Intelligent robots yesterday, today, and tomorrow
The evolution of ideas about the ways of developing robotics, its goals, and tasks is very similar to what is observed in such a field as artificial intelligence. It is explained by the fact that the original tasks turned out to be much more complex, requiring the creation of entirely different models, methods, technologies, and above all - artificial intelligence technologies. Artificial intelligence (AI) technologies have always been closely related to robotics. It is no accident that one of the directions of AI is still considered to be the goal-oriented behavior of robots (creation of intelligent robots capable of autonomously performing operations to achieve the goals set by humans). In general, a robot can be defined as a technical complex designed to perform various movements and some intellectual functions of a human and possessing the necessary for this purpose actuating devices, control and information systems, along with the means for solving computational and logical tasks.
At present, there are three generations of robots:
- Programmatic. Rigidly programmed (cyclogram).
- Adaptive. Ability to automatically reprogram (adapt) depending on the environment. Initially, only the basics of the program are set.
- Intellectual. A task is input in a general form, and the robot can make decisions or plan its actions in an uncertain or complicated environment it recognizes.
Intelligent robots
It is generally accepted that an intelligent robot has a so-called model of the external world or internal environment, which allows the robot to act in an uncertain information environment. If this model is implemented in the form of a knowledge base, then it is appropriate that this knowledge base is dynamic. In this case, the correction of the rules of behavior of the robot in the conditions of a changing environment naturally implements the mechanisms of self-learning and adaptation. Thus, an intelligent robot is a robot that includes an innovative control system (IS). IS means a computer system for solving problems that a human can not solve in real-time, or their solution requires automated support or gives results comparable to human decisions. In addition, among other things, it is implied that for the tasks to be solved, the IS does not assume the completeness of knowledge, and the IS itself should have the ability: to order data and expertise with the allocation of essential parameters; to learn from positive and negative examples, to adapt under changes in the set of facts and knowledge, etc. If we define the intelligence of a robot in more simple words - the system can solve problems formulated in a general way. So, despite the many proposed intelligence criteria, the most robust requirement remains that the human role in interaction with the IS should be reduced only to formulating the task.
The architecture of intelligent robots
Today, it is believed that an intelligent robot should include: Actuators - these are manipulators, undercarriage, and other devices, with the help of which the robot can act on the objects around it. And by their structure, these are complex technical devices with servo drives, mechatronic parts, sensors, and control systems. By analogy with living organisms, these are the arms and legs of the robot. Sensors are systems of vision, hearing, touch, distance sensors, locators, and other devices that allow information from the surrounding world. The control system is the brain of the robot, which must receive information from the sensors and control the actuators. This part of the robot is usually implemented by software. The control system of an intelligent robot should include the following components: A world model - reflects the state of the world around the robot in terms that are easy to store and process. The world model performs the function of remembering the condition of objects in the world and their properties. Recognition system - this includes image recognition systems, speech recognition, etc. The recognition system's task is identifying, i.e., "recognition" of objects surrounding the robot and their position in space. As a result of the components of the recognition system, a model of the world is built. The action planning system performs a "virtual" transformation of the world model to obtain some action. In this case, the achievability of the goal is usually checked. The result of action planning work is the construction of plans, i.e., sequences of elementary actions. Action execution system - tries to execute the planned actions by issuing commands to actuators and controlling the execution process simultaneously. If the execution of an elementary action proves impossible, the whole process is interrupted, and a new (or partially new) planning must be performed. Goal control system - defines the hierarchy, i.e., the importance and order of achieving the set goals. Essential properties of the management system are the ability to learn and adapt, i.e., the ability to generate sequences of actions for the goal set and to adjust its behavior to changing environmental conditions to achieve the goals set.
AI technologies for intelligent robots
Intelligent systems are a necessary component to solve the problems of creating a model of the world, a system of action planning and goal management. The knowledge base in intelligent systems is one of the main parts of the world model and its transformation functions. Image recognition has long been a necessary part of complex robotic systems. Volumetric vision systems provide information about the orientation of objects in space. Significant changes are currently taking place in this area. Recognition and speech generation is necessary for effective communication with humans. Without these technologies, full-fledged contact with humans is impossible. Significant advances have been made in the area of text-based speech generation. Things are worse with speech recognition because it is a more complex task.
Multi-agent systems are used to collectively control many robots that can work individually or as a team.
Intelligent robot control
Operations are performed by the intelligent robot in the following sequence:
- a plan of operations is made,
- taking into account the recognition of the working conditions and the state of the object, the order of operations is determined,
- the robot actuators are actuated.
In contrast to intelligent robots, in conventional robots, the plan of operations and the order of actions depend on the willful decision of a human operator, and the actuation of actuators is automatic.
A characteristic feature of an intelligent robot control system is that it has a function of recognizing an object and its state using sensor devices and determining (based on this information and commands received from the operator) the actions to be performed. It essentially makes the robot versatile and capable of adapting to its environment and provides simple control of the robot.
The main points in the control system of an intelligent robot are:
- control of the actuators,
- recognition and sensing devices,
- command processing and decision making.
Information exchange between a human and an intelligent robot
Recognition and sensory devices
An intelligent robot must primarily possess the ability to adapt to its environment. In this, of course, sensory devices that perceive external information play an essential role.
The fields of application of intelligent robots are vast. They can be used both in industry and in scientific research. Depending on the robot's intended use, a wide variety of sensing devices are required. Here, we will limit our discussion to sensor devices and pattern recognition systems for industrial robots, which mainly perform handling operations.
- a) Sensor devices required for intelligent industrial robots. The following three main types of sensing devices can be distinguished:
vision devices,
hearing devices,
touch devices.
The primary sensory devices required by an intelligent robot are as if the three senses.
- b) Recognition function of an intelligent robot. The signals received by the sensory devices are usually information corresponding to some "image." Based "on this information, it is possible to identify the features of the object and make the necessary measurements of the image.
In recognizing objects using vision and hearing devices, one can distinguish between the recognition of the thing of the operation itself and the recognition of signals from the operator. In the second case, we are talking about the glory of recorded calls, for example, on paper in letters or graph form, or about the recognition of signs the voice gives.
- Vision Devices. Visual information about an image can be obtained using a television camera as an input device. There are also plans to use devices in which laser beams will be used as a visual input device.
The information for recognition is mainly light and shade (contrast) information, but color information and position information (by glowing points in the image) can also be used.
Recognized objects are three-dimensional objects. The recognition process proceeds in the following (sequence:
- preprocessing (contour image),
- recognition,
- measurement (of the necessary parameters of the object, e.g., its size, location, etc.).
In the problem of recognition, there are many issues of great interest, such as recognizing complex shapes, recognizing objects when some are placed on top of the other, etc.
For an intelligent robot to function, it must "understand" its environment. The robot remembers the real world of its territory in the form of some model, but vision alone is not enough to assess the environment.
Hearing devices. Hearing devices are more often used for measurement and detection purposes than for pattern recognition. In addition to ultrasonic measurements concerning objects' location and dimensions, hearing devices are used to recognize the end of operations and detect unusual noises by picking up sounds and noises in the workplace with microphones.
Touch devices. They are mainly used for location detection and direct measurement. However, in addition to this, tactile information can be used to recognize the surface condition of objects and their properties (weight, elasticity, etc.).
Command processing and decision making
An intelligent robot makes a plan of operations and performs work based on the commands given by a human operator and recognition results. The orders delivered to the robot can have a variety of forms: from simple and concrete to quite abstract. As for the methods of using recognition results, they can also be very different: from changing programs with the help of simple sensors, giving the robot some properties of adaptation, to automatically performing heuristic programming.
Information exchange between a human and an intelligent robot
To perform operations using an intelligent robot, exchanging information between a human and a robot is necessary. Intelligent robots have a significant difference in "intelligence levels." According to this, there are means of information exchange corresponding to one or the other level. In the information exchange between a human and an intelligent robot, there is information that the human has to give to the robot and knowledge that the intelligent robot directs to the human. The first includes commands given to the robot and learning information. The second provides messages to the human about the progress of operations, the end of operations, results, etc., or requests for help from the human. Visual and auditory communication can facilitate good information exchange between an intelligent robot and a human. In optical touch, the means for the smart robot to receive information from the human can be letter recognition and figure and drawing distinction. The standard for mutual information transfer combines a display with a light pen. The advantage of visual communication is that if a human uses letters, drawings, figures, etc., while giving commands or teaching, it helps to organize his thinking. In auditory communication, voice recognition is the means of conveying information. Using speech synthesis techniques, it is possible to give commands and teach the robot with a human voice. In terms of specifying the communication methods listed above, developing information preprocessing hardware and researching information structure and machine languages are essential.
Significant milestones in the history of intelligent robots
A significant step forward in the development of technology and the most famous were three "turtles" created by the English biophysicist and neurophysiologist G. Walter in 1950 - 1951. These devices are self-moving electromechanical toys, capable of crawling to or from light, avoiding obstacles, entering the "feeder" to recharge dead batteries, and the like. The " turtles " demonstrated learning properties despite a straightforward device and analog control system; the "turtles" showed teaching properties. The emergence of intelligence in robots is related to the development of the computer. The "Shakey" robot was created in 1969 at the Stanford Research Institute (USA) and was then called an integral robot or mobile automaton using the principles of artificial intelligence. This robot consisted of a moving part, a computer, and corresponding software. The robot was designed to study real-time control processes in a complex environment. All the functions that the robot must perform can be divided into three classes: problem-solving, perception, and simulation. The robot control system that serves problem-solving uses the information recorded in the model to plan and calculate the sequence of actions. As the environment changes by the active movements of the robot itself or for other reasons, the model must be transformed to remember these changes. In addition, new, current information about the environment, which the robot acquires as it learns it, must be added to the model. In 1969, the Electrotechnical Laboratory (Japan) began to develop an "industrial intelligent robot" project. The goal was to create a robot with artificial intelligence to perform assembly work with visual control. The robot's manipulator has six degrees of freedom and is controlled by a mini-computer (main memory 32000 words, external magnetic memory 273000 words). The manipulator is equipped with tactile sensors. Two TV cameras equipped with red-green-blue filters for color recognition of objects are used as a visual perception system. The robot could recognize simple things bounded by planes and cylindrical surfaces under special lighting. In 1972-1975, the Kyiv Institute of Cybernetics created a transport autonomous integrated robot (TAIR) model. The robot demonstrated purposeful movement in the natural environment, avoiding obstacles, etc. Structurally, TAIR was a three-wheel self-propelled cart equipped with a system of sensors: an optical rangefinder, a navigation system with two radio beacons and a compass, contact sensors, cart angle sensors, a timer, etc.
Applications of intelligent robots
Industrial robots started to be widely introduced into the production sphere in the seventies of the last century. These robots were controlled automatically by numerical control systems. The use of adaptive elements made it possible to expand the capabilities of industrial robots.
The automotive and electronic industries are the primary consumers in the field of industrial robotics. Currently, many different types of industrial robots are produced for manipulation, welding, painting, packaging, grinding, polishing, etc., with a wide range of applications in terms of precision and the nature of the operations performed.
Military robots
A good example is a program implemented by the Defense Advanced Research Projects Agency (DARPA), the leading research center of the US Department of Defense. Its essence is a creation of an army of robots. According to experts, the transition to a full-fledged robot-technical army should occur by 2025. Thanks to the introduction of robots, the most critical factor in military operations is excluded - the presence of live soldiers on the battlefield. Using satellite communications, such an army can be controlled from anywhere in the world. Unmanned aerial vehicles and light and medium ground weapons are already available. A new generation of heavily armored vehicles will also be available shortly.
Security robots
Currently, these types of robots are widely used for security. A modular reconnaissance robot was used at Germany's 2006 FIFA World Cup. In total, during the World Cup, at least 20 such machines were present at Berlin's Olympic Stadium at any one time.
In recent years, unmanned aerial vehicles (UAVs) have been actively used in various countries to ensure internal security - from patrolling borders, ports, pipelines, and other strategic facilities to monitoring the population.
Household robots
Robots for the home are becoming as much an attribute of the household as a TV or a refrigerator. Every year, new and improved models of household helpers are developed. Thanks to this, the price trend of recent years has been downward, and they are becoming more and more accessible to the general population.
The robotics industry in Japan is not the last place to create robotic human helpers. Experts point out that the task of large-scale research into the problem of human-robot-assistant coexistence, including its psychological and social aspects, is on the agenda.
Robots for games and entertainment have been gaining popularity in recent years. Many different robot toys for various ages are being created.
Medical robots
Medical robots are becoming an integral part of modern clinics. The da Vinci surgical system from Intuitive Surgical is widespread. It is a precision manipulator for doctors. The system accurately controls its movements and "sees" a three-dimensional image, which helps make the surgery much less painful. And this, in turn, contributes to a speedier recovery, making the treatment process more accessible and cheaper.
In Japan, home robots are widely used for patient care, and in US clinics, the machines are applied to deliver registration cards, dispense medications, perform surgery, and other functions.
Space robots
Space robots conduct exploration of space and other planets. Automated interplanetary stations and planetoids to study the surface of the solar system planets are essentially complex robots. The long delay in radio signal propagation virtually eliminates real-time telecontrol, so the planet rover must be able to make decisions on the spot, i.e., have the properties of an intelligent robot.
State of the Art and Prospects
Today, intelligent robots have moved out of the realm of pure science and are becoming as necessary elements of everyday life as television and cellular communications. However, some key challenges must be overcome to trigger a full-fledged boom. The difficulties of communication and coordination remain. A gripping mechanism also needs to be worked out. Still, robots' abilities are evolving rapidly. According to research, a 1.5-kilogram human brain can perform about 100 trillion operations per second-almost three times as many as the world's most powerful computer. Machines can achieve this processing power. However, a robot does not need all the human brain's capabilities to be functional enough. Sales figures give the scale of the change that is taking place. Experts estimate that the volume of the "household robots" market in Japan, for example, will reach an annual volume of 8 trillion yen ($70 billion) by 2025. The demand for personal and mobile robots is constantly growing. Today we live in a rapidly changing world, of which robots with artificial intelligence will be an integral part. We cannot stop these changes, but we can direct them to improve human life.