For UAV systems, due to the different requirements for autonomy on different platforms and different task scenarios for different users, the requirements for the realization of autonomous control systems are also different. However, it is still possible to study and analyze the capability requirements of UAV autonomous control systems from a macro perspective, so as to provide an important reference and reference for technical research and guide the development and design of UAV autonomous control systems in engineering practice.
Up to now, many domestic experts and scholars have conducted in-depth research on the capabilities of autonomous control systems for UAVs, and have formed a number of relatively complete viewpoints. Researcher Wang Yingxun of Beijing University of Aeronautics and Astronautics believes that autonomous drones should have four basic capabilities: security capabilities, perception capabilities, decision-making capabilities, and coordination capabilities; Professor Zhu Huayong of the National University of Defense Technology  believes that the future technical requirements for autonomous control of UAV systems It is mainly reflected in the realization of the following four capabilities: comprehensive environmental awareness and intelligent battlefield situation awareness, autonomous navigation, planning and control capabilities under complex conditions, human-computer intelligent integration and learning adaptability, and multi-platform distributed collaboration Ability; Researcher Fan Yanming of Shenyang Institute of Aviation Industry believes that the system must have three main capabilities: independent information acquisition capability, independent independent information processing and decision-making capability, and independent behavior execution capability based on the autonomous behavior of drones. Researcher Ma Weihua and Liu Jiarun of Beijing Aerospace Automatic Control Institute believe that the capabilities of aerospace intelligent control systems can be summarized into five aspects: perception and understanding capabilities, movement and control capabilities, learning and adaptation capabilities, planning and decision-making capabilities, communication and collaboration ability. Regarding the intelligent control system, Academician Wu Hongxin clearly pointed out that it should have the following abilities: perception and cognition, online planning and learning, reasoning and decision-making, and coordinated control of multiple actuators.
Summarizing the above-mentioned various points of view, it is not difficult to find that the capability requirements of the UAV autonomous control system and the embodiment of UAV autonomy are highly unified, and a brief analysis can be made from the following perspectives.
First of all, if it is implemented based on the OODA cycle, since "no one is on board" and "people are on the loop", for the autonomous control system of UAVs, it is natural to expect that all links can best be derived from the UAV system. The master completes and forms a closed loop of control. In this way, in order to realize an autonomous OODA cycle, it is necessary to have the corresponding perception and cognition ability, evaluation and judgment ability, planning and decision-making ability, and control execution ability.
Secondly, considering the actual use of drones, in addition to the drone itself, the main elements of its application scenarios generally include the natural environment, missions, hostile forces, friendly forces, and operations/users. The realization of the UAV autonomous control system must comprehensively consider the influence of the above-mentioned elements, especially the effective resource management, scheduling and control for flight and mission must be realized through the interaction, integration and coordination of one's own forces. Therefore, in addition to the capabilities necessary for the realization of the above-mentioned OODA cycle, the UAV autonomous control system should also have the corresponding human-machine fusion capability and multi-aircraft coordination capability.
In addition, in addition to the basic capabilities required for the realization of autonomous control functions, it is also expected that the UAV autonomous control system will have certain fault-oriented fault-tolerant repair capabilities, as well as more intelligent learning and evolution capabilities.
To sum up, the main capability requirements of the UAV autonomous control system can be summarized into 8 items, namely: perception and cognitive ability, evaluation and judgment ability, planning and decision-making ability, control execution ability, human-machine integration ability, and multi-machine coordination ability. , Fault tolerance, learning and evolution ability.
2.1 Perceived cognitive ability
Perception is a means of obtaining information from the outside world, and cognition is the formation of knowledge/cognition through what is perceived. Perception and cognition is the basis for the realization of autonomous control systems for UAVs. Especially under complex and uncertain conditions, only with the corresponding perception and cognition ability, the UAV can obtain sufficient and correct flight/mission environment information, its own motion and system status information, as well as operation instructions and mission target information, etc., support The realization of the desired functions and performance of the autonomous control system.
The objects of perceptual cognition are related information from various sources, so correspondingly, perceptual cognitive ability can be understood as related to information acquisition, recognition/discrimination, and information-based modeling. Among them, the perception ability focuses on the collection and acquisition of front-end information. On the one hand, it solves the problem of "whether or not" information, and on the other hand, it needs to distinguish and extract useful information from various information to solve the problem of "good or bad"; Cognitive ability focuses more on the processing and understanding of back-end information. Perception ability can be considered as the basis of cognitive ability. Compared with perception ability, cognitive ability is more complex and abstract, and contains a certain degree of subjective color. For example, it can be based on perceptual information to establish a certain degree of environment/threat/task. Cognitive models of preferences, etc.
2.2 Assessment and judgment
The assessment and judgment ability of the UAV autonomous control system is an extension of the perception and cognition ability. After obtaining the corresponding information based on the perception and cognition and establishing a cognitive model, it needs to deal with the enemy's situation/intention, environment/enemy threat, self-health, etc. Make effective assessments and judgments.
From the perspective of data fusion, evaluation and judgment ability belongs to the category of high-level data fusion. Obviously, the strength of the assessment and judgment ability directly affects the operation of the autonomous control system, and misjudgment and misjudgment may bring catastrophic consequences. The typical assessment and judgment ability is situation assessment ability and threat estimation ability. Among them, the situation assessment is based on various information such as "enemy + our side + environment + mission" to realize the integration of multi-layer views reflecting the battlefield/competition/operation situation; threat estimation needs to integrate threat subjects, behaviors, capabilities, intentions, and situation It is a higher level of integration to achieve "perception-understanding-prediction" based on multiple factors such as events, events, etc.
2.3 Planning and decision-making ability
Planning and decision-making capabilities are also an important manifestation of the intelligence of autonomous control systems. To reduce human participation in real-time control and enhance autonomous control capabilities, UAVs must make plans and decisions on their own under uncertain circumstances. The strength of this capability reflects the greatness of "pre-setting" and "adaptive changes". the difference.
Planning and decision-making abilities oriented to goals and tasks. Its realization relies on human experience, intelligent control methods, and the support of software and hardware. The implementation is mainly based on the information transmitted by the data link, the relevant data of the original database, and the related acquisition of perception and cognition. Information and the results of evaluation judgments. In the UAV autonomous control system, typical planning and decision-making capabilities include trajectory planning, mission planning, and tactical maneuvering decisions.
2.4 Control execution ability
For UAV autonomous control systems, the control execution capability is mainly for the maneuvering of UAVs, which is the ability to change its position and motion state based on the results of planning and decision-making. It is usually tightly coupled with the controlled object, which not only needs to achieve a certain degree of rapidity, agility and maneuverability, but also has corresponding requirements for attributes such as control accuracy, stability, and robustness.
The pros and cons of control execution ability not only depend on the design of control modes and control algorithms, but also on effective sensing devices and actuators. For example, for future unmanned combat autonomous aerial vehicles, in order to take into account the requirements of maneuverability and stealth, advanced designs such as variant configurations may be adopted. The complexity, nonlinearity, and uncertainty brought about by the system affect the system. The ability to control execution is a huge challenge.
2.5 Human-machine integration capability
In the actual use of drones, human participation is inseparable, and the principle of "people-centered" should always be implemented. Therefore, although most of the time "people are on the loop", the ability of man-machine fusion is still indispensable for autonomous control systems. Through the realization of human-machine fusion capabilities, communication and collaboration bridges can be established between drones and operating users, and between drones and manned systems.
The human-machine integration capability can be understood as the organic combination of intelligent system technology and platform control technology, and its realization involves human-machine interface, human-machine division of labor, human-machine collaboration and other related technical fields. The US Air Force also clearly pointed out in important reports such as "Autonomous Horizon" that man-machine symbiosis/community is an important future development direction for autonomous systems.
2.6 Multi-machine collaboration capability
Facing increasingly complex tasks and application environments, the use mode of UAV systems has gradually evolved from a single platform to a more flexible multi-platform (manned/unmanned, unmanned/unmanned) collaborative operation mode. Therefore, the autonomous control system is also Corresponding multi-machine coordination capabilities must be established according to actual task requirements.
UAV systems with multi-aircraft coordination capabilities can complete some complex tasks that a single unmanned platform cannot accomplish, such as coordinated perception, coordinated attack, and coordinated interference. To realize this ability, it is necessary to solve the problems of complexity, distribution, heterogeneity, etc. This poses a series of challenges to the related communication, information processing, management and scheduling.
2.7 Fault tolerance
Fault tolerance is the ability to automatically/autonomously handle failures. The autonomous control system of drones should have a certain degree of fault tolerance and even repair capabilities for sudden system failures, battle losses, etc., so as to be able to autonomously handle failures in flight. Implementation provides effective protection.
In order to have fault tolerance, the corresponding functions of real-time detection, isolation, recovery and prediction of faults/errors must be established in the autonomous control system of UAVs, which are mainly realized through active fault tolerance and passive fault tolerance related technologies. For example, common fault tolerance methods include fault detection/diagnosis/isolation, system redundancy/backup, self-repair/reconfiguration control, system degradation processing, etc.
2.8 Learning to evolve
The ability to learn and evolve is one of the important manifestations of the high degree of intelligence of an autonomous control system. It refers to the ability to improve the relevant performance of the system through autonomous learning, correction and continuous evolution. Among them, learning refers to the processing, processing and refining of existing experience and information to form the knowledge that one has mastered; evolution refers to the continuous iterative optimization and improvement of one's own knowledge.
With the advancement of artificial intelligence and machine learning technology, autonomous drone control systems are expected to gradually have a certain degree of learning and evolution capabilities, especially for certain specific mission scenarios, such as autonomous air combat decision-making. The ALPHA autonomous air combat simulation system has demonstrated very powerful related capabilities. The engineering realization of learning and evolution capabilities is not far away, and it is worth looking forward to.
It needs to be emphasized that the above eight capability requirements of the UAV autonomous control system are not independent of each other, but the relationship of mutual penetration, interaction and mutual promotion, which should be considered in the process of system research, design and implementation. . The first four abilities mentioned are progressive, the former is the basis of the latter, and the latter is the purpose of the former; on the basis of the first four abilities, the task-oriented man-machine integration ability and multi-machine coordination ability can be realized; The ability of fault tolerance and learning and evolution can further improve the first six abilities. Only the coordinated development of the above-mentioned multiple capabilities can achieve and promote the realization and progress of the autonomous control system.