A meta-task management method and system for variable autonomy level of unmanned system

Abstract

A meta-task management method and system for variable autonomy levels of unmanned systems, the main steps include: establishing a meta-task set of unmanned systems; establishing an autonomy level set and a task execution plan set of unmanned systems in combination with the working background of the unmanned system, the closeness of human-machine collaboration and the workload distribution; establishing a meta-task base state with a hierarchy system for the meta-task execution plan set of the unmanned system, and establishing a state sequence with level transfer restrictions; establishing a meta-task-oriented unmanned system state monitoring model and evaluation plan; establishing a real-time optimal autonomy level reasoning model based on unmanned system state monitoring; and obtaining a meta-task state autonomous transfer architecture of the unmanned system autonomy level. This method can improve the unmanned system's understanding and control of the autonomy level, request operators to participate in task execution in a more efficient way, and better balance the workload between operators and unmanned systems.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned systems, and in particular to a meta-task management method for unmanned systems with variable autonomy levels. Background Art

[0002] The application of unmanned systems (such as unmanned vehicles, drones, and unmanned ships) has gradually penetrated into social life, and its application background covers various civil or military fields such as logistics warehousing and field reconnaissance. The key technology for unmanned systems with certain autonomous capabilities to replace humans in performing high-risk tasks and detecting dangerous areas is that the system can manage itself. The National Institute of Standards and Technology of the United States has proposed the following definition of the autonomy of unmanned systems: Autonomy is the ability of unmanned systems to perceive, observe, analyze, communicate, plan, make decisions and act, and complete the tasks assigned to it by humans through human-computer interaction. Autonomy can be differentiated according to factors such as the complexity of the task, the difficulty of the environment, and the degree of human-computer interaction to complete the task, thereby indicating the state and quality of the self-management of the unmanned system.

[0003] There are currently two methods for managing the autonomy level of unmanned systems. One is the system-active method, in which the operator does not monitor the progress of the task. Only when the system actively seeks help can the operator intervene to help complete the task until the problem is solved and release control. In this mode, the operator can perform tasks in parallel with the unmanned system. The tasks faced by this mode must not be time-sensitive. If the tasks that arise require urgent help or the system itself has a perception delay, then this mode is not applicable. One is the hybrid active method, which requires the operator to monitor the progress of the task in real time and have the right to actively take over the system authority. This active control capability allows the operator to avoid clear faults before the system detects them. However, the operator needs to devote himself to the entire task, which greatly increases the operator's workload. Summary of the invention

[0004] The present disclosure provides an autonomy level management mechanism and system application for unmanned system meta-tasks, aiming to improve the unmanned system's understanding and grasp of the autonomy level and request operators to participate in task execution in a more efficient manner, which not only satisfies the unmanned system's ability to adjust the autonomy level, but also reduces the operator's monitoring burden and makes the entire unmanned system stable and safe, greatly balancing the workload between the operator and the unmanned system, and ultimately ensuring that the unmanned system and the operator can complete the task with high quality.

[0005] The autonomous level management method of the unmanned system provided by the present disclosure mainly includes the following steps:

[0006] S1, establish a set of meta-tasks for unmanned systems;

[0007] S2, combining the working background of the unmanned system, the closeness of human-machine collaboration and workload distribution, establish the set of autonomous levels of the unmanned system and the set of meta-task execution plans;

[0008] S3, for the set of meta-task execution plans of the unmanned system, establish the meta-task base state with the autonomy hierarchy system, and establish the state sequence with the autonomy level transition restriction;

[0009] S4, establish a task-oriented unmanned system status monitoring model and evaluation scheme;

[0010] S5, establish a real-time optimal autonomy level reasoning model for unmanned systems based on condition monitoring;

[0011] S6, obtain the autonomous transfer architecture of the task state of the unmanned system autonomy level.

[0012] Further, step S1 is specifically:

[0013] S11, establish the hierarchical task architecture of the unmanned system, and divide the operation process of the unmanned system into four stages: observation, positioning, decision-making, and behavior according to the OODA cycle. Each stage contains a series of action sets, such as: the observation action of the unmanned system represents reconnaissance work such as data collection and sensing; the positioning action of the unmanned system represents understanding work such as data analysis and processing; the decision-making action of the unmanned system represents planning work such as behavior planning and the next step deployment based on cognitive results; the behavior action of the unmanned system represents implementation work such as implementation based on planning results.

[0014] All achievable tasks of the unmanned system are decomposed into four levels: observation, positioning, decision-making, and behavior according to the nature of the meta-tasks. Each level is composed of a certain number of minimum meta-tasks. The meta-tasks of all levels are aggregated to form the task set of the unmanned system.

[0015] S12, the specific tasks of the unmanned system are disassembled, and each action is classified into one of the unmanned system observation action, the unmanned system positioning action, the unmanned system decision action, and the unmanned system behavior action. The actions of the same task at the same level can form an action sequence, and different tasks at the same level can form different action sequences.

[0016] Furthermore, step S2 is specifically as follows:

[0017] S21: According to the task set and combined with the task background, analyze the urgency and difficulty of each task, divide the workload distribution of operators and unmanned systems in the OODA loop, and set different autonomy levels for different workloads. At the same time, the highest autonomy level is set as the fully autonomous mode with the lowest operator workload, and the lowest autonomy level is set as the fully manual mode with the lowest unmanned system workload. In the middle, set the semi-autonomous mode with reference to different workload ratios.

[0018] S22: Since the sub-state set under the same task is the same, but the sub-state combination sequence under different autonomy levels is inconsistent, that is, the meta-task execution order is inconsistent. It is necessary to determine the state sequence of each task at any autonomy level to ensure that the task can be executed normally at any autonomy level, forming a complete task solution collection containing all levels.

[0019] Furthermore, step S3 is specifically as follows:

[0020] S31: First, based on the ROS-Smach state machine, multiple functional layers are added on the basis of the state template as the basic functions of the state, including loops, forced jumps, data monitoring and other sub-functions, and they are called base states. Then each base state can execute commands normally and realize normal switching between tasks. At the same time, during the execution of tasks, it can monitor the changes in the system's autonomy level as the current level of the state.

[0021] S32: Then, according to the specific task, each meta-task is implemented on the base state, thereby forming a state sequence of the entire task. A link transfer relationship is implemented between each two sub-states in the state sequence, and a transfer level buffer is added. The operator determines the level limit required for jumping between two sub-states in the preset stage; after the sub-state completes the expected task of the current state, it monitors whether the current level meets the jump limit. If it does, it enters the next sub-state, otherwise it enters the self-loop state and waits for the operator's instructions or until the current level meets the jump limit.

[0022] Furthermore, step S4 is specifically as follows:

[0023] S41: Establish the state model of the unmanned system, including the four levels of the OODA loop. Observation level: various sensor fault diagnosis status, input signal stability; Positioning level: collision safety status, environmental complexity analysis; Decision level: task clarity and task completion; Behavior level: actuator fault diagnosis status, behavior safety prediction.

[0024] S42: Establish an evaluation scheme for the indicators in the state model according to the specific task type, extract characteristic values ​​with evaluation meanings by referring to expert experience or historical data formulas, and limit them to [0,1].

[0025] Furthermore, step S5 is specifically as follows:

[0026] S51: According to each monitoring state data in the unmanned system state model, different reasoning models are established for different monitoring state combinations, which are divided into hard processing reasoning and soft processing reasoning.

[0027] S52: Hard processing reasoning means that after a hardware sensor or hardware actuator fails, it analyzes whether the current meta-task will be affected by the failure. If it can be solved by semi-manual intervention of the operator, the target autonomy level is outputted in a directional manner. Otherwise, the lowest autonomy level is directly outputted and fully handled manually by the operator.

[0028] S53: Soft processing reasoning represents measuring the current task index of the unmanned system, extracting the characteristic value of each index, using expert experience or historical experience data as different interval thresholds, adding the characteristic values ​​and comparing them with the interval thresholds, and selecting the characteristic value and the interval it belongs to as the current optimal autonomy level.

[0029] Furthermore, step S6 is specifically as follows:

[0030] S61: Establish the unmanned system autonomy level transfer rules, stipulating that when the optimal autonomy level is lower than the current level, the unmanned system automatically adjusts the current level to the optimal autonomy level; when the optimal autonomy level is higher than the current level, the unmanned system can only issue inquiries and adjust the autonomy level only after the operator's command allows.

[0031] S62: Design a task state autonomous transfer framework. For a specific task solution set, decompose it into meta-tasks under the OODA loop. Each level of the meta-task solution set includes solutions at different levels. Parallel representation of all execution plans for the same task forms a parallel solution. A level selection base state is established before the parallel solution, which is responsible for selecting the corresponding execution plan according to the current autonomous level.

[0032] S63: Finally, each meta-task state is registered in the state machine, including the level selection base state corresponding to the meta-task when the level changes, and the level limit of state transfer between meta-tasks. When the meta-task state detects that the current level has changed, the state automatically jumps to the level selection base state corresponding to the current meta-task, and a more appropriate task solution takes over the current meta-task. At the same time, when the meta-task successfully executes the current state, it needs to access the registration content first to check the level limit required for transferring to the next meta-task state. If it is not satisfied, it stays in the current meta-task.

[0033] The unmanned system autonomous level management system using the above method includes: an unmanned system level scheme set, an unmanned system state model, an unmanned system level reasoning model, and an unmanned system autonomous transfer architecture, wherein:

[0034] The unmanned system level scheme set is used to allocate the task workload of the unmanned system and the operator, set multiple autonomy levels with different allocation ratios, and select appropriate levels from the full level scheme set to form a level scheme to form a multi-autonomy level unmanned system.

[0035] The unmanned system state model is used to select sub-detection states related to the current task from the four levels of the unmanned system's OODA loop. It is responsible for storing the models and parameters of all measurement indicators and providing a data source for subsequent autonomous hierarchical reasoning models.

[0036] The unmanned system level reasoning model is used to jointly reason about the state model and current level of the unmanned system to detect whether the current autonomy level meets the characteristic values ​​of the state model. The module also saves the threshold of the metric and inputs the reasoning results into the next step of the autonomy level transfer architecture.

[0037] The autonomous transfer architecture of unmanned systems is used to establish the level transfer rules of unmanned systems, embed the level transfer mechanism into the system task state, allow the task states to transfer to each other, and establish a parallel task plan for each task. The plan is selected by selecting the base state, supporting the unmanned system to select different level plans to solve the current task under any task.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) the concept of autonomy level is introduced into the meta-task management of unmanned systems, and the meta-task execution plans of unmanned systems are managed in a hierarchical manner; (2) the autonomy level management architecture based on the meta-task of the unmanned system is based on the basic idea of ​​the state machine. Without affecting the global architecture, new base states and state sequences are formed by adding state functions, and the specific execution plan is confirmed by selecting the base state by level, thereby realizing the autonomous level transfer; (3) by realizing real-time analysis and hierarchical reasoning of specific tasks, the dynamic allocation of the burden ratio of the human-machine system is realized, and the meta-task can be executed dynamically and smoothly; (4) an autonomy level management method and system based on the state machine is established, which can be oriented to different unmanned systems and tasks, and thus has a certain degree of universality. It can be applicable to any unmanned system built around the state machine for any task, and plays a positive role in autonomy level management. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The above and other objects, features and advantages of the present disclosure will become more apparent through a more detailed description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, wherein like reference numerals generally represent like components throughout the exemplary embodiments of the present disclosure.

[0040] Figure 1 is a flow chart of an unmanned system autonomy level management method according to the present disclosure;

[0041] Figure 2 Added schematic diagram for exemplary task state machine base state function;

[0042] Figure 3 Decomposition diagram of the OODA loop for ground reconnaissance missions for unmanned systems;

[0043] Figure 4 A schematic diagram of the meta-state sequence of an unmanned system with transfer restrictions;

[0044] Figure 5 A collection of ground reconnaissance mission options for unmanned systems;

[0045] Figure 6 This is a schematic diagram of the unmanned system status monitoring model;

[0046] Figure 7 Schematic diagram of the autonomous transfer architecture of the unmanned system meta-task state machine. DETAILED DESCRIPTION

[0047] The preferred embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0048] The present disclosure provides a meta-task management method for an unmanned system with a variable autonomy level, as shown in the attached Figure 1 As shown in the figure, firstly, a meta-task set of the unmanned system is established; then, in combination with the working background of the unmanned system, the closeness of human-machine collaboration and the workload distribution, an autonomy level set and a meta-task execution plan set of the unmanned system are established; for the meta-task execution plan set of the unmanned system, a meta-task base state with a hierarchy is established, and a state sequence with hierarchy transfer restrictions is established; a meta-task-oriented unmanned system state monitoring model and evaluation plan are established; an inference model of the real-time optimal autonomy level of the unmanned system is established; finally, a meta-task state autonomous transfer architecture of the unmanned system autonomy level is established.

[0049] More specifically, the following steps are included:

[0050] S1: Establish a meta-task set for unmanned systems. The unmanned systems are divided according to their platform attributes, including but not limited to unmanned vehicles, drones, and unmanned ships. Here, an unmanned vehicle is used as an example, which has the ability to perform ground reconnaissance, target identification, path planning, tracking and navigation, and other tasks.

[0051] S11: Establish a hierarchical task architecture for unmanned systems, which is divided into four levels: observation, judgment, decision-making, and action according to the OODA cycle.

[0052] S12: All achievable tasks of the intelligent ground unmanned system are allocated to four levels according to the nature of the meta-tasks. The observation stage includes sensor driving and sensor data reading. The judgment level includes autonomous positioning, mapping, semantic information generation, target recognition, and target position input. The decision-making level includes path planning, mission planning, navigation planning, and tactical planning. The action level includes chassis driving, path following, robotic arm operation, artillery attack, and remote control operation. The meta-tasks of all levels are summarized to form the task set of the unmanned system.

[0053] As attached Figure 3 As shown in the figure, the ground reconnaissance mission of the unmanned system is broken down into four levels: observation, judgment, decision-making, and action. Observation includes sensor driving and data reading, judgment includes autonomous positioning and target recognition, decision-making includes path planning and navigation planning, and action includes ground driving and path following. Meta-task actions are connected to form an action sequence.

[0054] S2: Establish a set of autonomy levels and a set of meta-task execution plans for specific tasks of unmanned systems;

[0055] S21: As attached Figure 5 As shown in the figure, the ground reconnaissance mission is set to three autonomous levels according to the workload ratio, namely, fully autonomous mode, semi-autonomous mode, and fully manual mode. In the fully autonomous mode, the operator burden is 0, in the fully manual mode, the operator burden is 100, and in the semi-autonomous mode, the mission is completed in cooperation.

[0056] S22: Since the meta-task set under the same task is the same, but the execution sequence of meta-tasks under different autonomy levels is inconsistent. In the fully autonomous mode, the execution order is: sensor data input-autonomous mapping and positioning-autonomous target recognition-autonomous navigation-autonomous following-autonomous recovery. In the semi-autonomous mode, the execution order is: sensor data input-autonomous mapping and positioning-operator target input-autonomous path planning-operator path confirmation-autonomous following-operator confirmation recovery. In the fully manual mode: sensor data input-operator remote control.

[0057] S3: Establish a task-based state with a hierarchy to form a state sequence with hierarchical restrictions;

[0058] S31: As attached Figure 2 As shown in the figure, based on the state machine provided by ROS-Smach, multiple functional layers are added as the basic functions of the state by modifying the state template code, including forced loop, forced jump, data monitoring and other functions, and they are called base states. Each base state can execute commands normally and realize sequential switching between tasks. At the same time, during execution, it can monitor the changes in the system's autonomy level as the current level of the state.

[0059] S32: Implement each meta-task on the base state, thereby forming a sub-state sequence of the entire task. A link transfer relationship is implemented between each two sub-states in the sub-state sequence, and a transfer level buffer is added. The operator determines the level limit required for jumping between two sub-states at the preset stage, as shown in the attached figure. Figure 4 After completing the expected task of the current state, the sub-state monitors whether the current autonomy level meets the jump limit. If so, it enters the next sub-state, otherwise it enters the self-loop state and waits for the operator's instructions or until the current level meets the jump limit.

[0060] S4: Establish a condition monitoring model and evaluation scheme for intelligent ground unmanned systems.

[0061] As attached Figure 6 As shown in the figure, a state monitoring model of an intelligent ground unmanned system is established. According to the mission background of ground reconnaissance, the state model contains four levels of the OODA loop. Observation level: various sensor fault diagnosis status, input signal stability; judgment level: collision safety status, environmental complexity analysis; decision level: task clarity and task completion; action level: actuator fault diagnosis status, behavior safety prediction.

[0062] Set the sensor fault diagnosis status to a Boolean value, including the operating status of the laser radar, camera, odometer, etc. 1 is the normal operating status and 0 is the fault status.

[0063] Input signal stability is a Boolean value, indicating whether the input data is invalid due to large-scale fluctuations. 1 is a normal stable state, and 0 is an unstable state.

[0064] The collision safety status is a Boolean value, indicating whether the intelligent ground unmanned system has contacted and collided with the environment, affecting the hardware safety status of the system. 1 is recorded as no expansion state, and 0 is collision state.

[0065] Environmental complexity analysis CpxE is expressed as a floating point number, which indicates whether the complexity of the environment around the intelligent ground unmanned system will affect the path planning and autonomous following. The formula is as follows:

[0066] CpxE=∑Obstacle occupancy of the current 30 surrounding sectors

[0067] Among them, this formula divides the 360-degree environment of the intelligent unmanned platform into 30 equal areas, each of which has a weight of 1 / 30. If there is an obstacle in the area, it is set to 1, otherwise it is 0. The higher the value, the more complex the environment.

[0068] The mission clarity ClrT is a floating point number, which represents the reliability of the intelligent ground unmanned system to follow the identified target, indicating whether the identified target can be finally reached.

[0069] ClrT = ∑ obstacle occupancy in 30 sectors around the target point

[0070] Among them, this formula divides the 360-degree environment of the target to be identified into 30 equal areas, each of which has a weight of 1 / 30. If there is an obstacle in the area, it is set to 1, otherwise it is 0. The higher the value, the more difficult it is to follow the target.

[0071] The actuator fault diagnosis status is a Boolean value, including the operating status of the motor drive and the unmanned system chassis. 1 is recorded as normal operation status and 0 is recorded as fault status.

[0072] S5: Build an autonomous hierarchical reasoning model for specific tasks of intelligent ground unmanned systems.

[0073] According to the monitoring state data in the unmanned system state model, different reasoning models are established for different monitoring state combinations, including hard processing and soft processing.

[0074] The hard processing reasoning model indicates that after the parameters with Boolean values ​​in the state model are transformed from normal operation to failure, the first step is to match whether the currently executed task will be affected. If not, the autonomy of the current task is maintained, and then it is further determined whether the current task can be partially completed through operator intervention. If so, it jumps to semi-autonomous mode, otherwise it jumps to full manual mode.

[0075] The soft processing reasoning model represents the parameters in the state model whose values ​​are floating point numbers. Take the environmental complexity analysis CpxE and task clarity ClrT as examples: It is known that CpxE and ClrT fluctuate between [0,2]. The higher the value, the more difficult the task is to perform. According to expert experience, the full manual interval is set to [1.6,2.0], the semi-autonomous interval is set to [0.6,1.6], and the full autonomous interval is set to [0,0.6]. The interval after the eigenvalues ​​are added is selected as the optimal autonomy level. At the same time, the single eigenvalue is restricted. For example, when Clr is greater than 0.9, it jumps to the full manual mode for the operator to check the situation.

[0076] S6: Establish an autonomous transfer architecture of the meta-task state of the autonomous level of the intelligent unmanned ground system, embed the target transfer state pool and the transfer authority state pool in the meta-task state machine, establish an autonomous level selection base state, and select the execution plan in step S2 according to the autonomous level obtained in step S5 to execute the meta-task.

[0077] S61: First, establish the unmanned system autonomy level transfer rules, stipulating that when the optimal autonomy level is lower than the current level, the unmanned system automatically adjusts the current level to the optimal autonomy level; when the optimal autonomy level is higher than the current level, the unmanned system can only issue inquiries and can adjust the autonomy level only after the operator's command allows.

[0078] S62: Design a task state autonomous transfer framework. For a specific task solution set, decompose it into meta-tasks under the OODA loop. Each level of the meta-task solution set includes solutions at different levels. Parallel representation of all execution plans for the same task forms a parallel solution. A level selection base state is established before the parallel solution, which is responsible for selecting the corresponding execution plan according to the current autonomous level.

[0079] S63: Register each meta-task state in the state machine, including setting the level selection base state corresponding to the meta-task when the level changes, and the level limit of state transfer between meta-tasks. When the meta-task state detects that the current level has changed, the state automatically jumps to the level selection base state corresponding to the current meta-task, and a more appropriate task execution plan takes over the current meta-task. At the same time, when the meta-task successfully completes the current state, it needs to access the registration content first to check the level limit required for transferring to the next meta-task state. If it is not satisfied, it stays in the current meta-task.

[0080] The above technical scheme is only an exemplary embodiment of the present invention. For those skilled in the art, it is easy to make various types of improvements or modifications based on the application methods and principles disclosed in the present invention, and it is not limited to the method described in the above specific embodiment of the present invention. Therefore, the method described above is only preferred and does not have a restrictive meaning.

Claims

1. A meta-task management method for variable autonomy levels of unmanned systems, The following steps are involved: S1, based on the OODA loop structure, split the specific tasks of the unmanned system into different task stages, and further determine the meta-tasks contained in each task stage; S2, determine the set of autonomy levels for the unmanned system to perform the task, and the execution plan of each meta-task at each autonomy level; S3, based on the state machine theory, establishes a meta-task base state with a hierarchy, and establishes a state sequence with hierarchical transition restrictions; S4, setting the status monitoring variables of the unmanned system in different mission stages, including Boolean and numerical types, to measure the mission execution status under the current autonomy level; S5, establish an autonomy level reasoning model for a specific task of an unmanned system, which is used to indicate the switching of the autonomy level when the system state meets certain conditions during the execution of the task at the current autonomy level; S6, according to steps S1-S5, establish an autonomous level transfer architecture for the unmanned system, embed the target transfer state pool and the transfer authority state pool in the meta-task state machine, establish an autonomous level selection base state, and select the execution scheme in step S2 according to the autonomous level obtained in step S5 to perform the meta-task execution; The step S6 specifically includes: Establish the autonomous level transfer rules for unmanned systems, stipulating that when the optimal autonomy level is lower than the current level, the unmanned system automatically adjusts the current level to the optimal autonomy level; when the optimal autonomy level is higher than the current level, the unmanned system sends a request to the operator and upgrades to the optimal autonomy level after obtaining permission; Design a task state autonomous transfer framework: After determining the set of solutions for a task, decompose it into meta-tasks in the four task stages of the OODA cycle, and then determine the execution plans of the meta-tasks in each stage at different autonomy levels. List all the execution plans representing the same meta-task to form a parallel plan, and then establish an autonomy level selection base state for each meta-task, which selects the corresponding execution plan according to the current autonomy level; Merging all meta-task states into the state machine, including: registering each meta-task state in the state machine, the registration content including the autonomous level selection base state corresponding to the meta-task and the autonomous level restriction of state transfer between meta-tasks; When the meta-task state detects that the current level has changed, the state automatically jumps to the level selection base state corresponding to the current meta-task, and a more appropriate execution scheme takes over the current meta-task. At the same time, when the meta-task successfully completes the current state, it needs to access the registration content first to check the level restriction required to transfer to the next meta-task state. If it is not met, it stays in the current meta-task.

2. The method according to claim 1, It is characterized in that The step S1 specifically includes: Based on the OODA cycle, the tasks of the unmanned system are divided into four stages: observation, positioning, decision-making, and action; The tasks of the unmanned system are decomposed into a series of meta-tasks, and are respectively classified into the corresponding task stages. The meta-tasks do not need to be decomposed further, can keep the state of the unmanned system or its execution unit relatively stable, and are directly executed by the object.

3. The method according to claim 1, It is characterized in that The step S2 specifically includes: For specific tasks of unmanned systems, feasible autonomy levels are set, including: fully autonomous mode with the highest autonomy level at the lowest operator workload, fully manual mode with the lowest autonomy level at the lowest unmanned system workload, and semi-autonomous mode with different workload ratios in between; Determine the execution plan for each meta-task at any autonomy level, ensure that the meta-task can be executed normally at any autonomy level, and form a complete set of meta-task execution plans covering all autonomy levels.

4. The method according to claim 1, It is characterized in that The step S3 specifically includes: Based on the ROS-Smach state machine, multiple functional layers including loop, forced jump, and data monitoring are added on the basis of the state template as the basic functions of the state. As the base state, each base state can execute commands normally and realize the normal switching between meta-tasks. At the same time, during the execution of the task, the changes in the system autonomy level can be monitored as the current level of the state; According to the specific task, each meta-task is implemented on the base state to form the state sequence of the entire task. A link transfer relationship is implemented between every two sub-states in the state sequence, and a transfer level buffer is added. The operator presets the level limit required for jumping between two sub-states. The level limit is used for: after the sub-state has completed the expected task of the current state, it monitors whether the current level meets the jump limit. If it does, it enters the next sub-state, otherwise it enters the self-loop state and waits for the operator's instructions or until the current level meets the jump limit.

5. The method according to claim 1, It is characterized in that The step S4 specifically includes: S41: Based on the four task layers of the OODA loop, a state monitoring model for unmanned systems is established, where the observation layer: monitors the fault diagnosis status of various sensors and / or the stability of input signals; the judgment layer: monitors the collision safety status and / or environmental complexity analysis; the decision layer: monitors the task clarity and / or task completion; the action layer: monitors the actuator fault diagnosis status and / or behavior safety prediction; S42: According to the specific task type, an evaluation scheme is established for the indicators in the condition monitoring model, and characteristic values ​​with evaluation meanings are extracted by referring to expert experience or historical data formulas, and are limited to between [0,1].

6. The method according to claim 1, It is characterized in that The step S5 specifically includes: According to the various monitoring indicators in the unmanned system status monitoring model, different reasoning models are established for different monitoring status combinations, including hard processing and soft processing, among which: Hard processing reasoning model: After determining that a hardware failure has occurred in the unmanned system based on monitoring indicators, analyze whether the currently executed task will be affected. If not, maintain the autonomy of the current task; otherwise, determine whether the current task can be partially completed through operator intervention. If so, jump to semi-autonomous mode, otherwise jump to full manual mode; Soft processing reasoning model: Measure the current task indicator metrics in the unmanned system, extract the characteristic values ​​of each indicator metric, add them and compare them with the interval threshold under a specific level of a specific task. If the current value is lower than the threshold of the high autonomy level and falls into the low level value interval, the current autonomy level of the unmanned system will be changed.

7. An unmanned system meta-task autonomy level management system using the method described in any one of claims 1 to 6, It is characterized in that include: Unmanned system level scheme collection module, unmanned system state model module, unmanned system autonomous level reasoning module, unmanned system autonomous transfer architecture module, including: The unmanned system level scheme collection module is used to set multiple autonomy levels with different allocation ratios for the task workload distribution of the unmanned system and the operator, and select the appropriate autonomy level from the full level scheme set to form a level scheme to form a multi-autonomy level system unmanned system; The unmanned system state model module is used to select sub-detection states related to the current task from the four levels of the unmanned system's OODA loop, and is responsible for storing the models and parameters of all metrics, providing a data source for subsequent autonomous hierarchical reasoning models; The unmanned system level reasoning model is used to jointly reason about the state model and current level of the unmanned system to detect whether the current autonomy level meets the characteristic value of the state model. This module also saves the threshold of the metric and inputs the reasoning result into the next step of the autonomy level transfer architecture. The autonomous transfer architecture of unmanned systems is used to establish the level transfer rules of unmanned systems, embed the level transfer mechanism into the system task state, allow the task states to transfer to each other, and establish a parallel task plan for each task. The plan is selected by selecting the base state, supporting the unmanned system to select different level plans to solve the current task under any task.

Images