-80 at the front?
    2. -1
      23 November 2022 00: 52
      Quote: Vladimir61
      Cool! But some experts predicted the "decline" of tanks in modern warfare. One would like to believe that ours do not stand still and "Armata" will eventually become massive and one of the best tanks in the world.

      Of course, we will hit the enemy with Chubais's nanotechnologies. And let's add the optimism of the former head of Roscosmos, somewhere around 2050. :)
  3. Two
    0
    21 November 2022 06: 28
    hi And then everything described was a fantasy!
    1. +1
      21 November 2022 08: 45
      Quote: Dos
      hi And then everything described was a fantasy!

      Yes it was fantastic
  4. -1
    21 November 2022 06: 32
    The function is certainly useful, since it takes on part of the burden of monitoring the environment. True, with the current practice, when the tank has turned into a protected mobile firing point (often even firing from a closed position), this is not a fact that it will be in full demand
    1. +2
      21 November 2022 13: 31
      Quote: Ka-52
      not the fact that it will be fully demanded
      The device increases the awareness of the tank crew about the environment, and this is vital for the tank.
  5. +1
    21 November 2022 07: 55
    And how will this system behave if the sensors are damaged? Will the tanks retain full-fledged manual mode if some of the equipment is damaged?
  6. +3
    21 November 2022 08: 04
    a tank is such a thing that needs automation very much, for example, an automatic gun loader, automatic fire control system, computerized motor control Live and learn ! And I thought that ASUO is automated system ... and here, it turns out, "automatic"! How can I argue with literate people!
  7. 0
    21 November 2022 08: 13
    If they can bring it to mind, then the system will be a good helper for the crew. Pendos have a rubber budget and can afford expensive experimental projects.
    1. +5
      21 November 2022 08: 39
      Judging by the described, the system is not so expensive. It is only required to encourage good developers, and not to persecute them, as is customary with us.
      1. 0
        22 November 2022 01: 10
        In principle, it will not be expensive, because it is universal. Technical vision in a modern conflict is necessary for any technique. Moreover, optics with heat packs are already worth it. You just need to train the neural network.
  8. +3
    21 November 2022 09: 04
    Without such systems, a modern tank and military equipment are impossible in principle. In addition to detecting vehicles and tank-dangerous infantry, they can detect UAVs. Of course, this will greatly facilitate the work of the crew, but not enough. New tasks are added. The UAV must be part of the armament of the tank, as well as weapons against them. We need 4 crew members (driver, commander, gunner), system operator. In this regard, the most promising option is the new European tank.
    1. +1
      21 November 2022 19: 22
      Quote from cold wind
      Need 4 crew members (driver, commander, gunner), systems operator
      Well, if you dream, then under the armor you need to leave two commanders who will approve the actions of automation, and control, guidance, and so on are the task of computers.
      1. +1
        21 November 2022 19: 43
        Why dream? Look here. Two people in the car driver and commander. Can be controlled completely remotely.



        There are 2 different types of tanks.
        First: high-tech EMBT type, with various cannon weapons, KAZ, electronic warfare and air defense systems, built-in UAVs. Conditionally commander.
        Second: unmanned, relatively cheap and, accordingly, massive, highly specialized. Consumables of war.

        Another direction is self-propelled guns of the RCH 155 SPH type (pictured), which themselves can act as tanks.
  9. +3
    21 November 2022 09: 09
    There is still AI in terms of evaluating the image of the battlefield.
    To work effectively, it needs to be trained by uploading a large number of such images, preferably in video format.
    The direction is promising, in the long run on uninhabited combat vehicles, so we should expect assemblers of such a video on the tanks of the Armed Forces of Ukraine soon.
    This material cannot be replaced by anything, it is worth its weight in gold, so the Americans will drag out the war with all their might, with any external rhetoric.
  10. -6
    21 November 2022 12: 01
    I understand correctly that this electronic external module is an excellent target where to throw a sapper shovel laughing
  11. 0
    21 November 2022 12: 56
    Signature definition of enemy armored vehicles on the battlefield. The capabilities of the system allow you to identify the model of the combat vehicle in the sight. Source: covar.com


    As usual with Lohokid Martin, they "forgot" to add: in the ideal conditions of the test site, in a strictly defined perspective. Turn the tank sideways, turn its turret, place a source of heat signature behind it (an explosion funnel), and close the tank with a banal fence of their boards. And you will understand why fire-and-forget systems are still not the same as fire-and-hit systems.

    I have said and will repeat in all topics related to "artificial intelligence": everything that is being done now in this direction is an endless repetition of the schemes worked out back in the 70s. In an attempt to solve a methodological problem by brute force: more megahertz, more megabytes.
    And the problem here is not at all a lack of computing power. The problem is that no one really understands how a person recognizes objects. And without this understanding, it is impossible to create a reliable object recognition algorithm. Developers will wander over and over again in the vicious circle of formal analysis, which revolves around the banal "white spot" in the center of the frame. Yes, perhaps the developers of this system will teach their computer to recognize the Abrams tank in an open field. They can even teach him to distinguish it from the T-72. But what will they do when they “see” in the radio channel, for example, something like this?

    And to arrange this by placing corner reflectors on the tank is not so difficult. And if you give an order to the crews to place reflectors on their own, then EVERY tank will have an individual signature in the radio channel.
    Algorithms for formal image analysis are well known, and as soon as this system reaches the final implementation, the technique will acquire signature distortion systems. In the Navy, these methods were used back in the middle of the last century, painting ships.
    And they are, characteristically, much cheaper, if not an order of magnitude. I'm not talking about the fact that the real battlefield is replete with additional signal sources in both the thermal and radio channels.
    1. +4
      21 November 2022 19: 26
      Quote: abc_alex
      Turn the tank sideways
      The tank is garbage, let the ATGM operator be looked out for. And by the way, the tank is recognized sideways.
      Quote: abc_alex
      The problem is that no one really understands how a person recognizes objects. And without this understanding, it is impossible to create a reliable object recognition algorithm.
      No one now algorithmizes the tasks of pattern recognition, they are left to the mercy of neural networks. And neural networks do not program, but teach. Yes, the result is not ideal (they can mess up), but it allows you to solve the problem in most cases.
      1. 0
        22 November 2022 01: 10
        Quote: bk0010
        No one now algorithmizes the tasks of pattern recognition, they are left to the mercy of neural networks. And neural networks do not program, but teach. Yes, the result is not ideal (they can mess up), but it allows you to solve the problem in most cases.


        The other day I was testing a neural network trying to draw. So, this network is not capable of recreating the figure of a person, even from an anatomically correct sketch, without spawning extra limbs. Despite the fact that she is asked to draw a person with the text. This suggests that the network does not understand what exactly it draws, that is, it does not recognize the object, but only formally combines graphic primitives, tritely selecting cells in the "excel table" of correspondences. She does not learn in the literal sense of the word, she only fills many cells of this table with graphic primitives. All the same megabytes and megahertz. The hope of the creators is only that one day they will be able to hammer all possible options into this table, create connections between them and "THIS" will happen. But I repeat, just as some learn to recognize objects, others learn to mask them. This is the eternal struggle of sword and armor.
        1. +1
          22 November 2022 19: 33
          Quote: abc_alex
          This suggests that the network does not understand what exactly it draws.
          Yes.
          Quote: abc_alex
          but only formally combines graphic primitives, tritely choosing cells in the "excel table" of correspondences
          No
          Quote: abc_alex
          She does not learn in the literal sense of the word, she only fills many cells of this table with graphic primitives
          She's just learning. There is not a single graphic primitive in the neural network, only connections and coefficients.
          Quote: abc_alex
          All the same megabytes and megahertz
          This is important: they are still not enough for serious networks. Outstanding results are shown by networks on supercomputers, but they are stationary.
          1. 0
            24 November 2022 01: 37
            Quote: bk0010
            She's just learning. There is not a single graphic primitive in the neural network, only connections and coefficients.


            Connections between what and what? :) You can connect something with something. There is simply no connection at all.

            Quote: bk0010
            This is important: they are still not enough for serious networks. Outstanding results are shown by networks on supercomputers, but they are stationary.


            Of course of course. We need more megahertz and even more megabytes. The main thing is not to think, how does a person manage without all this? Yes, almost instantly. I think that without really understanding how our intelligence works and how neural connections function in our brain, building artificial intelligences and neural networks simply does not make sense, and neither megahertz nor megabytes will help.
            1. 0
              24 November 2022 20: 09
              Quote: abc_alex
              Connections between what and what?
              Between neurons
              Quote: abc_alex
              The main thing is not to think, how does a person manage without all this?
              A person cannot do without it, "it" is in his brain and in the peripheral nervous system, and in quantities many times greater than modern neural networks.
              Quote: abc_alex
              Yes, almost instantly.
              Far from instantaneous: in humans, the speed of signal propagation is 300 m/s (features of intercellular ionic conduction), in machines, the speed of light (at the peak, of course, potentially a million times higher).
              Quote: abc_alex
              I think that without really understanding how our intelligence works and how neural connections function in our brain, building artificial intelligences and neural networks simply does not make sense.
              It has: already now they have been assigned a lot of boring work or that which cannot be algorithmized with the required efficiency.
    2. 0
      21 November 2022 19: 29
      This system replaces the sailor-observer, only in the tank. The fifth member of the tank crew.
    3. +3
      22 November 2022 01: 24
      Fuck, you basically propose to score on the development and implementation of AI in the army? And then we wonder where the counter-battery radars are, where the strike UAVs are, where the principle of "fire and forget" ... The key factor is that AI is an assistant. Roughly speaking, you will simply have probable targets marked by the system on the screen, and whether this is a real target or not, the person will decide anyway. Well, besides, the neural network is trainable. Now, if they developed a completely autonomous tank that would itself find and destroy targets without a person, then I agree, as long as it is damp, although everything is going towards this, and it is criminal to miss this direction.
    4. +2
      22 November 2022 20: 29
      If no one understood how a person recognizes objects, then they would not be able to introduce a face recognition system for citizens in medical masks in the Moscow metro. Which is much more difficult than determining the presence of a tank or a soldier.
      This atlas will not work by itself.
      It is logical to assume that it will be integrated into a single data exchange system with other machines, UAVs, satellites, intelligence officers and other observers.
      If it is possible to establish an uninterrupted connection, then the calculations can be carried out on remote servers.
  12. 0
    21 November 2022 13: 06
    The idea is good, but the author did not understand it. Or publications of amers were cut. In fact, only half of the actually applicable algorithm has been published. But it's close... close to efficient equipment. Not like Rogozin's under-tankettes and androids.
    1. 0
      26 November 2022 22: 12
      The system has already been implemented on the F-35 strike fighter.
      But there is a powerful radar that scans ground objects.
      There is no such thing on the tank. Only optical devices.
      Therefore, it is more difficult for a computer (which is also weaker than an aircraft) to process the results.
  13. +1
    21 November 2022 13: 30
    A serious thing and an increase in efficiency, but I'm interested in how is it with the friendly fire errors of this device?
  14. rtv
    +1
    21 November 2022 13: 43
    Of course, the neural network can have “bugs” and false positives, but they are successfully corrected by running new scenarios and signatures.

    Well, you don't have to be naive. A neural network is a black box and no one in the world can explain why it produced a particular result. It can be very helpful in some cases, harmful in others, and until you come across these specific cases you may not know about them. There was a case when the neural network learned to recognize data by the background, roughly speaking, it did not recognize the number in the picture, but the background on which this figure is printed. And in the tests everything worked just amazing, but in practice it turned out to be a complete zero. A neural network is just a tool that you need to be able to use. And here the question arises - who will retrain these neural networks directly in the course of hostilities? How will I prepare data for training? On the range, everything looks completely different than in battle. So there are more questions than answers.
  15. 0
    21 November 2022 17: 47
    I remembered something about an aviation anecdote about a crow. If the automation fails, will the tank be combat-ready? Automation relaxes, it’s not always possible to intercept control “painlessly” if it fails.
  16. 0
    21 November 2022 18: 04
    A good way to destroy these bells and whistles is to remotely determine, aim and fire a guided projectile 10..15 kilometers away.
    Quote: Lech from Android.
    I wonder if Atlas will be able to distinguish an inflated model of a tank from a real one.

    If the copy is warmed up the same way. The same sections of the body. To the same temperatures as the real one.
  17. 0
    21 November 2022 20: 26
    [quote=abc_alex] [quote]
    The problem is that no one really understands how a person recognizes objects. And without this understanding, it is impossible to create a reliable object recognition algorithm. [/ Quote]
    Nevertheless, face recognition systems work great, and "recognize" the faces specified in the search, even partially covered, with makeup, and from different angles.
    It is possible that submillimeter waves will be used to recognize equipment on the battlefield. They do not shine through metal, and therefore any weapon will be naked for them, regardless of any disguise. It is much easier to set a reference image of a technique than a human face.
    The main problem now is in compact radiation sources.
    Receivers that render the picture are already there.
    1. +1
      22 November 2022 01: 37
      Quote from cpls22
      Nevertheless, face recognition systems work great, and "recognize" the faces specified in the search, even partially covered, with makeup, and from different angles.


      Because the principles of recognition are developed by professionals in the field of face recognition, and amateurs are engaged in disguise. If professionals of the appropriate qualification are engaged in camouflage, systems that reliably mask faces will appear very quickly. Remember, first AONs appeared on phones and then suddenly anti-AONs :)

      Quote from cpls22
      It is possible that submillimeter waves will be used to recognize equipment on the battlefield. They do not shine through metal, and therefore any weapon will be naked for them, regardless of any disguise. It is much easier to set a reference image of a technique than a human face.


      So it seems like this range is strongly absorbed by the atmosphere. Not? They write that it is completely absorbed by the atmosphere at a distance measured in meters, not even tens of meters ... :)
      Owl I repeat, the problem is not to copy the tank at the training ground. The problem is to correlate correctly a changing graphic image of an object, shaded and illuminated, partially covered or camouflaged in real combat conditions. Since the automaton does not understand what exactly it is looking for, but only analyzes the array of pixels in the center of the frame, it will not be able to understand that there is a burning barrel of rags in front of the tank. For him, it will be a single object. Do you understand? There is no recognition, which means there is no semantic filtering.

      Quote from cpls22
      The main problem now is in compact radiation sources.


      Again no. You or me or any other person do not need a radio channel, an IR signature, or a terahertz emitter to recognize a tank. We have enough information supplied in the optical range. And we can almost instantly recognize the Abrams, distinguish it from the T-72 or Merkava. Because we KNOW it, but do not recognize it. If someone can understand this recognition mechanism, no compact terahertz emitters will be needed, a high-resolution television camera with good optics will suffice. In the meantime, attempts will be made to reduce everything to a formal analysis, neither millimeter, nor kilometer, nor nanometer emitters will help. The problem is not in obtaining information, but in its analysis.
      I want to note that this does not mean at all that it makes no sense to develop machine vision systems. Including multispectral. This allows you to increase the amount of information received by a person. But so far, there is no other effective system of analysis, except for a person.
      1. 0
        31 December 2022 20: 42
        And you, Alex, are an expert in these matters! Write more, pls, we all read! wink
  18. -1
    21 November 2022 20: 55
    It is worth noting that ATLAS will still be controlled by a person - a gunner and / or tank commander.

    Until the software considers the actions of the human operator to be insufficiently effective.
    Or until a human changes the software.
  19. -1
    21 November 2022 21: 00
    Quote: rtv
    Of course, the neural network can have “bugs” and false positives, but they are successfully corrected by running new scenarios and signatures.

    Well, you don't have to be naive. A neural network is a black box and no one in the world can explain why it produced a particular result. It can be very helpful in some cases, harmful in others, and until you come across these specific cases you may not know about them. There was a case when the neural network learned to recognize data by the background, roughly speaking, it did not recognize the number in the picture, but the background on which this figure is printed. And in the tests everything worked just amazing, but in practice it turned out to be a complete zero. A neural network is just a tool that you need to be able to use. And here the question arises - who will retrain these neural networks directly in the course of hostilities? How will I prepare data for training? On the range, everything looks completely different than in battle. So there are more questions than answers.


    There is a lot of text, but I agree that automation based on trained AI is not yet perfect. And you can get oncoming fire from your automated trunks.
  20. 0
    21 November 2022 21: 12
    And then I thought, since this direction is promising, and the tanks are on average and so on. Isn't it easier to add an automatic loader to the tank, but leave the 4th crew member to them ?! His task will be to "observe" the situation. The fourth crew member will be responsible for monitoring the situation through systems around the entire circumference of the tank. Including he will be engaged in the control of the drone of the tank.

    As a result, the crew of 4 consists of:
    1. Driver - responsible for driving the tank and maintaining it in working condition
    2. Gunner's gunner - responsible for firing the cannon and maintaining the gun and loading mechanism in working condition
    3. Observer operator - responsible for surveillance systems (in the future, and electronic warfare systems), monitors the entire field around the tank, including terrain folds or obstacles (using drones), maintains all mechanisms in working order
    4. Commander - manages the entire team, has the ability, if necessary, to take control of any of the positions from his commanding position. It mainly fires and observes through the machine gun turret.

    Moreover, the difference between the observations of the "observer" and the commander is that:
    The commander observes everything within the line of sight. О especially carefully monitors dangerous directions
    So the "observer" not only covers the neighboring sectors of observation (to cover a larger area of ​​​​control). But it also conducts surveillance OUTSIDE OF THE LINE OF SIGHT with the help of a drone projectile or a conventional drone (behind the folds of the terrain, behind obstacles, or behind smoke / aerosol curtains) and conducts target designation for other crews or crew members. Moreover, the drone can be equipped with an ultraviolet laser, so that the commander and gunner, looking at the battlefield where the drone flies, see the beam on the screen through the cameras. And the enemy, with his own eyes, would not see any beam. This is a little less productive than a full-fledged laser control unit, but it’s cheaper, and therefore the drone itself is cheaper and massive.

    Moreover, the launch of the drone can be organized through shooting from a special mortar. The drone, in the folded state, is stored next to the workplace of the Operator-observer. As soon as the need arises to obtain aerial reconnaissance. The operator puts a "drone projectile" (hereinafter referred to as DS) into the breech and launches. The DS flies out through a separate channel and itself in flight opens the propellers, wings / plumage and starts up. The operator on his screen begins to see the picture from the drone and gets the opportunity to control it. At what to manage both directly through the sidestick, and set the program for automatic patrolling of the area on the map or along the route.

    In this case, as an option, it is possible to organize a system with a hybrid-combined drone power supply. You can just put the DS in the mortar and shoot. Then he, like a regular drone, flies around the tank, or even flies away from him on command or under direct control. But the time of his work in this form is limited, and as soon as the charge comes to an end. The drone in manual mode, or automatically lands or flies to the tank and lands on the prepared net at the stern of the tank, where, clinging to the net, it lies until after the battle (or at the moment of respite), someone from the crew takes and puts the drone inside where it is can be charged. At the same time, with manual control, you can get used to sending the drone to fly up to the hatch of the commander or gunner of the tank, and he would stretch out his hand (not even the whole body, but only his hand) would simply grab it, put it inside and close the hatch behind him.

    And the second option is when the operator quickly attaches an additional "module" to the DS from the back. And in this form, he puts it inside the mortar. The DS is fired again, but this time after unfolding and launching, it automatically turns on but does not fly far, but hangs in the air. Because now it is fastened to the tank with a long power cord. In this case, the drone will not be able to fly far (the length of the cord will be limited), but it can fly around the tank indefinitely (Since it is powered by the tank’s power plant, although it doesn’t fly far now) until either the enemy does something to it, or the crew . In this case, if necessary, on command, the DS is disconnected from the cord and starts flying on batteries as in the first case. And as soon as the drone flies off a few meters, the module that held the DS on the "leash" is automatically fired from the mortar.

    Moreover, each DS can be equipped with a radio beacon for target designation. Suppose a tank has discovered a well-fortified enemy pillbox, which you cannot take with a tank gun. Then the operator transmits the data of the DS direction finder to the gunners via a communication channel, and puts this DS on the roof of the bunker. Artillerymen either launch a rocket into the square, which, flying up, aims at the DS pelleting and flies right into the DS. Or it shoots at a given area with large-caliber artillery shells, which themselves are accurately brought to the DS bearing data, which shines on the roof of the bunker. Or, for example, a column of enemy armored vehicles was found that is going somewhere. Then the drone, turning on its direction finder again, flies OVER the column, so that the same missiles / shells would hit directly into the column.
    1. +2
      21 November 2022 21: 36
      What you wrote is called EMBT, photo above. (The KF-51 also has an operator)
      Hull from Leopard 2, heavily redesigned turret from Leclerc.
      Immediately the ability to put a 120 mm or 140 mm gun, tested on Leclerc, respectively, under 130 mm needs to be completed.
      It costs AZ, but added 4 crew members, an operator of on-board systems (turrets with a 30mm cannon, other reconnaissance equipment). The driver and operator are together in front, the commander and gunner are in the hull below the turret.
      You need to add a block with 4 Hero-120 UAVs a la KF-51 (there is also a quadric). They can be returned by parachute or hit targets. The operator if something runs after them.
      1. 0
        21 November 2022 22: 40
        So here it is, a promising direction. Of course, a 30 mm gun is not a fact that is needed.
        At what it is not needed not in general for tanks, but specifically for this one. Because of the ability to put 140 mm. gun, it becomes clear that the tank is already very heavy. I think 30 mm. autocannon when mounted 140 mm. guns will only interfere with their weight.

        But I liked the performance. Although not perfect. The execution of the tower itself is unsuccessful, there are many weakly protected and uncovered zones. But they are already working on the direction itself. We should hurry up. At least try to conduct research on the difference in the quality of crews with the 4th member of the team as an "operator".
        For example, take several crews and divide them into 2 teams. There will be 2 teams in which, well, let's say there will be 5 crews on 5 cars. Teams, purely because we initially study the effectiveness of the composition and develop strategies, are put on the BMP 3.
        Teams as I said 2. One team is "red", they will go in the standard composition of the commander, driver mechanic and gunner (so as not to remake more fluid and expensive tanks). And the "blue" team will be equipped with equipment for monitoring and controlling drones, as well as a 4th crew member.

        Then 2 teams run through the same tests at the training ground - 1 round.
        And at the end, "battles" between the teams are played out. A few 1v1 duels, a few duos, then a team-on-team, and a 1vs-multiple battle. But on the basis of the results obtained, theorists are sent data and they work out the statistics in a form that is convenient for understanding. And the crews of the "blue" together with the data obtained by theorists in the company with a couple of staff officers. Based on the received data. develop a preliminary program for the strategy and techniques of such crews in battle. The "blue" team is working on new techniques and techniques, and the "red" team is training based on their experience.
        And then they do round 2.
        They do the same thing as before (tests, races and shooting at the range, and then friendly matches), but with the use of created techniques and techniques. The statistics are collected again and compared with the first one.
        Based on the difference in statistics, the details of the further development of the ideas of 4-crew tanks or completely new BMs are being worked out.

        At what for tests. While you can use BMP 2 or 3. Why?
        Like a tank, it is a tracked vehicle.
        In an unused infantry section, there is enough space to accommodate the 4th crew member and equipment for him
        We have a lot of infantry fighting vehicles themselves, and it will not be difficult for research to allocate 8-10 vehicles for this.
        On the BMP 2/3 chassis, you can install a module with a cannon and a machine gun in order to more accurately simulate the work of the gunner and tank commander
        1. 0
          21 November 2022 22: 50
          The M-230 weighs 59 kg (KPVT 52 kg), respectively, there are no problems with its installation and the installation of a 120/130/140 mm gun does not depend on it. Questions about the price, the need and the number of shells. The air defense capabilities and work against tank-dangerous infantry of the 30 mm gun are extremely high, they are not comparable with machine guns. For this gun, shells were made with a radio proximity sensor to the drone.
          On EMBT there is a European analogue, on AbramsX directly M230.


          EMBT has extremely high firepower.
          Main caliber - 120/140 mm gun
          Optional - 30mm autocannon on separate turret
          Coaxial with the main gun 12,7 mm machine gun
          Coaxial 7,62 mm machine gun with a panoramic sight.
          Three pairs of eyes can conduct reconnaissance and independently fire.
          1. +1
            21 November 2022 23: 56
            Then all the more an excellent direction. The main problem is probably only in the final price.
          2. 0
            31 December 2022 19: 29
            The main burden again falls on the one who must press the button and record a victory for himself, or do they divide the booty into three? Who will put the horns on the wall? Take turns, or play cards? So, as there are four hooves, then one nails the horns to the wall, and two more nail the hooves. Each one in front and one back, ears on jelly for everyone. And if it's her and not him, how will they share it?
    2. +1
      22 November 2022 01: 48
      Quote: Mustachioed Kok
      3. Observer operator - responsible for surveillance systems (in the future, and electronic warfare systems), monitors the entire field around the tank, including terrain folds or obstacles (using drones), maintains all mechanisms in working order


      Everything would be fine, but here's the problem. A person is not able to track several directions at the same time. We are naturally adapted to focus attention in a rather narrow sector of observation. And a tank in modern combat can be attacked from almost any azimuth. This is why automatic recognition systems are needed - to look in all directions at once and, if necessary, for example, turn the gun in the right direction, issuing a request for a shot.
      And one person cannot control drones either. Can fly a drone. One. We cannot psycho-physiologically analyze two streams of visual information at the same time. We are a single channel system. Albeit multispectral. Within this one channel, we are almost perfect. But if you need multi-channel - alas. We cannot focus attention to the left and right at the same time :) Therefore, the operator will not be able to control the entire field around the tank.
      1. 0
        22 November 2022 09: 11
        Well, first of all, he is not the only one who conducts surveillance in the tank. There is also a commander. As well as a mechanic, a driver and a shooter that are observing in their directions.
        Secondly, no one canceled the automatic recognition system. The "operator" conducts direct observation or patrol of a remote area on the orders of the commander.
        1. 0
          24 November 2022 01: 56
          You yourself write:

          is observing the whole field around tank


          I'm saying he's incapable of doing it.
          Now you're talking
          Quote: Mustachioed Kok
          There is also a commander. As well as a mechanic driver and shooter


          Well, firstly, the driver has something to do without it. Like an arrow. Secondly, if you entrust the control of the "look" of the operator to the tank commander, then this is just shifting the problem from the left hand to the right. Since now I will say about the tank commander that he is not able to control the entire battlefield. :)
          When they gave the task for the BMPT, that is why they demanded TWO independent weapons operators at TWO firing points. To control the left and right hemisphere.
          What am I for? Moreover, the allocation of ONE person to control the entire battlefield does not make sense. It is not enough.

          Quote: Mustachioed Kok
          "Operator" conducts direct observation or patrol of a remote area by order of the commander.


          And here we immediately remember that our commander is also a person and he is also not all-perspective. :)
          If we do not want to increase the crew to the number of BMPT-1 (6 people), then we need to follow a different path. The system needed focusing a person's attention. For example, recognizing a large vehicle with a gun at azimuth 97, it will inform the commander of a "dangerous object on the starboard side" and turn the tank's turret to this azimuth, bringing the target into the commander's or gunner's zone of attention.

          Perhaps there would be a need for a 4th crew member if our developers gave the tank a tethered quadcopter-UAV for observation.
  21. +1
    22 November 2022 07: 19
    The next war will be a war of hackers breaking into other people's defense servers.
  22. 0
    22 November 2022 08: 10
    More and more external devices are being hung on the tank behind the armor, which are blown away by fragments from the nearest explosion and then, in the old fashioned way, if they managed to train.
  23. 0
    22 November 2022 09: 31
    And we don’t even have normal thermal imaging sights on all tanks .... well, so that there are two separate observation and aiming channels ....
  24. 0
    22 November 2022 09: 36
    So it seems like this range is strongly absorbed by the atmosphere. Not? They write that it is completely absorbed by the atmosphere at a distance measured in meters, not even tens of meters ... :)
    Hard to believe. Absorption is proportional to the decrease in wavelength. UV - is extinguished by the atmosphere, the visible range - by precipitation, near IR - overcomes precipitation, far IR - penetrates through thin barriers, terahertz - shines through organic matter.
    Somewhere I saw a satellite image of the coastal zone - ships, reinforced concrete buildings are clearly visible there, in which massive metal objects look like canaries in cages.
    What disguise would you oppose to this?
    Carry barrels in front of you? Visual perception of a person is based on many patterns (a kind of cliches) associated with many other information arrays, including through the speech areas of the brain. Therefore, a person can see a white-maned horse in a cloud. But for a highly specialized combat system, this is not necessary.
    It is enough to clearly recognize at least fragments of basic 3D objects.
    For individuals, such a system already exists, and experts tried to deceive it.
    The main problem of practical application is the inability to obtain a reference 3D image of the wanted person. In the case of technology, this is easily overcome.
  25. 0
    22 November 2022 16: 27
    Quote: Timur_kz
    You just need to train the neural network.


    The system itself, IMHO, is inexpensive, but with the "training" of the neural network, I believe, there will be the main difficulty.
  26. 0
    22 November 2022 17: 10
    now there are a lot of people with registrars running and driving. They will collect videos, put a hundred people in jail - to recognize and mark equipment, and then they will upload these markers to the neural network. Long, tedious, not without pitfalls, but BigData will do its job. The main thing is to collect it, so that the Americans will stretch this story as long as they can. They need a smoldering war.
  27. 0
    23 November 2022 22: 57
    Anti-radar missile, etc. With such an arsenal as a light bulb on the battlefield.
  28. 0
    31 December 2022 19: 23
    "Despite the fact that the gunner should also be engaged in the search and identification of targets, it is on this crew member that the lion's share of the load falls in this process - that's what the all-round view is given to him." And if this is a gunner, then on which organ does the lion's share of the load fall?
  29. DO
    0
    8 February 2023 02: 17
    And, apparently, the Americans are quite serious about the fact that they will be serially equipped with tanks and other combat vehicles. As the saying goes, will see.

    The fruits of "we will look" at the "drone toys" in previous years, we are fully tasting now, having to buy them for the NWO by import.