"Building killer robots is a very, very bad idea." Creating killer robots is a very, very bad idea Dmitry, mastermind and jack of all trades

Dmitry Melkin and Pavel and Boris Lonkin had no questions about who to take on the team to participate in robot battles. The guys knew each other from Baumanka, then together they assembled and installed solar-powered power plants. One day, Dmitry saw an announcement about a robotics competition and applied. Friends supported the initiative, and a month later the first combat robot of the Solarbot team, Brontosaurus, stood in the garage.

The first robot is lumpy

The Brontosaurus weighed a whole centner and, as its creators now admit, was not distinguished by either reliability or ingenious design solutions. No wonder: it was assembled partly on a whim, partly on fuzzy screenshots from videos from English Robot Wars competitions.

After Brontosaurus, having counted and rediding the main nodes several times, Dmitry, Boris and Pavel assembled their second robot. For its resemblance to a shell, it was called Shelby, from the English shell - “shell”. Shelby, the son of difficult mistakes, first defeated everyone at the “Battle of Robots – 2016” in Perm, organized by the Moscow Technological Institute (MIT) and Promobot, and then, together with the machines of two other Russian teams, became a participant in international competitions in China. How the winning robot works and what it cost to make it, its creators tell.

Dmitry, ideological inspirer and jack of all trades:

“Our great pride is the Shelby chassis. We fiddled with the running gear of its predecessor literally after each battle. When we made Shelby, the chassis was turned, sorted out and reassembled many times, but now you can forget about it altogether. In future projects, we will only have to work on maintaining reliability and increasing power. It would be nice, for example, if our new robot could move not one, but two enemy robots at once.”

Shelby's chains are from mopeds, the wheels are from a racing kart, and the electric motors are from radio-controlled models of cars. Parts for combat robots are not produced, so you have to look for them in flea markets and on the Internet. Good parts are very expensive, and designers tend to make them themselves.

Boris, designer, durable:

“Shelby is a type of flipper, “flipper”. It is equipped with a pneumatic system that pushes the lid up with force. This is the main weapon of the robot and its way of stabilizing: by tipping over, it can roll over with one jerk and stand on wheels. But we could not create high pressure in the pneumatic cylinder to make the impact of the cover powerful - there were no necessary valves. There was only one thing left: to make the system work as quickly as possible. The solution turned out to be simple: we got rid of excess hydraulic resistance and modified the factory valves. In the future, of course, a high pressure valve will be needed. Ready-made is expensive, about 200 thousand rubles, so now we are thinking about our own design.

Combat robots are not a cheap hobby: you need at least 200-300 thousand rubles plus consumables, spare wheels and everything that breaks down and is replaced in battle. And that's without taking into account the time and effort involved. “To assemble a robot, a team of three people needs to stop going to work for two months,” Solarbot engineers laugh. It will not be possible to save even on electronic stuffing.

Pavel, programmer:

“The main advantage of Shelby electronics is that there are very few of them. In order not to pick up a soldering iron after each fight, you need to provide the robot with the necessary minimum of “brains”. Shelby has simple factory controllers, and only the valves are controlled by a small board. It is very difficult to disable it. Even when in China, instead of the usual lead batteries, we were given powerful lithium batteries and the wires could not stand it after a couple of minutes, the robot’s electronics were not affected.”

Fighting robot Shelby

Speed ​​up to 25 km/h Effort on the pneumatic cylinder rod 2 t Engine power 2.2 kW Stock of pneumatic shocks without changing the cylinder 30−35 Remote control and its body is made only of a metal profile.

The Solarbot team has built a hardy iron soldier, but it also has a limit to its strength. In China, he suffered from the rotating knives of Chinese spinners, in Perm - from the claws of a matanga robot, which cuts a metal profile like butter with an eight-ton force. There are lacerated wounds on his iron ribs. The creators are preparing the fate of the exhibit for him: he will participate in festivals (summer Geek Picnic is in the near future), and a new fighter will replace him in the arena - also a flipper, only faster, more powerful and even more reliable. The lifting force of the lid will be twice that of the Shelby, the motor power will increase from 2.2 to 2.8 kW, and the speed will increase. With a new robot, the Russian team dreams of getting to Robot Wars in England.

But the future flipper is not the ultimate dream of Solarbot. Now Dmitry is negotiating with other teams and looking for sponsors: if everything goes well, then the first “megabot” will appear in Russia - as big and formidable as Japanese, American and Chinese multi-ton monsters.

Thanks to the support of the Moscow Institute of Technology, the Russians for the first time got to the international tournament of fighting robots FMB Championship 2017 in China. The fight was hosted by Shelby, Destructor from Kazan and Energy from St. Petersburg, which advanced to the semi-finals.

WikiHow is a wiki, which means that many of our articles are written by multiple authors. When creating this article, 14 people worked on editing and improving it, including anonymously.

Have you ever wanted to build a fighting robot? You probably thought it was too expensive and dangerous. However, most combat robot competitions have a 150 gram weight class, including RobotWars. This class is called "Antweight" in most countries and "FairyWeight" in the USA. They are much cheaper than large combat robots and not as dangerous. Therefore, they are ideal for those new to combat robots. This article will show you how to design and build an Antweight combat robot.

NOTE: This article assumes that you have already read and built a simple RC robot. If not, come back and at first make it. It should be noted that this article not is a recommendation to use a certain part of your robot. This is to encourage creativity and diversity among robots.

Steps

Understand the rules. Before designing a competition robot, you must understand all the rules. They can be found. The most important build rule to follow is the size/weight requirement (4"X4"X4" 150 grams), and the metal armor rule, which says you can't have armor more than 1mm thick.

What weapon will you use? An important part of a combat robot is a weapon. Come up with an idea for a weapon, but make sure you don't overstep the rules. For your first antweight bot, it's highly recommended to use "flipper" or even "pusher". A flip weapon, if properly designed, can be the most effective weapon in the Antweight class. The pushing weapon is the simplest, as it is not a moving weapon. The whole robot acts like a weapon and pushes the robots around. This is effective as the rules state that half of the arena must be without walls. You will be able to push another robot out of the arena.

Choose your details. Yes you need select your details before design. However, don't buy them. Bye. Just select the details and the corresponding project. If something doesn't fit or doesn't work while you're designing, you'll save money by being able to replace parts. And again, not buy parts now!

• Select a servo. It is generally recommended for Antweight beginners to use a servo instead of a motor, as with a servo you don't need a speed controller which saves you money and some weight for your robot. You should look for "micro" servos as they will save you a lot of weight. Make sure the servo is "360" modifiable. For combat robots, it is recommended to take a high torque servo instead of high speed, so that it is easier to push other robots even if you have different weapons. Servo can be bought
  • If you can't find the perfect servo for your needs, check out another section of the site that sells Futaba servos. Futaba is another servo brand. They are sometimes sized differently than the HiTec brand servos.
• Select a weapon motor. If you have an active weapon (not "pusher" for example), then you will probably need a motor to move the weapon. If you have a weapon that needs to move really fast (like a spinning weapon), then you should equip a DC motor (brushless usually works better, but brushes will work too) with a speed controller. It is not recommended to use spinning weapons for your first antweight mech as they are difficult to build and balance properly. However, if you want to make a flip weapon, then you will need a servo. It is recommended to purchase a micro servo with a particularly high torque so that it can flip another robot with ease. Another thing to look out for when choosing a servo for a weapon is the type of gears. If you are using nylon gears and the motor is heavily stressed, the gears can stretch over time. Try to choose more durable gears made of metal.
• Choose wheels. When choosing wheels, remember the rule that the robot must fit in a 4"X4"X4 cube. This means that your robot must have smaller diameter wheels. It is recommended to use wheels with a diameter of 2". Make sure the wheels can be easily mounted to the servo and protected. Another great technique used by combat robots of any size is the ability to ride upside down. Yes, the controls will be a bit reversed, but you can prevent losing the root contest. To do this, make your robot lower than your wheels so it can ride upside down. You can buy wheels
• Select a receiver/transmitter. When buying a receiver, make sure it is "fail safe". This is a mandatory rule in most competitions and safety. AR500 receiver not has this feature. You will need to purchase a BR6000 Bot Receiver, or another fail-safe receiver. The recommended transmitter is the SpektrumDX5e. If you built the remote controlled robot from the previous wikiHow article, you can use this transmitter again, but you will have to buy a new receiver.
• Select a battery. It is highly recommended to purchase a LiPo battery instead of a NiHM battery. LiPo batteries are lighter. However, they are more dangerous, expensive and require a special charger. Invest in a LiPo battery and charger to save on weight.
• Choose a material. The material from which the chassis and armor a combat robot is very important, as it protects your electrical components from being pierced by enemy weapons. There are three options to choose from: (note: there are more options, but these three are the most suitable for this weight class) aluminum, titanium and polycarbonate. Aluminum is light and strong, but can be expensive and hard to cut. Plus, he can be quite not 1mm in thickness. Titanium is light and very strong, but it is difficult to cut and very expensive. And the rule of 1 mm thickness also applies to it. Polycarbonate, or Lexan, is a lightweight, inexpensive, easy-to-cut, shatter-proof, durable plastic that is sometimes used in bulletproofing. Polycarbonate is also plastic, so it can be of any thickness, but it is recommended to use a thickness of 1 mm. The use of polycarbonate is highly recommended. It is as strong as the plastic used to make the walls of an antweight competition arena. When you buy, make sure you take a little extra in case you miscalculate. All of these materials can be purchased

1. Collect characteristics. Now that you have selected all the details, you need to remove the dimensions and weight. They must be listed on the website where you purchased them. Convert all inches to millimeters using a converter. Write down the specifications (in mm) of all your parts on a piece of paper. Now, convert the weight values ​​(ounces, pounds) to grams using the converter. Write down the weight specifications on paper.

   Design. You want the design to be as accurate as possible. This means that you should try to make a 3D design on a computer rather than a 2D design on paper. However, a 3D project doesn't have to look complicated. A simple project of prisms and cylinders will do.

   1. Add up the weight of all parts (in grams) and make sure the total is less than 150 grams.
   2. If you don't have CAD, download the free version of Sketchup.
   3. Learn the basics of Sketchup with free lessons.
   4. Create all the parts you will be using in Sketchup with the dimensions you noted down.
   5. Design your chassis and armor. Make sure to make it smaller than 4X4X4 inches.
   6. Place all components in the chassis/armor 3D model to see if they fit. This will help you decide where the components will be placed.

2. Order your details. If all your components match your design perfectly, order the parts. If not, choose new parts.

   Collect it. Now you need to assemble your chassis/armor. Put all your components in the places provided in your project. Connect everything and test. You should try to assemble everything so that you can easily remove the components if they need to be replaced. And the components will need to be replaced more often than a regular robot, as this robot will fight. Attacking robots can damage yours. It is recommended to use velcro tape (Velcro) to store the parts.

   Practice management. No matter how good your robot is, if you fall, you lose. Before you even think about competition, you necessary practice management. Use the upside down cups as cones and drive around them. Use the styrofoam as targets and attack him (try this on a small table to practice pushing and try not to fall yourself). You can even buy a cheap RC car (on a different frequency from your robot), have another person drive it, and try to push or destroy the car without falling off. If you know another person with an Antweight robot, have friendly duels with them (if possible, replace spinning weapons with less destructive plastic ones).

3. Compete. Find competitions in your area and have fun destroying other robots! Remember that if you are going to compete in the US, you should look for Fairyweight events, not Antweight events.

   • If you want your robot to be able to punch, it is advisable to attach a servo to a spherical "shoulder" and have the arm set at a 90 degree angle to do uppercuts.
   • Will your robot be more defensive or offensive? Since weight is limited, you may want to use more of it for weapons or armor. Try to balance these characteristics on your first robot.
   • Any robot can be improved. Just because your first robot model doesn't work, don't throw it out completely. Maybe you just need to replace the motor. Even if you have a fully functioning robot, you can still improve it. Look for motors that are more suitable for your purposes, if the new motor is not used in the project, just leave it and you will be able to build another robot. Try to upgrade some parts (usually front, back and weapons) of the armor to aluminum, or even titanium, for more "turntable protection".
   • Remember that you can place your robot diagonally in the cube.
   • Order spare parts for your robot. Since this is a combat robot, your parts may be damaged in battle. If you have spares on hand, you can replace parts faster.

   The rules say that the robot must fit in a 4X4X4 inch cube, however it can expand with remote control. You can benefit from this. For example, your flip weapon sticks out too much. Try to design it so that the flipper can go straight up and be less than four inches high. But when the flipper is lowered (after the cube is raised), the length will become more than four inches.

   • After building your first robot and having a clear understanding of combat robots, try to build another one. But this time be unique. Try to make it different from the robots of other people in this weight class. If you are really ambitious, you can try making a flying robot! Flying robots are allowed by the rules, but they are rarely built.
   • If you're using SketchUp, you can find the perfect models of servos and other components on Warehouse. Just look up the name of the servo (or component you want) and see if anything matches. Not everything is there, but what you find will usually look better and give you a cleaner model. Make sure the model you found is the same size as the real item.
   • If you are experienced in mechanics and combat robots, you can try building a walking robot. If you make a combat robot that walks, you get extra weight to work with.

   Warnings

   • LiPo batteries very dangerous. Not charge them using a NiHM or Nicad battery charger.
   • Even micropneumatics are dangerous. If you use pneumatics, follow the safety precautions.
   • Combat robots of even this size can be dangerous. If you are using a spinning weapon, move away when you operate it. Turn it off when working on weapons.
   • Always wear safety goggles when cutting material or operating the robot.
   • Some arenas are considered unsafe for spinning weapons. Don't try to use spinning weapons in these arenas.
   • LiPo batteries can ignite if punctured. When designing a robot, try to position the battery in a place that won't get pierced. If the battery catches fire, the rules say you not you can touch the robot while it is on fire. You will not be able to get it, which means all other components can be destroyed. Protect your battery like it's the heart of a robot!

A large gathering of scientists, industry leaders and NGOs have launched a campaign to stop the killer robots, dedicated to preventing the development of combat autonomous weapon systems. Among those who signed up were: Stephen Hawking, Noam Chomsky, Elon Musk and Steve Wozniak.

These big names are getting a lot of attention and giving legitimacy to the fact that killer robots, once considered science fiction, are actually fast approaching reality.

An interesting study published in the International Journal of Cultural Studies takes a different approach to the idea of ​​"killer robots" as a cultural concept. The researchers argue that even the most advanced robots are just machines, like everything else that has ever been done by mankind.

“The point is that ‘killer robot’ as an idea didn’t come out of thin air,” said co-author Tero Karppi, an assistant professor of media theory at the University at Buffalo. "This was preceded by methods and technologies to make the thinking and development of these systems possible."

In other words, we're worried about killer robots. The authors explore the theme of killer robots in films such as The Terminator or I, Robot, in which they suggested that far into the future, robots will eventually enslave the human race.

“Over the past decades, the increased use of unmanned weapons has dramatically changed warfare, bringing new humanitarian and legal challenges. There has now been rapid progress in technology, as a result of efforts to develop fully autonomous weapons. These robotic weapons will have the ability to select fire on a target on their own, without any human intervention."

The researchers respond that these alarmist dystopian scenarios reflect a "techno-deterministic" worldview, where technological systems are given too much autonomy, which can become destructive not only for society, but for the entire human race.

But what if we code machine intelligence in such a way that robots can't even tell the difference between a human and a machine? This is an intriguing idea: if there is no "us" and "them", there can be no "us against them".

Indeed, Karppi suggested that we may be able to control how future machines will think of humans on a fundamental level.

If we want to make changes in the development of these systems, now is the time. Simply ban lethal autonomous weapons and address the root causes of this dilemma. To really avoid the development of autonomous killing machines.

Ride on the unmanned "Yandex. Taxi" by Skolkovo, military engineers figured out how to adapt the technology of unmanned vehicles to create new weapons.

In fact, technology is not quite what it seems. The problem with all technological evolution is that the line between commercial robots "for life" and military killer robots is incredibly thin, and it costs nothing to cross. So far, they choose the route of movement, and tomorrow they will be able to choose which target to destroy.

This is not the first time in history that technological progress calls into question the very existence of mankind: first, scientists created chemical, biological and nuclear weapons, now - "autonomous weapons", that is, robots. The only difference is that, until now, weapons of "mass destruction" were considered inhuman - that is, they did not choose who to kill. Today, the perspective has changed: much more immoral seems to be a weapon that will kill with particular discrimination, choosing victims to its own taste. And if some militant power was stopped by the fact that, if it used biological weapons, everyone around would suffer, then with robots everything is more difficult - they can be programmed to destroy a specific group of objects.

In 1942, when the American writer Isaac Asimov formulated the three laws of robotics, it all seemed exciting, but completely unrealistic. These laws stated that a robot cannot and must not harm or kill a human. And they must unquestioningly obey the will of man, except in cases where his orders would be contrary to the above imperative. Now that autonomous weapons have become a reality and may well fall into the hands of terrorists, it turned out that the programmers somehow forgot to put Asimov's laws into their software. This means that robots can be dangerous, and no humane laws or principles can stop them.

A Pentagon-designed missile detects targets on its own thanks to software, artificial intelligence (AI) identifies targets for the British military, and Russia is showing off unmanned tanks. Colossal funds are spent on the development of robotic and autonomous military equipment in various countries, although few people want to see it in action. Just as most chemists and biologists are not interested in their discoveries eventually being used to create chemical or biological weapons, so most AI researchers are not interested in creating weapons based on them, because then a serious public outcry would damage their research programs.

In his speech at the start of the United Nations General Assembly in New York on September 25, Secretary-General António Guterres called AI technology a "global risk" along with climate change and rising income inequality: "Let's call a spade a spade," he said. “The prospect that machines will determine who lives is disgusting.” Guterres is probably the only one who can call on the military departments to change their minds: he previously dealt with conflicts in Libya, Yemen and Syria and served as the High Commissioner for Refugees.

The problem is that with the further development of technology, robots themselves will be able to decide who to kill. And if some countries have such technologies, while others do not, then uncompromising androids and drones will predetermine the outcome of a potential battle. All this contradicts all Asimov's laws at the same time. Alarmists may seriously worry that a self-learning neural network will get out of control and kill not only the enemy, but all people in general. However, the prospect of even quite obedient killer machines is not at all rosy.

The most active work in the field of artificial intelligence and machine learning today is not in the military, but in the civilian sphere - in universities and companies like Google and Facebook. But much of this technology can be adapted for military use. This means that a potential ban on research in this area will also affect civil developments.

In early October, the Stop the Killer Robots Campaign, a US non-governmental organization, sent a letter to the United Nations demanding that the development of autonomous weapons be internationally restricted. The UN signaled its support for the initiative, and in August 2017, Elon Musk and the United Nations International Conference on Artificial Intelligence (IJCAI) joined. But in fact, the US and Russia oppose such restrictions.

The last meeting of the 70 member countries of the Convention on Certain Conventional Weapons (on “inhumane” weapons) took place in Geneva in August. Diplomats failed to reach consensus on how a global AI policy could be implemented. Some countries (Argentina, Austria, Brazil, Chile, China, Egypt and Mexico) expressed support for a legislative ban on the development of robotic weapons, France and Germany proposed to introduce a voluntary system of such restrictions, but Russia, the USA, South Korea and Israel said they were not going to limit the research and development that is being done in this area. In September, Federica Mogherini, the European Union's top foreign and security policy official, said that weapons "affect our collective security" and that life and death must remain in the hands of the individual anyway.

Cold War 2018

US defense officials say the United States needs autonomous weapons to maintain its military advantage over China and Russia, which are also investing in similar research. In February 2018, Donald Trump demanded $686 billion for national defense in the next fiscal year. These costs have always been quite high and have only come down under the previous president, Barack Obama. However, Trump - unoriginally - argued the need to increase them by technological competition with Russia and China. In 2016, the Pentagon budgeted $18 billion for the development of autonomous weapons over three years. This is not much, but here you need to take into account one very important factor.

Most of the developments in the field of AI in the United States are carried out by commercial companies, so they are widely available and can be sold commercially to other countries. The Pentagon does not have a monopoly on advanced machine learning technologies. The American defense industry no longer conducts its own research in the way it did during the Cold War, but uses the achievements of start-ups from Silicon Valley, as well as Europe and Asia. At the same time, in Russia and China, such research is under the strict control of the defense departments, which, on the one hand, limits the influx of new ideas and the development of technologies, but, on the other hand, guarantees state funding and protection.

The New York Times estimates that military spending on autonomous military vehicles and unmanned aerial vehicles will exceed $120 billion over the next decade. This means that the discussion ultimately comes down not to whether to create autonomous weapons, but to what degree of independence to give them.

Fully autonomous weapons don't exist today, but Vice Chairman of the Joint Chiefs of Staff General Paul J. Selva of the Air Force said back in 2016 that in 10 years, the United States will have the technology to create such weapons that can decide who and when to kill. And as countries debate whether to limit AI or not, it may be too late.

Clearpath Robotics was founded six years ago by three college friends who shared a passion for making things. The company's 80 specialists are testing cross-country robots like the Husky, a four-wheeled robot used by the US Department of Defense. They also make drones and even built a Kingfisher robotic boat. However, one thing they will definitely never build: a robot that can kill.

Clearpath is the first and so far the only robot company to pledge not to build killer robots. The decision was made last year by co-founder and CTO Ryan Garipay and, in fact, even attracted experts to the company who liked Clearpath's unique ethical stance. The ethics of robot companies have come to the fore recently. You see, we are one foot in the future, in which there will be killer robots. And we are not yet ready for them.

Of course, there is still a long way to go. Korean Dodam systems, for example, is building an autonomous robotic turret called the Super aEgis II. It uses thermal imaging cameras and laser rangefinders to detect and attack targets up to 3 kilometers away. The US is also reportedly experimenting with autonomous missile systems.

Two steps away from the "terminators"

Military drones like the Predator are currently operated by humans, but Garipai says they will become fully automatic and autonomous very soon. And it worries him. Very. “Deadly autonomous weapon systems could roll off the assembly line now. But deadly weapon systems that will be made in accordance with ethical standards are not even in the plans.

For Garipai, the problem lies in international rights. There are always situations in war where the use of force seems necessary, but it can also endanger innocent bystanders. How to create killer robots that will make the right decisions in any situation? How can we determine for ourselves what the correct solution should be?

We are already seeing similar problems in the example of autonomous transport. Let's say a dog crosses the road. Should the robot car swerve so as not to hit the dog but put its passengers at risk? What if it's not a dog, but a child? Or a bus? Now imagine a war zone.

“We can't agree on how to write a manual for such a car,” says Garipai. “And now we also want to move to a system that should independently decide whether to use lethal force or not.”

Make cool things, not weapons

Peter Asaro has spent the last few years lobbying for a ban on killer robots in the international community, being the founder of the International Committee for the Control of Robotic Armies. He believes the time has come for "a clear international ban on their development and use." This will allow companies like Clearpath to keep doing great things without worrying that their products can be used to violate human rights and threaten civilians, he said.

Autonomous missiles are of interest to the military because they solve a tactical problem. When remote-controlled drones, for example, operate in combat, it is not uncommon for an adversary to jam the sensors or network connection so that the human operator cannot see what is happening or control the drone.

Garipai says that instead of developing missiles or drones that can decide on their own which target to attack, the military needs to spend money on better sensors and anti-jamming technologies.

“Why don't we take the investment that people would like to make to build autonomous killer robots and invest it in improving the efficiency of existing technologies? he says. “If we set a goal and overcome this barrier, we can make this technology work for the benefit of the people, not just the military.”

Recently, talk about the dangers of artificial intelligence has also become more frequent. worries that a runaway AI could destroy life as we know it. Last month, Musk donated $10 million to artificial intelligence research. One of the big questions about how AI is how it will merge with robotics. Some, like Baidu researcher Andrew Ng, worry that the coming AI revolution will take people out of their jobs. Others, like Garipai, fear that it could take lives.

Garipay hopes that his colleagues, scientists and machine builders, will think about what they are doing. Therefore, Clearpath Robotics took the side of the people. “While we as a company cannot bet $10 million on this, we can bet our reputation.”