When many people hear the three words "robot", words such as "cool appearance", "powerful" and "high-end" will appear in their minds. They think that robots are as high-end as the "terminator" in science fiction movies. Cool. In fact, in this article, we will explore the basic concepts of robotics and understand how robots accomplish their tasks.
1. The components of the robot
From the most basic level, the human body includes five main components:
- body structure
- Muscle system, used to move body structures
- Sensory system to receive information about the body and surrounding environment
- Energy source, used to provide energy to the muscles and senses
- The brain system, used to process sensory information and direct muscle movement
Of course, humans have some intangible characteristics, such as intelligence and morality, but on a purely physical level, this list is quite complete.
The components of a robot are very similar to humans. A typical robot has a movable body structure, a motor-like device, a sensing system, a power source, and a computer "brain" that controls all these elements. In essence, robots are "animals" made by humans. They are machines that mimic the behavior of humans and animals.
Bionic Kangaroo Robot
The definition of a robot is very wide, ranging from industrial robots serving factories to small household cleaning robots. According to the current broadest definition, if something is considered a robot by many people, then it is a robot. Many roboticists (people who make robots) use a more precise definition. They stipulate that the robot should have a reprogrammable brain (a computer) to move the body.
According to this definition, the difference between robots and other mobile machines (such as cars) lies in their computer elements. Many new cars have an onboard computer, but only use it to make minor adjustments. The driver directly controls most parts of the vehicle through various mechanical devices. Robots are different from ordinary computers in terms of physical characteristics. They are each connected to a body, while ordinary computers are not.
Most robots do share some common characteristics
First, almost all robots have a body that can move. Some have only motorized wheels, while others have a large number of movable parts, which are generally made of metal or plastic. Similar to human bones, these independent parts are connected by joints.
The wheels and shafts of the robot are connected by some kind of transmission. Some robots use motors and solenoids as transmission devices; others use hydraulic systems; still others use pneumatic systems (systems driven by compressed gas). The robot can use any of the above-mentioned transmission devices.
Second, the robot needs an energy source to drive these transmissions. Most robots will be powered by batteries or wall outlets. In addition, hydraulic robots require a pump to pressurize the liquid, while pneumatic robots require a gas compressor or compressed gas tank.
All transmission devices are connected to a circuit by wires. This circuit directly supplies power to the electric motor and solenoid, and operates the electronic valve to start the hydraulic system. Valves can control the path of fluid under pressure in the machine. For example, if the robot wants to move a leg driven by hydraulic pressure, its controller will open a valve, which leads from a hydraulic pump to the piston barrel on the leg. The pressurized fluid will push the piston and rotate the leg forward. Generally, robots use pistons that can provide two-way thrust so that parts can move in two directions.
The robot's computer can control all the components connected to the circuit. In order to make the robot move, the computer will turn on all the necessary motors and valves. Most robots are reprogrammable. If you want to change the behavior of a robot, you only need to write a new program into its computer.
Not all robots have sensor systems. Few robots have vision, hearing, smell or taste. One of the most common senses that a robot has is the sense of movement, which is its ability to monitor its own movement. In the standard design, wheels with grooves are installed at the joints of the robot. There is a light-emitting diode on one side of the wheel, which emits a beam of light that passes through the groove and shines on the light sensor on the other side of the wheel. When the robot moves a specific joint, the grooved wheels will turn. During this process, the groove will block the beam. The optical sensor reads the flashing pattern of the beam and transmits the data to the computer. The computer can accurately calculate the distance the joint has rotated based on this model. The basic system used in a computer mouse is the same.
These are the basic components of the robot. Robot experts have countless ways to combine these elements to create infinitely complex robots. The robotic arm is one of the most common designs.
2. How the robot works
The term "Robot" in English comes from the Czech word robota, which is usually translated as "forced laborer". It is very appropriate to describe most robots. Most of the robots in the world are used for heavy repetitive manufacturing tasks. They are responsible for tasks that are very difficult, dangerous or boring for humans.
The most common manufacturing robot is a robotic arm. A typical robotic arm consists of seven metal parts connected by six joints. The computer will rotate a stepper motor connected to each joint to control the robot (some large robotic arms use hydraulic or pneumatic systems). Unlike ordinary motors, stepping motors move precisely in increments. This allows the computer to accurately move the robotic arm so that the robotic arm repeats the exact same actions continuously. The robot uses motion sensors to ensure that it moves exactly the correct amount.
This six-joint industrial robot is very similar to a human arm. It has parts equivalent to shoulders, elbows and wrists. Its "shoulder" is usually mounted on a fixed base structure (rather than a moving body). This type of robot has six degrees of freedom, that is, it can rotate in six different directions. In contrast, the human arm has seven degrees of freedom.
The joints of a six-axis industrial robot
The role of the human arm is to move the hand to different positions. Similarly, the role of the robotic arm is to move the end effector. You can install various end effectors suitable for specific application scenarios on the robot arm. There is a common end effector that can grasp and move different objects. It is a simplified version of a human hand. Robotic hands often have built-in pressure sensors to tell the computer how hard the robot grasps a particular object. This prevents objects in the hands of the robot from falling or being crushed. Other end effectors include blowtorches, drill bits and paint sprayers.
Industrial robots are designed to perform exactly the same tasks repeatedly in a controlled environment. For example, a robot might be responsible for screwing the lid on a peanut butter jar that is transported on an assembly line. In order to teach the robot how to do this work, the programmer will use a handheld controller to guide the robotic arm to complete the entire set of actions. The robot accurately stores the sequence of actions in the memory. After that, whenever a new can is transferred from the assembly box, it will do this set of actions repeatedly.
Robot arm is one of the basic parts used in manufacturing cars
Most industrial robots work on car assembly lines and are responsible for assembling cars. In carrying out a large number of such tasks, robots are much more efficient than humans because they are very precise. No matter how many hours they have been working, they can still drill holes in the same position and tighten screws with the same force. Manufacturing robots also play a very important role in the computer industry. Their extremely precise hands can assemble a very small microchip.
The manufacturing and programming of robotic arms is relatively low because they only work in a limited area. If you want to send robots to the vast outside world, things become a bit complicated.
The first problem is to provide a viable motion system for the robot. If the robot only needs to move on flat ground, wheels or tracks are often the best choice. If the wheels and tracks are wide enough, they are also suitable for rough terrain. But robot designers often want to use leg-like structures because they are more adaptable. Making legged robots also helps researchers understand the knowledge of natural kinematics, which is a useful practice in the field of biological research.
The legs of the robot are usually driven back and forth by hydraulic or pneumatic pistons. The pistons are connected to different leg parts, like muscles attached to different bones. It is undoubtedly a difficult problem to make all these pistons work together in the right way. In infancy, the human brain must figure out which muscles need to contract at the same time to prevent falling when walking upright. In the same way, the designer of the robot must figure out the correct combination of piston motion related to walking and program this information into the robot's computer. Many mobile robots have a built-in balance system (such as a set of gyroscopes) that tells the computer when to correct the robot's movements.
Boston Dynamics' latest upgraded Atlas humanoid robot
The motion mode of biped walking is inherently unstable, so it is extremely difficult to realize in the manufacture of robots. In order to design robots that walk more steadily, designers often turn their eyes to the animal kingdom, especially insects. Insects have six legs, they often have extraordinary balance abilities and can adapt to many different terrains.
Some mobile robots are remotely controlled, and humans can direct them to perform specific tasks at specific times. The remote control device can communicate with the robot using cables, radio or infrared signals. Long-range robots are often referred to as puppet robots, and they are very useful when exploring environments that are full of danger or inaccessible to humans, such as the deep ocean or the interior of a volcano. Some robots are only partially remote controlled. For example, the operator may instruct the robot to reach a certain place, but will not guide it, but will let it find its own way.
NASA develops remotely controllable space robot R2
Automatic robots can move autonomously without relying on any controller. The basic principle is to program the robot to respond to external stimuli in some way. An extremely simple collision reaction robot can interpret this principle well.
This robot has a collision sensor that is used to check obstacles. When you start the robot, it generally moves along a straight line zigzag. When it hits an obstacle, the impact force will act on its collision sensor. Every time a collision occurs, the robot's program will instruct it to back up, turn to the right, and then continue forward. According to this method, the robot will change its direction whenever it encounters an obstacle.
Advanced robots will use this principle in a more sophisticated way. Robot experts will develop new programs and sensing systems to create robots with higher intelligence and greater perception. Today's robots can show their talents in various environments.
Simpler mobile robots use infrared or ultrasonic sensors to sense obstacles. These sensors work similarly to animal echolocation systems: the robot emits a sound signal (or a beam of infrared light) and detects the reflection of the signal. The robot calculates the distance between it and the obstacle based on the time it takes for the signal to reflect.
Advanced robots use stereo vision to observe the world around them. Two cameras can provide depth perception for the robot, and image recognition software enables the robot to determine the location of objects and recognize various objects. The robot can also use microphones and odor sensors to analyze the surrounding environment.
Some autonomous robots can only work in the limited environment they are familiar with. For example, lawn mower robots rely on buried landmarks to determine the range of pasture. The robot used to clean the office requires a map of the building to move between different locations.
Higher-level robots can analyze and adapt to unfamiliar environments, even in areas with rugged terrain. These robots can associate specific terrain patterns with specific actions. For example, a rover robot uses its vision sensor to generate a map of the ground ahead. If the map shows a rough terrain pattern, the robot will know that it should take another path. This system is very useful for exploring robots working on other planets.
There is an alternative robot design scheme that uses a looser structure and introduces randomization factors. When this kind of robot gets stuck, it will move its appendages in all directions until its actions have an effect. It uses force sensors and transmission devices to work closely together to complete tasks, rather than being guided by a computer through programs. This is similar to when an ant tries to bypass an obstacle: the ant does not seem to act decisively when it needs to pass an obstacle, but keeps trying various methods until it bypasses the obstacle.
Three, home-made robots
In the last few parts of this article, we take a look at the most compelling areas in the robotics world: artificial intelligence and research robotics. Over the years, experts in these fields have made great progress in robotics, but they are not the only manufacturers of robots. For decades, although few people have taken this as a hobby, they are enthusiastic. They have been making robots in garages and basements all over the world.
Home-made robots are a rapidly developing subculture and have considerable influence on the Internet. Hobbyists use a variety of commercial robot tools, mail-order parts, toys, and even vintage video recorders to assemble their own works.
Like professional robots, there are many types of home-made robots. Some robot enthusiasts who can only work on weekends have created very sophisticated walking machines, while others have designed domestic robots for themselves, and some enthusiasts are keen on making competitive robots. Among competitive robots, people are most familiar with remote-controlled robot fighters, as you see in the "BattleBots" program. These machines are not "real robots" because they have no reprogrammable computer brains. They are just enhanced remote control cars.
More advanced competitive robots are controlled by computers. For example, a football robot does not need human input at all when playing a small football game. The standard robotic soccer team consists of several individual robots that communicate with a central computer. This computer uses a camera to "observe" the entire stadium, and distinguish the football, the goal, and the players of one's side and the opponent's side according to the color. The computer processes such information at all times and decides how to command its team.
Adaptability and versatility
The personal computer revolution is marked by its excellent adaptability. Standardized hardware and programming languages allow computer engineers and amateur programmers to build computers for their specific purposes. Computer parts are somewhat similar to craft supplies, and their uses are countless.
Most robots to date are more like kitchen appliances. Robot experts make them for specific purposes. But they are not very adaptable to completely different application scenarios.
This situation is changing. A company called Evolution Robotics created a precedent in the field of adaptive robot software and hardware. The company hopes to open up its own niche market with an easy-to-use "robot developer toolkit."
This toolkit has an open software platform that provides various commonly used robot functions. For example, roboticists can easily give their work the ability to track targets, listen to voice commands, and bypass obstacles. From a technical point of view, these functions are not revolutionary, but what is unusual is that they are integrated in a simple software package.
This kit also comes with some common robot hardware, which can be easily combined with software. The standard kit provides some infrared sensors, motors, a microphone and a camera. Robot experts can use a set of reinforced mounting kits to assemble all these parts, this kit includes some aluminum body parts and durable wheels.
Of course, this kit is not for you to make mediocre works. It costs more than US$700 and is by no means a cheap toy. However, it is a big step towards a new type of robotic science. In the not-too-distant future, if you want to build a new type of robot that can clean the room or take care of your pets when you are away, you may be able to do it just by writing a BASIC program, which will save you a lot of money.
Fourth, artificial intelligence
Artificial Intelligence (AI) is undoubtedly the most exciting field of robotics, and undoubtedly the most controversial: everyone agrees that robots can work on assembly lines, but there are disagreements about whether it can be intelligent.
Just like the term "robot" itself, it is also difficult for you to define "artificial intelligence". The ultimate artificial intelligence is the reproduction of the human thought process, that is, an artificial machine with human intelligence. Artificial intelligence includes the ability to learn any knowledge, reasoning ability, language ability and the ability to form one's own opinions. Robot experts are still far from achieving this level of artificial intelligence, but they have made great progress in the limited artificial intelligence field. Nowadays, machines with artificial intelligence can imitate certain elements of intelligence.
Computers already have the ability to solve problems in limited areas. The implementation process of using artificial intelligence to solve problems is complex, but the basic principles are very simple. First, an artificial intelligence robot or computer will collect facts about a certain situation through sensors (or manual input). The computer compares this information with stored information to determine its meaning. The computer calculates various possible actions based on the collected information, and then predicts which action will work best. Of course, a computer can only solve problems that its program allows it to solve, and it does not have the ability to analyze in a general sense. The chess computer is an example of such a machine.
Some modern robots also have limited learning capabilities. The learning robot can recognize whether a certain action (such as moving the leg in a certain way) achieves the desired result (such as bypassing an obstacle). The robot stores such information, and when it encounters the same situation next time, it will try to make actions that can be successfully responded to. Likewise, modern computers can only do this in very limited situations. They cannot collect all types of information like humans do. Some robots can learn by imitating human actions. In Japan, roboticists demonstrate dance moves to a robot and let it learn how to dance.
Some robots have interpersonal communication capabilities. Kismet is a robot made by the Artificial Intelligence Laboratory of the Massachusetts Institute of Technology. It can recognize human body language and the tone of speech, and respond accordingly. The authors of Kismet are very interested in the way of interaction between adults and infants. The interaction between them can be completed only by tone and visual information. This low-level interaction can be used as the basis of a human-like learning system.
Kismet robot
Kismet and other robots made by the MIT Artificial Intelligence Laboratory use an unconventional control structure. These robots do not use a central computer to control all actions, their low-level actions are controlled by low-level computers. Project director Rodney Brooks believes that this is a more accurate model of human intelligence. Most human actions are made automatically, rather than being determined by the highest level of consciousness.
The real problem with artificial intelligence lies in understanding how natural intelligence works. The development of artificial intelligence is different from the manufacture of artificial hearts. Scientists do not have a simple and specific model for reference. We know that there are tens of billions of neurons in the brain, and our thinking and learning are done by establishing electrical connections between different neurons. But we don't know how these connections achieve high-level reasoning capabilities, or even the realization principles of low-level operations. The neural network of the brain seems incomprehensibly complex.
Therefore, artificial intelligence is still largely a theory. Scientists propose hypotheses about the principles of human learning and thinking, and then use robots to experiment with their ideas.
Just as the physical design of robots is a convenient tool for understanding animal and human anatomy, research on artificial intelligence also helps to understand how natural intelligence works. For some robotics experts, this insight is the ultimate goal of designing robots. Others are imagining a world where humans and intelligent machines live together. In this world, humans use various small robots for manual labor, health care, and communication. Many roboticists predict that the evolution of robots will eventually make us completely semi-robots, that is, human beings fused with machines. There is reason to believe that future humans will implant their thoughts into robust robots and live for thousands of years!
In any case, robots will play an important role in our future daily lives. In the next few decades, robots will gradually expand to areas beyond industry and science and enter daily life, which is similar to the process that computers began to spread to families in the 1980s.
No comments:
Post a Comment