Introduction
From ancient legends or fables, it can be seen that humans have dreamed of flying like birds since their birth. Through generations of imagination, hard work, and continuous exploration, humans have tried various methods to fly freely in the sky, such as using kites, early rockets, flying cars, hot air balloons, and gliders, but none of them were able to achieve free flight. In 1903, the Wright brothers completed the first powered manned flight of humanity using their self-designed "flying machine - airplane", realizing the dream of human flight in the air. As human beings dreamed of flying in the sky, they also began to imagine that humans could control the flight of airplanes on the ground instead of airplanes, which gave rise to the concept of drones. In 1916, Sperry and Lawrence of the United States conducted their first drone flight, marking the beginning of human research on drones. After World War I and World War II, especially during the Cold War period, drones made significant progress; After 100 years of development, drones have formed a huge family system with numerous branches. They can be classified from multiple dimensions such as fuselage structure, volume and weight, flight altitude, range, duration, and purpose. For example, based on the structural characteristics of the fuselage, they can be divided into unmanned helicopters, unmanned fixed wing aircraft, unmanned multi rotor aircraft, unmanned airships, unmanned paraglider wing aircraft, and flapping wing drones; Classified by volume and weight characteristics, it can be divided into large unmanned aerial vehicles, medium unmanned aerial vehicles, small unmanned aerial vehicles, and micro unmanned aerial vehicles; According to their usage, they can be divided into military drones (intelligence, reconnaissance, surveillance, decoy, electronic countermeasures, communication relay, target aircraft, and unmanned aerial vehicles, etc.), civilian drones (such as police, firefighting, meteorology), and consumer grade drones (for leisure purposes such as aerial photography and gaming); It can also be divided into high-altitude drones, long-range drones, high-altitude long-range drones, etc.
The development of anything in the world is keeping pace with the times, and its essence, connotation, and appearance are constantly evolving and evolving. The same is true for drones. The definition and connotation of drones also need to be constantly improved dynamically. Drones have evolved from the era of simple human controlled aircraft flight to the current stage of automatic (autonomous) drone flight. They should also have a clear definition to adapt to the development of technology and the improvement of people's understanding, otherwise it will be detrimental to the development of drone technology and even hinder its progress. This article conducts research and analysis on the development and changes of drones, summarizes the connotation and definition of drones and their development, elaborates on the connotation of drone control, explores the relationship between autonomy and intelligence, provides the design concept and engineering elements of autonomous intelligent control, and proposes the implementation method and scheme of autonomous intelligent control for drones.
2. Definition and Connotation of Drones
On September 12, 1916, Sperry and Lawrence of the United States completed the first powered unmanned flight for humans, marking the beginning of a new era in drone research. When people first defined drones, they used the most intuitive physical concepts, only considering the physical position relationship between pilots and aircraft. That is to say, an aircraft without pilots on board is called a drone, initially expressed in English as a pilotlesaircraft, which is also the most basic connotation of drones.
With the development of radio remote control technology, aviation engineers have used radio to remotely control the flight of unmanned aerial vehicles on the ground, which has led to the emergence of the terms Remote Pilot Aerial Vehicle (RPAV) and Remote Pilot Aircraft System (RPAS). During this period, some people also used Uninhabited Aerial Vehicle as the term for unmanned aerial vehicles.
The Unmanned Aircraft System (UAS) was adopted by the US Department of Defense and FAA in the UAV Roadmap 2005-2030 [1], which defines it as a powered aircraft that does not carry an operator, uses aerodynamics to provide lift, can fly autonomously or remotely, can be used once or retracted, and carries lethal or non lethal payloads. This definition clarifies the most basic connotation of drones: 1) There are no pilots on board the aircraft; 2) Capable of completing certain mission tasks;
3) Can be reused.
According to the above definition, ballistic or semi ballistic aircraft, cruise missiles, and shells cannot be considered unmanned aerial vehicles because missiles cannot be recovered; At present, there is still doubt and debate about whether remote-controlled aviation model aircraft belong to drones, and there is no consensus. The main reason is that modern aviation model aircraft also use advanced control technology, and their functions are also changing, making it difficult to clearly identify them from a judicial and flight management perspective. In the author's opinion, if only performing entertainment activities within line of sight, remote-controlled aviation model aircraft can be considered not to be drones,
Until now, unmanned aerial vehicles (UAVs) have been used technically to represent drones, while drones are often used in civilian settings due to the regular noise of old military drone engines similar to the buzzing of drones [1,2].
There are two forms of expression for drones: unmanned aerial vehicle and unmanned aerial vehicle. Upon careful exploration, their meanings are different. Unmanned literally has two meanings: one is direct meaning, unmanned, unmanned on the aircraft; The second is unmanned operation; Uninhabited only means unmanned on the plane. Literally speaking, unmanned aerial vehicles should have two meanings at the same time, that is, if a person is not on the aircraft and does not control it, the aircraft can fly "normally". That is to say, the entire process of unmanned aerial vehicles from takeoff preparation, taxiing, takeoff, aerial flight, return landing, and exit shutdown (Figure 1) does not require human intervention. Therefore, using "unmanned aerial vehicle" to define drones is the most appropriate and better reflects the connotation of "real" drones. The English expression for Unmanned Combat Aircraft (UCAV) should be: unmanned combat aerial vehicle.
After 100 years of development, the evolution and development of drones are comprehensive, and their connotations have undergone significant changes. Regardless of the ever-changing tasks they perform, the most fundamental change is the change in their flight control methods. According to the changes in drone flight control methods, drones have gone through the following stages of development and formed corresponding types of drones:
(1) Remote control flying drone (stage);
(2) Remote control and local automatic flight unmanned aerial vehicle (stage);
(3) Fully automated flight unmanned aerial vehicle (stage);
(4) Fully automatic and local autonomous flight unmanned aerial vehicle (stage);
(5) Fully autonomous drone phase (the next phase is approaching).
At present, the highest level of unmanned aerial vehicles (UAVs) internationally is fully automatic and local autonomous flight UAVs. The types of UAVs are selected based on different task requirements, manpower and cost, and actual situations. These types of UAVs can coexist and complement each other, fully leveraging the advantages of each type of UAV.
At the current stage of technological development, narrowly defined, the flight of drones may not be directly related to humans, that is, there is a state of isolation between humans and drones; In a broad sense, as drones are a type of flying tool or weapon, in order for humans to use them, it is necessary to clarify the issue of "drone permissions": humans are the owners (masters) of drones, and the behavior of drones must be subject to human control. However, due to the limitations of human abilities, energy, and precise control of drones, it is impossible to control drones every minute or moment. Therefore, drones must have the ability to work independently (automatically).
From the above analysis, it can be seen that there are differences in the connotation and essence between remotely piloted vehicles and unmanned aerial vehicles. The schematic diagram of the structure of a remote-controlled drone is shown in Figure 2, and the schematic diagram of the structure of a "real" drone is shown in Figure 3. From Figures 2 and 3, it can be visually seen that the difference between the two is that remote-controlled flying drones operate under direct human control, that is, humans are in their working environment; However, drones work under human authorization, with humans outside their work environment rather than within it. This fundamental difference leads to differences in system structure, control functions, and implementation methods in their design.
No matter how a drone is defined, having the ability to work independently should be its essential attribute. However, the unchanging principle is that the drone is used by humans, and humans are the "masters" of the drone. The drone must obey human control, and the self independent work permission of the drone is naturally set by humans at any time. It is obvious that in order to achieve the function of the drone, the structure and ability of the drone are constructed during the "manufacturing" of it. Therefore, a standard unmanned aerial vehicle should have three operating modes: autonomous (automatic) mode, manual intervention mode, and manual control mode. The use of these three modes is set and selected by humans (operators). The mode selection of human (operator) should also take into account the complex situation of the actual environment, and follow the principle of "going out, not taking your life" for mode selection The definitions and meanings of
the three working modes above are as follows:
(1) Autonomous (automatic) mode is the default mode of unmanned aerial vehicle systems, which works according to the rules, concepts, and ideas formulated by humans to control the flight of unmanned aerial vehicles;
(2) The manual intervention mode is in the autonomous (automatic) mode, where humans actively correct deviations in autonomous (automatic) flight by adding a delta increment on the basis of default control;
(3) The manual control mode refers to the direct operation of the aircraft by humans in emergency situations where the control system fails and cannot autonomously (automatically) control the drone. In general, manual control is difficult to ensure the control effect of the aircraft, and the reason is simple: humans are not present at the flight site, making it difficult to accurately perceive the motion information of the aircraft, thus making it difficult to accurately control the aircraft. The use of the above three working modes by drones actually clarifies the issue of "human-machine permissions". As the "owner" of the drone, humans control the drone by formulating rules and strategies, and the drone autonomously (automatically) generates control instructions to control the flight of the aircraft according to the rules and strategies; When there is inconsistency with human imagination during flight, humans can make moderate corrections; When an emergency situation occurs during flight, people can directly control the aircraft, which is a desperate behavior.
Due to the fact that drones have different missions and tasks in different flight stages throughout the entire usage process, the three working modes execute different task commands in different flight stages. Based on this reason, when constructing the working logic structure of the drone system, the principle of "adapting to local conditions" should be adopted for the logical construction of the flight stage and the three working modes, that is, confirming the stage first, and then selecting the working mode. The adoption of the above mode also involves the issue of daily training concepts for drone operators. According to the setting of the above mode, the daily training method for drones should be mainly based on simulator virtual training, with a focus on task training, supplemented by flight special situation training, and supplemented by a small amount of physical flight training.
It should be emphasized that since drones can fly autonomously (automatically), it involves how to fly autonomously (automatically)? Is flying good or bad? The ability and level of autonomous (automatic) flight, which is the level of intelligence, raises the question of the connotation and interrelationship between autonomy and intelligence.
The Essence and Relationship of Autonomy/Intelligence
The essential connotation of autonomy/intelligence and its interrelationship criteria are the basic principles of drone design.
In general, autonomy and intelligence are two different categories of concepts. Autonomy expresses behavior patterns, and completing a certain behavior through one's own decision-making is called "autonomy"; Intelligence is the ability to complete behavioral processes, which refers to whether the methods and strategies used conform to natural laws or the behavioral rules of individuals (or groups). Finding a reasonable "path" to complete a task in a ever-changing environment is called intelligence. It is obvious that intelligence is hierarchical and hierarchical.
The relationship between autonomy and intelligence should be: autonomy comes first, intelligence comes second, and the two should complement each other; Autonomy may not necessarily be intelligent, but autonomy hopes for intelligence; Intelligence relies on autonomy, and the level of intelligence depends on the level of autonomy. Intelligence is a combination of autonomy and knowledge and its application. The general process of intelligence generation should be: under the premise of autonomy, comprehensively utilizing various abilities such as authority, initiative, passion, and insight, to feel information, extract information, accumulate knowledge, summarize knowledge, summarize features and refine them, improve and perfect knowledge structure, and integrate knowledge to achieve the goal of conforming to natural laws as much as possible.
Intelligence has relativity, and the intelligence of different individuals varies. These differences come from both the intelligence given to them at birth and the intelligence acquired through learning and improvement. Due to the fact that nature and its existence are a contradictory unity, "right" and "wrong", "good" and "bad", "smart" and "stupid" are all relative and can be transformed into each other. Therefore, in order to correctly understand, master, and apply "intelligence", it should be recognized that in human society, "high intelligence" is determined by human standards or worldviews, and there is a phenomenon of inconsistency between the so-called "high intelligence" and the true "high intelligence" in nature, as well as the phenomenon of "great intelligence being foolish". This is a problem of human cognitive ability. This is also one of the mysteries of human society, which also extends to the famous saying of the Chinese nation, "A wise man has a thousand worries, and there must be a mistake." This saying is both truth and axiom. Therefore, when designing individual intelligence strategies, it is necessary to use the other party's "mistake" to control and win the other party. Therefore, the difficulty is how to explore the other party's "mistake". Once the other party's "mistake" is mastered, it is easy to formulate strategies to win the other party.
It should be emphasized that people often confuse intelligence with intelligence, which is incorrect. Intelligence (IQ) and intelligence have completely different connotations. Intelligence refers to the ability to acquire, reason, and apply knowledge, while intelligence refers to the degree to which the results of acquiring, reasoning, and applying knowledge conform to natural laws. Intelligence is only one element of intelligence formation, and having intelligence does not necessarily mean intelligence.
High intelligence necessarily requires high intelligence.
When constructing drones, they should be endowed with "equivalent" autonomy and intelligence capabilities to meet human needs, and design should be based on the above relationship criteria between autonomy and intelligence.
Due to the fact that drones work under human authorization, they are endowed with intelligent abilities by humans. In the process of endowing them with intelligent abilities, they will "implant" factors that can cause "disasters" in drones. The so-called "disasters" are "natural disasters" and "man-made disasters", which are creations of nature and human beings have no ability to control. "Human disasters" are subjective thoughts that humans want to do. Due to issues with ability or sense of responsibility (negligence), the aircraft does not listen to human commands or does not "do things" according to human thoughts, resulting in things that people do not want to see, do not want, and cannot be controlled by humans. This does exist, emphasizing that these uncontrollable situations are temporary and cannot last forever; The reason is that drones require energy, and their level of intelligence is limited. Additionally, humans can use other means to control or destroy them. The statement that "machines will partially replace humans in the future" holds in a sense, but it is impossible for machines to control humans in the future, but it will bring "trouble" or "disaster" to humans. Such "trouble" or "disaster" is not entirely bad, for some people it is "bad", but for others it is "good" (provided that it can be controlled), using this characteristic to deal with "others".
The autonomy of specific individuals has authority and scope, and is limited by many factors, just like the concept of human autonomy. Firstly, humans are limited by natural laws or legal principles, and secondly, they are constrained by the group they belong to (state/system, social group/regulation, unit/regulation, family/ethics and morality); The autonomous authority of drones is first limited by human (user) limitations, as well as their own capabilities and usage environments (such as natural geographical environment, aircraft formation requirements, mission requirements, etc.).
The basic principle or bottom line for achieving autonomous behavior of drones is that they must have independent and autonomous information acquisition ability, independent and autonomous information processing and decision-making ability, and independent behavior execution ability. Independent and autonomous information acquisition ability is the foundation of autonomy. Without independent and autonomous information acquisition ability, one is like a castle in the air, a vassal or parasite; For drones, if information needs to be provided by the outside world, once the outside world no longer provides information, the drone will become "deaf and blind"; Independent and autonomous information processing and decision-making abilities are the core of autonomy, otherwise it cannot be called autonomy. One has no independent opinions, only listens to others' orders, and can only act as a puppet; The meaning of independent behavior execution ability is to independently obey the commands of one's own decision-making level, execute tasks based on one's own abilities, rather than just mechanically executing external commands. Therefore, becoming a complete autonomous intelligence for drones requires the composition of three layers above, which interact and depend on each other, forming a community of shared destiny. Figure 4 shows the information flow diagram and requirements for drone autonomous agents.
The basic principle or bottom line for the implementation of unmanned aerial vehicle intelligence is that at each level of the self agent, there must be basic intelligent functions and capabilities, and the three should be coordinated and complementary; The intelligent functions of each layer should implement basic natural laws or behavioral rules, and have the ability to learn and improve themselves. The ability to perceive and extract information independently and autonomously is reflected in the following three aspects: (1) Information sources should be natural attributes, and information sources cannot be artificially set with feature attributes, which will make it difficult to ensure the uniqueness, credibility, and security of information; (2) The information perception of information sources should be completed independently and cannot utilize other external information and auxiliary segments; (3) The extraction of information features should be completed independently and cannot utilize feature information provided by other external means.
It is obvious that the information provided by GPS, Beidou information, and data links is not autonomous information, and these information sensors (or devices) are not autonomous information sensors (or devices); The sensors (or devices) listed below belong to information autonomous perception sensors: inertial sensors; Visual perception equipment; Terrain matching perception equipment; Optical sensing (laser/infrared/ultraviolet, etc.) sensors; Astronomical information perception sensor; Electromagnetic sensing sensor.
Taking airport perception and feature extraction as an example to illustrate the meaning of independent and autonomous ability in information perception and extraction.
Installing reflective mirrors or other artificially set signs at airports to indicate the location characteristics of the airport and the parameter characteristics of the runway does not reflect autonomous characteristics, as these artificially set features are easily changed or destroyed and cannot be used as reliable and unique information attribute markers. The attribute features around the runway that are difficult to change should be used as the information source, and the size of the surrounding area should be selected based on the distance. From far to near, the area gradually decreases. Within the area, the attribute features that are difficult for humans to change should be used as the information extraction features, such as mountains, rivers, natural landscapes, and their interrelationships. This can achieve both autonomy and fault recognition and reconstruction, providing reliable and unique feature information for autonomous decision-making.
At present, humans are the "elves" of nature and the most reasonable autonomous intelligence complex. The design concept of drones should follow the relationship between human body structure and human thinking/behavioral logic, and the strategy and policy of complexity management should be "divide and rule". Human intelligent processing is hierarchical, prioritizing the processing and response of external information, prioritizing the handling of life and death issues, followed by improving the quality of life and environment. Under different conditions, the methods for handling life and death issues also vary. Therefore, as a tool for human use, the intelligent processing of drones should also be hierarchical and hierarchical [1,5], with priority. Suggest dividing the intelligence level of drones into three levels (levels) based on priority principle
The first level is the individual safety flight level of unmanned aerial vehicles, defined as "high reliability to live": able to fly safely, with safe altitude, speed, and attitude states; Having collision avoidance ability, able to autonomously and safely avoid stationary and moving objects; The ability to refuel in the air to ensure flight energy; Fault reconstruction and self repair capability; Special situation safe landing capability.
The second level is to complete group specific tasks, defined as "high-quality work": able to achieve four-dimensional navigation, situational awareness and cognition; Capable of achieving path planning and re planning; Task planning and re planning; The cognition of unconscious information.
The third level is to achieve collaborative tasks among aircraft groups, defined as "efficient work for collective missions": formation flying; Manned/unmanned collaborative operations; Group perception and situational sharing; Cluster joint operations.
In summary, in order to achieve the above three levels of autonomous intelligent control and "create" a complete and perfect drone, the following "creation" principle should be followed: while ensuring autonomy and intelligence, other means and sensors should be used to fully obtain and utilize all available information, which can achieve twice the result and effect with half the effort. It is emphasized that autonomy and intelligence must ensure the bottom line. To achieve this principle, drones must have three information loops: autonomous information loop, non autonomous (external assistance) information loop, and permission information loop (high intelligence should also have an unconscious information loop); The acquisition, processing, and application of information, as well as task decision-making, must be ensured to be completed autonomously. This requires the construction of an "agency" to achieve at least two "agents". The first is the "agent" on the ground to complete the construction and distribution of human instructions, and the second is the "agent" on the aircraft to achieve complete autonomous intelligent control. Its most critical role is the core functional "component". Without this "agent", drones cannot achieve true autonomous intelligent missions. Figure 6 illustrates the information flow and structure of drones.
Drones can function as a complete and independent entity in nature, and they should be integrated into the natural world. They should be closely related to the natural world, the people who are their masters, and their opponents. Otherwise, it is difficult to achieve a high level of autonomous intelligence. Only when the drone body is related to both nature, humans, and agents can it achieve true autonomous and intelligent functions. Figure 7 illustrates the "social" relationships of "real" drones.
4 Implementation Method for Autonomous Intelligent Control of Drones
Just like humans, anything with "life" in the world is composed of two parts: a carrier (human body) and a functional soul. Therefore, the autonomous intelligent control system of drones is no exception. It is composed of hardware carriers and the functional soul carried by the carrier (information acquisition and behavioral decision-making, control law and control logic, etc.) (see Figure 8 for the basic composition of the system). It should be pointed out that hardware carriers and functional souls interact with each other, and different carriers carry different functional souls. A highly intelligent soul requires a high-performance carrier, and the two are complementary. When creating an autonomous intelligent control system for unmanned aerial vehicles, both must be considered simultaneously and coordinated.
In order to achieve autonomous intelligent control of unmanned aerial vehicles, it is necessary to expand the carrier, expand the function, improve the intelligence level, and improve the fault reconstruction and self-healing capabilities. The system structure should adopt a distributed system, and the sensors (components) for information perception and acquisition, information processing analysis and decision-making computing units, and instruction execution components should be relatively independent and configured in a distributed manner. Distributed configuration here has two meanings, one of which is the distributed configuration of hardware carriers
Set (Figure 9 (a)), and secondly, functional control is also distributed, with main control center, sub control center, and auxiliary control (Figure 9 (b)) [4,6]. Relatively speaking, the composition and implementation of distributed carriers are easier, but for the functional soul of autonomous intelligence, it is very difficult to "create" the logic and information architecture in order to achieve the requirement of 3 layers and 10 levels of autonomous intelligence. After analysis and research, a feasible logic and information architecture is shown in the four ring structure in Figure 10. The first and second rings complete the first level of autonomous intelligence control, achieving "high reliability and survival"; The first, second, and third loops complete the second level of autonomous intelligent control, achieving "high-quality work"; The first, second, third, and fourth rings complete the autonomous intelligent control of the third level.
To achieve efficient work for the collective mission. The framework structure of autonomous intelligent control above is based on the "divide and conquer" strategy, which first divides into levels, and then adopts different intelligent decision-making strategies at each level to simplify the complexity of the system.
5 Discussion and Conclusion
On the basis of years of research, the author of this article has reorganized the definition and connotation of drones, analyzed and elaborated on the relationship between autonomous and intelligent control of drones, and initially formed the basic concept and idea of realizing autonomous intelligent control of drones. The architecture of engineering implementation of autonomous intelligent control system has been constructed, and some results have been adopted in actual aircraft control systems, which is only the starting work; Although autonomous intelligent control has a research history of many years, the depth of research is far from sufficient, mostly staying at the theoretical level, and the engineering application level is still in its infancy. It is urgent to conduct in-depth research on specific information perception methods, intelligent decision-making strategies, and implementation methods (algorithms) of autonomous intelligence.
