A method for constructing a multi-task collaborative model for heterogeneous unmanned swarms

Abstract

The present invention discloses a method for constructing a multi-task collaborative model of a heterogeneous unmanned cluster. The method describes the common problems in the unmanned cluster collaborative process in a procedural manner, and can meet complex modeling requirements such as diverse task scenarios, heterogeneous unmanned platforms, multi-task integration, and different evaluation standards. In particular, for the problem of constructing a task collaborative model faced by heterogeneous unmanned clusters when performing multiple types of tasks, it is proposed to classify the controllable and uncontrollable elements in the unmanned cluster collaborative process, and classify the commonality and independence of the state space and action space respectively. Finally, the dynamic operation of the model is realized through the state transfer rules, and the multi-task collaborative model of the heterogeneous unmanned cluster is constructed. This method is of great significance for expanding the unmanned cluster modeling theory and promoting the application of unmanned cluster modeling to unmanned cluster engineering.

Description

Technical Field

[0001] The present invention belongs to the field of swarm intelligence, and in particular refers to a method for constructing a swarm collaboration model for heterogeneous unmanned clusters facing multi-task scenarios. Background Art

[0002] The main characteristics of the environment that unmanned swarms will face in the future are high dynamics, strong confrontation, unknown environment, and time sensitivity. They will need to be ready to respond to harsh conditions including sudden threats, electronic interference, and platform damage. In a diverse environment, the unmanned swarm task elements are interrelated, constrained and influenced by each other, making it very difficult for unmanned swarms to make collaborative decisions. The main issues involved include: 1) Difficulty in model establishment: Unmanned swarms are composed of many drones with different characteristics, types and purposes. The task process depends not only on their own capabilities, but also on the constraints of the natural environment and target state, which leads to a sharp increase in the difficulty of unmanned swarm collaborative modeling; 2) Difficulty in problem solving: The cluster collaborative decision-making problem is a multi-parameter, multi-constrained non-deterministic polynomial (NP) problem, which is prone to combinatorial explosion under multiple input conditions, and requires the study of efficient dimensionality reduction methods and optimization strategies; 3) High degree of task coupling: The tasks in the cluster are usually interrelated and constrained by each other, that is, there are complex constraints and different task requirements, such as time, space, load matching relationship and task priority requirements. Multi-task coupling greatly increases the difficulty of self-coordination; 4) Information loss: UAV swarms mostly perform tasks in a highly dynamic and highly confrontational environment. The time, space and state of the object are all unknown in advance, making it extremely difficult to perform task coordination in the absence of information or partial information.

[0003] The current unmanned swarm collaboration model is basically limited to the autonomous or semi-autonomous logical hierarchical architecture in a structured environment, using rigid and fixed models and processes, pre-defining scenarios and applications, resulting in limited generalization of task collaboration, insufficient adaptability to unknown environments, and poor effectiveness of task execution. Therefore, how to more efficiently and accurately establish an unmanned swarm multi-task collaboration model will be an important basis for achieving refined modeling of unmanned swarms. Summary of the invention

[0004] In view of the current problems of difficulty in constructing unmanned cluster collaborative models, diverse model elements, and weak scenario adaptability, the present invention proposes a method for constructing a heterogeneous unmanned cluster multi-task collaborative model. The method provides a process-based description of common problems in the unmanned cluster collaboration process, and can meet complex modeling requirements such as diverse task scenarios, heterogeneous unmanned platforms, multi-task integration, and different evaluation standards. It is of great significance to expand the unmanned cluster modeling theory and promote the application of unmanned cluster modeling to unmanned cluster engineering.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A method for constructing a heterogeneous unmanned cluster multi-task collaboration model includes the following steps:

[0007] (1) Classify the types of tasks that heterogeneous unmanned swarms are facing;

[0008] (2) extracting the task elements required for each task type involved in step (1);

[0009] (3) Create two sets of task elements, namely a known controllable task element set and an unknown uncontrollable task element set;

[0010] (4) Incorporating the task elements of each type of task in step (2) into the two types of task element sets created in step (3) according to known controllable elements and unknown uncontrollable elements, and repeatedly accumulating the number of occurrences of the task elements;

[0011] (5) Setting a positive integer threshold N, taking the task elements whose cumulative times in the two types of element sets in step (4) are greater than or equal to N as common elements, and taking the task elements whose cumulative times are less than N as independent elements, so as to classify the task elements in the two types of element sets;

[0012] (6) Merge the common elements and independent elements in the two types of element sets, and further construct the common state space and the independent state space;

[0013] (7) According to the changing factors of the UAV platform, payload action, target motion, and environmental change, construct the common action space and independent action space corresponding to the common state space and independent state space respectively;

[0014] (8) Construct the quantitative indicators of the impact of the two action spaces on the state space in step (7), as well as the quantitative indicators of the impact of the state space change on the task, to complete the construction of the entire collaborative model.

[0015] Furthermore, the heterogeneous unmanned cluster refers to a group of multiple unmanned systems consisting of different types of unmanned platforms, different payloads, or both, and the types of tasks it faces include collaborative reconnaissance, target search, communication coverage, and electronic countermeasures.

[0016] Furthermore, the two mission element sets created in step (3) are respectively a known controllable mission element set and an unknown uncontrollable mission element set; the known controllable mission element set includes the number of drones, payload movements, drone speed, drone attitude, communication topology, and communication bandwidth settings; the unknown uncontrollable mission element set includes weather, terrain, no-fly zones, electromagnetic environment, non-cooperative targets, and sudden obstacles.

[0017] Furthermore, the positive integer threshold N set in step (5) is:

[0018] N=[M / max(z,w)]N=[M / max(z,w)]

[0019] Where M is the number of mission types, z is the number of payload types, w is the number of unmanned platform types, max indicates the maximum value, and [] indicates rounding down.

[0020] Furthermore, in step (6), the state space is constructed as follows:

[0021] For all mission elements, all possible states that may occur at each moment during the mission execution are extracted; for known controllable elements, the state of the UAV platform includes flight speed, flight direction, flight altitude, remaining energy, and flight attitude; the state of the payload includes on, off, and running status; for unknown uncontrollable elements, the environmental state includes wind speed, wind direction, visibility, and terrain undulations; the obstacle state includes distance, size, and motion attributes.

[0022] Furthermore, in step (7), the action space is constructed as follows:

[0023] For all mission elements, all actions that can be performed in a certain state during the mission execution are extracted to form an action space. Among them, the actions of the drone include hovering, climbing, horizontal movement, moving direction, and moving speed. The actions of the payload include opening, closing, and adjusting the angle. The environmental actions include sudden obstacles, obstacle movement direction, wind speed changes, and changes in electromagnetic interference intensity.

[0024] Furthermore, in step (8), the quantitative indicators of the impact of the two action spaces on the state space are used to achieve a quantitative description of the effect of the action on the state change, including quantitative indicators of changes in the distance between drones, communication status, endurance time, and coverage position caused by the action of the drone platform; quantitative indicators of changes in the task type status, the number of corresponding targets, and the degree of task completion caused by the action of the payload; quantitative indicators of changes in the target position, speed, and coverage status of the drone caused by the action of the target; and quantitative indicators of changes in the distance between the drone and the obstacle and the height of the drone above the ground caused by the change of the environment.

[0025] The quantitative indicators of the impact of state space changes on the mission include the coverage target ratio in the communication coverage mission, the mission completion time in the collaborative search mission, the safety level of UAVs in the formation flight mission, and the UAV loss ratio in the coordinated strike mission.

[0026] The beneficial effects of the present invention are:

[0027] 1. The present invention provides a process-based description of common problems in the unmanned cluster collaboration process, especially the problem of building a task collaboration model faced by heterogeneous unmanned clusters when performing multiple types of tasks. It proposes to classify the controllable and uncontrollable factors in the unmanned cluster collaboration process, and classify the commonality and independence of the state space and action space respectively. Finally, the dynamic operation of the model is realized through the state transfer rules, thereby realizing the construction of a multi-task collaboration model for heterogeneous unmanned clusters.

[0028] 2. The present invention can meet complex modeling requirements such as diverse mission scenarios, heterogeneous unmanned platforms, multi-task integration, and different evaluation standards.

[0029] 3. This invention is of great significance for expanding the theory of unmanned swarm modeling and promoting the application of unmanned swarm modeling to unmanned swarm engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Construct a flow chart for the multi-task collaboration model of heterogeneous unmanned swarms.

[0031] Figure 2 Detailed schematic diagram of the process of building a multi-task collaboration model for heterogeneous unmanned swarms. DETAILED DESCRIPTION

[0032] The present invention is mainly used to solve the problem of constructing a task collaboration model when heterogeneous unmanned clusters perform multiple types of tasks. It classifies the controllable and uncontrollable factors in the collaboration process of the unmanned clusters, and classifies the commonality and independence of the state space and the action space respectively. The dynamic operation of the model is realized through the state transfer rules, thereby realizing the construction of a multi-task collaboration model for heterogeneous unmanned clusters.

[0033] like Figure 1 As shown, a method for constructing a heterogeneous unmanned cluster multi-task collaboration model includes the following steps:

[0034] (1) Classify the types of tasks that heterogeneous unmanned swarms are facing;

[0035] (2) extracting the task elements required for each task type involved in step (1);

[0036] (3) Create two sets of task elements, namely a known controllable task element set and an unknown uncontrollable task element set;

[0037] (4) Incorporating the task elements of each type of task in step (2) into the two types of task element sets created in step (3) according to known controllable elements and unknown uncontrollable elements, and repeatedly accumulating the number of occurrences of the task elements;

[0038] (5) Setting a positive integer threshold N, taking the task elements whose cumulative times in the two types of element sets in step (4) are greater than or equal to N as common elements, and taking the task elements whose cumulative times are less than N as independent elements, so as to classify the task elements in the two types of element sets;

[0039] (6) Merge the common elements and independent elements in the two types of element sets, and further construct the common state space and the independent state space;

[0040] (7) According to the changing factors of the UAV platform, payload action, target motion, and environmental change, construct the common action space and independent action space corresponding to the common state space and independent state space respectively;

[0041] (8) Construct quantitative indicators of the impact of the two action spaces on the state space in step (7), and include quantitative indicators of the impact of state space changes on tasks, completing the construction of the entire collaborative model.

[0042] The heterogeneous unmanned cluster described in step (1) refers to a group of multiple unmanned systems consisting of different types of unmanned platforms, different payloads, or both. The mission types it faces include different tasks such as collaborative reconnaissance, target search, communication coverage, and electronic countermeasures.

[0043] The two task element sets created in step (3) are respectively a known controllable task element set and an unknown uncontrollable task element set. The two types of elements include but are not limited to the following: the known controllable task element set mainly includes human-controllable factors, such as the number of drones, payload movements, drone speed, drone attitude, communication topology, communication bandwidth settings, etc.; the unknown uncontrollable task element set mainly includes external uncontrollable factors that affect the task, such as weather, terrain, no-fly zones, electromagnetic environment, non-cooperative targets, sudden obstacles, etc.

[0044] The positive integer threshold N set in step (5) complies with the following rule: the number of task types is M, the number of load types is z, and the number of unmanned platform types is w, then N = [M / max(z,w)], where max means taking the maximum value and [] means rounding down.

[0045] Step (6) The construction of the state space is implemented as follows: for all mission elements, all possible states that may occur at each moment during the mission execution are extracted. The known controllable elements such as the state of the UAV platform include flight speed, flight direction, flight altitude, remaining energy, flight attitude, etc., and the state of the payload includes on, off, and running states; unknown uncontrollable elements such as environmental states include wind speed, wind direction, visibility, terrain undulations, etc., and obstacle states include distance, size, motion attributes, etc.

[0046] The construction of the action space of step (7) is implemented as follows: for all task elements, all actions that can be performed in a certain state during the task execution are extracted to form an action space. For example, the actions of the drone include hovering, climbing, horizontal movement, moving direction, moving speed, etc. The actions of the payload include opening, closing, adjusting the angle, etc. The environmental actions include sudden obstacles, obstacle movement direction, wind speed changes, electromagnetic interference intensity changes, etc.

[0047] Step (8) constructs a quantitative index of the impact of action space on state space, aiming to achieve a quantitative description of the effect of action on state change, including but not limited to the UAV platform action can cause changes in the distance state between aircraft, communication connection state, endurance time state, and coverage position state; the payload action can cause changes in the task type state, the number of corresponding targets, and the task completion state; the target action can cause changes in the target position, speed, whether it is covered by the UAV, etc., and the environmental change causes changes in the distance state between the UAV and the obstacle, the UAV height above the ground, etc. The quantitative parameters of the impact of the action on the state are designed based on domain knowledge and typical task scenarios; the quantitative impact of the state space change on the task is usually related to the task type, such as the coverage target ratio in the communication coverage task, the task completion time in the collaborative search task, the UAV safety level in the formation flight task, and the UAV loss ratio in the collaborative strike task. For example, if a drone flies 3 kilometers to the left (belongs to the action space), the distance between the drone and its neighboring drones will change, the distance change will change the communication state between the drones, and the flight time will be shortened due to power consumption (all belong to the state space, which is equivalent to the drone flying 3 kilometers to the left, causing the entire cluster state to change. The specific change parameters can be set according to actual conditions). Other payload actions, target actions, and environmental actions are similar.

[0048] Here is a more specific example:

[0049] In order to accurately describe the implementation process, the necessary prerequisites are first specified: suppose the number of individuals in the target heterogeneous unmanned cluster is 20, and the unmanned cluster is composed of the same type of UAV platforms carrying different types of payloads, namely communication coverage payload, target perception payload and confrontation payload. The achievable task types are determined by the payload type, that is, communication coverage, collaborative perception and collaborative confrontation tasks can be performed, and each UAV only carries one of the payloads, that is, each UAV can only perform the corresponding type of task. In terms of quantity, there are 6 communication coverage UAVs, 7 collaborative perception UAVs, and 7 collaborative confrontation UAVs.

[0050] Based on the above constraints, the following Figure 2 This method is described in detail. A method for constructing a heterogeneous unmanned cluster multi-task collaboration model includes the following steps:

[0051] Step 1: Classify the task types that heterogeneous unmanned clusters are facing.

[0052] According to the prerequisites, it can be known that the number of individuals in the unmanned swarm is 20, including 6 communication coverage drones, 7 collaborative perception drones, and 7 collaborative confrontation drones, which are respectively equipped with communication coverage payloads, target perception payloads, and confrontation payloads, and can achieve three types of tasks: communication coverage, target perception, and collaborative confrontation.

[0053] Step 2: Extract the task elements required for each task type involved in step 1.

[0054] The communication coverage mission elements include: UAV platform, inter-machine data link, communication coverage payload, communication coverage target, obstacles, terrain, electromagnetic environment, and weather; the collaborative perception mission elements include: UAV platform, inter-machine data link, target perception payload, perception target, obstacles, terrain, electromagnetic environment, and weather; the collaborative confrontation mission elements include: UAV platform, inter-machine data link, confrontation payload, confrontation target, obstacles, terrain, electromagnetic environment, and weather;

[0055] Step 3: Create two task element sets, namely a known controllable task element set and an unknown uncontrollable task element set.

[0056] The set of known controllable mission elements mainly includes human-controllable factors, such as UAV platform, payload movement, UAV speed, UAV attitude, inter-machine data link communication topology, communication bandwidth setting, etc.; the set of unknown uncontrollable mission elements mainly includes external uncontrollable factors that affect the mission, such as weather, terrain, no-fly zones, electromagnetic environment, non-cooperative targets, sudden obstacles, etc.

[0057] Step 4: Incorporate the task elements of each type of task in step (2) into the two types of task element sets created in step (3) according to known controllable and unknown uncontrollable, and repeatedly accumulate the number of occurrences of the task elements.

[0058] The known controllable mission element set includes: UAV platform 3 times, inter-machine data link 3 times, communication coverage payload 1 time, target perception payload 1 time, and countermeasure payload 1 time; the unknown uncontrollable mission element set includes: obstacle 3 times, terrain 3 times, electromagnetic environment 3 times, weather 3 times, communication coverage target 1 time, perception target 1 time, and countermeasure target 1 time;

[0059] Step 5: Set a positive integer threshold N, and take the task elements whose cumulative times in the two types of element sets in step (4) are greater than or equal to N as common elements, and the task elements whose cumulative times are less than N as independent elements, so as to classify the task elements in the two types of element sets.

[0060] The set positive integer threshold N complies with the following rules. The number of task types is M, the number of payload types is z, and the number of unmanned platform types is w, then N = [M / max(z,w)], where max means taking the maximum value and [] means rounding down. Therefore, according to the description in step 1, the number of task types is 3, the number of payload types is 3, and the number of unmanned aerial vehicle platform types is 1. After calculation, N = [3 / max(3,1)] = 1. The common elements in the known controllable task element set include: unmanned aerial vehicle platform, inter-machine data link; the independent elements in the known controllable task element set include: communication coverage payload, target perception payload, and countermeasure payload; the common elements in the unknown uncontrollable task element set include: obstacles, terrain, electromagnetic environment, and weather; the independent elements in the unknown uncontrollable task element set include: communication coverage target, perception target, and countermeasure target;

[0061] Step 6: Merge the common features and independent features in the two types of feature sets, and further construct the common state space and the independent state space.

[0062] The common elements after the merger include: UAV platform, inter-machine data link, obstacles, terrain, electromagnetic environment, and weather; the independent elements after the merger include: communication coverage payload, target perception payload, countermeasure payload, communication coverage target, perception target, and countermeasure target.

[0063] The common state space constructed includes: 1) UAV platform: number, speed, position, and flight time; 2) inter-machine data link: bandwidth, communication distance, communication topology, and number of communication hops; 3) obstacles: size and distance from the UAV; 4) terrain: elevation information; 5) electromagnetic environment: strength distribution, frequency band; 6) weather: wind speed, visibility, rain and snow.

[0064] The constructed independent state space includes: 1) communication coverage payload: coverage range, band; 2) target perception payload: perception range, posture; 3) confrontation payload: confrontation range, band; 4) communication coverage target: quantity, location; 5) perception target: type, location; 6) confrontation target: band, location.

[0065] Step 7: According to the changing factors such as UAV platform, payload action, target motion, and environmental change, construct the common action space and independent action space corresponding to the common state space and independent state space respectively.

[0066] The common action space constructed includes: 1) UAV platform: movement direction, movement speed; 2) inter-machine data link: connected, non-connected; 3) obstacles: sudden appearance; 4) terrain: static existence, no movement; 5) electromagnetic environment: changes in strength; 6) weather: weather changes.

[0067] The constructed independent state space includes: 1) communication coverage payload: on, off; 2) target perception payload: on, off, attitude adjustment; 3) confrontation payload: on, off, band selection; 4) communication coverage target: the movement direction and speed of each target; 5) perception target: the movement direction and speed of each target; 6) confrontation target: the movement direction and speed of each target.

[0068] Step 8: Construct the transfer rules of the two action spaces and state spaces in step 7, and further construct the rules for the impact of state space changes on task completion to complete the construction of the entire collaborative model.

[0069] This step mainly sets the degree of influence of the action on the state according to the task type and expert knowledge, and refers to indicators such as coverage target ratio, task completion time, drone safety level, drone loss ratio, etc., aiming to achieve a quantitative description of the effect of action on state transition. For example, in this implementation, the action of the drone platform (movement direction, movement speed) will cause the state of the drone platform (speed, position, endurance time) to change; obstacles (sudden appearance) will affect the safety and flight direction of the drone platform; communication coverage targets (movement direction and speed of each target) will cause changes in the target position; the detailed design of this step refers to Figure 2 Transfer Rules section.

[0070] In summary, the present invention provides a procedural description of common problems in the unmanned cluster collaboration process, which can meet complex modeling requirements such as diverse mission scenarios, heterogeneous unmanned platforms, multi-task integration, and different evaluation standards. In particular, for the problem of constructing a task collaboration model faced by heterogeneous unmanned clusters when performing multiple types of tasks, it is proposed to classify the controllable and uncontrollable elements in the unmanned cluster collaboration process, and classify the commonality and independence of the state space and action space respectively. Finally, the dynamic operation of the model is realized through the state transfer rules, and the multi-task collaboration model of the heterogeneous unmanned cluster is constructed. This method solves the current problems of difficulty in constructing unmanned cluster collaboration models, diverse model elements, and weak scene adaptability. It is of great significance to expand the unmanned cluster modeling theory and promote the application of unmanned cluster modeling to unmanned cluster engineering.

Claims

1. A method for constructing a multi-task collaborative model for heterogeneous unmanned clusters. It is characterized in that The following steps are involved: (1) Classify the types of tasks that heterogeneous unmanned swarms are facing; (2) extracting the task elements required for each task type involved in step (1); (3) Create two sets of task elements, namely a known controllable task element set and an unknown uncontrollable task element set; (4) Incorporating the task elements of each type of task in step (2) into the two types of task element sets created in step (3) according to known controllable elements and unknown uncontrollable elements, and repeatedly accumulating the number of occurrences of the task elements; (5) Setting a positive integer threshold N, taking the task elements whose cumulative times in the two types of element sets in step (4) are greater than or equal to N as common elements, and taking the task elements whose cumulative times are less than N as independent elements, so as to classify the task elements in the two types of element sets; (6) Merge the common elements and independent elements in the two types of element sets, and further construct the common state space and the independent state space; (7) According to the changing factors of the UAV platform, payload action, target motion, and environmental change, construct the common action space and independent action space corresponding to the common state space and independent state space respectively; (8) Construct the quantitative indicators of the impact of the two action spaces on the state space in step (7), as well as the quantitative indicators of the impact of the state space change on the task, to complete the construction of the entire collaborative model.

2. According to the method for constructing a heterogeneous unmanned cluster multi-task collaborative model according to claim 1, It is characterized in that The heterogeneous unmanned swarm refers to a group of multiple unmanned systems consisting of different types of unmanned platforms, different payloads, or both. The mission types it faces include collaborative reconnaissance, target search, communication coverage, and electronic countermeasures.

3. According to the method for constructing a heterogeneous unmanned cluster multi-task collaborative model according to claim 1, It is characterized in that The two mission element sets created in step (3) are respectively a known controllable mission element set and an unknown uncontrollable mission element set; the known controllable mission element set includes the number of drones, payload action, drone speed, drone attitude, communication topology, and communication bandwidth setting; the unknown uncontrollable mission element set includes weather, terrain, no-fly zones, electromagnetic environment, non-cooperative targets, and sudden obstacles.

4. According to the method for constructing a heterogeneous unmanned cluster multi-task collaborative model according to claim 1, It is characterized in that The positive integer threshold N set in step (5) is: N = [M / max(z,w)] Where M is the number of mission types, z is the number of payload types, w is the number of unmanned platform types, max indicates the maximum value, and [] indicates rounding down.

5. According to the method for constructing a heterogeneous unmanned cluster multi-task collaborative model according to claim 1, It is characterized in that In step (6), the state space is constructed as follows: For all task elements, extract all possible states that may occur at each moment during task execution; for known controllable elements, the states of the UAV platform include flight speed, flight direction, flight altitude, remaining energy, and flight attitude, and the states of the payload include on, off, and operating state; for unknown uncontrollable elements, the environmental states include wind speed, wind direction, visibility, and terrain undulation, and the obstacle states include distance, size, and motion attributes.

6. A method for constructing a heterogeneous unmanned cluster multi-task cooperation model according to claim 1, characterized in that, in step (7), the action space is constructed as follows: For all task elements, extract all actions that can be executed in a certain state during task execution to form an action space. Among them, the actions of the UAV include hovering, climbing, horizontal movement, movement direction, and movement speed, the actions of the payload include turning on, turning off, and adjusting the angle, and the environmental actions include sudden obstacles, obstacle movement direction, wind speed change, and electromagnetic interference intensity change.

7. A method for constructing a heterogeneous unmanned cluster multi-task cooperation model according to claim 1, characterized in that, in step (8), the quantization indexes of the influence of the two action spaces on the state space are used to quantitatively describe the effect of the action on the state change, including the quantization indexes of the changes in the inter-aircraft distance, communication connection state, endurance time, and coverage position caused by the actions of the UAV platform, the quantization indexes of the changes in the task type state, the corresponding target quantity, and the task completion degree caused by the payload actions, the quantization indexes of the changes in the target position, speed, and UAV coverage state caused by the target actions, and the quantization indexes of the changes in the distance between the UAV and the obstacle and the UAV's height from the ground caused by the environmental change; The quantization indexes of the influence of the state space change on the task include the coverage target ratio in the communication coverage task, the task completion time in the collaborative search task, the UAV safety level in the formation flight task, and the UAV loss ratio in the collaborative strike task.

Images