A UAV Distributed Decision-making Method Based on a Dynamic Task List

Abstract

The present invention discloses a method for distributed decision-making of unmanned aerial vehicles (UAVs) based on a dynamic task list, which is applied to an environment where multiple UAVs cooperate to autonomously sense and perform support tasks. The present invention includes the following steps: the UAVs update their own positions in real time and detect ground targets; the UAVs obtain the target information sensed by neighboring UAVs in real time; based on the local sensing information obtained by the UAVs themselves and neighboring UAVs, a support task list for the UAVs themselves is determined, where the content in the support task list is sorted from largest to smallest according to the defined value, and those with higher values are given priority to go; the support task list shared by neighboring UAVs is obtained in real time, and the tasks currently being executed by the UAVs themselves are adjusted. The present invention can ensure that each UAV autonomously executes tasks based on local information and avoid task conflicts with other UAVs.

Description

Technical Field

[0001] The present invention belongs to the field of multi-agent cooperation, and particularly refers to a method based on distributed decision-making of unmanned aerial vehicles (UAVs), which can be applied to multi-UAV cooperation and the autonomous perception of UAV swarms to support the environment of ground targets. Background Art

[0002] With the application of unmanned equipment in multiple fields such as civilian and military, the importance of UAV swarm control has become increasingly prominent. Compared with manned aircraft or manual task execution, decision-makers using UAV means do not need to consider the limitations brought by the safety of pilots' lives, technical proficiency, etc., and can assign UAV swarms to perform various tasks that are dull, in harsh environments, highly dangerous, and deep behind the target, achieving unexpected task effects.

[0003] At present, UAV swarm control is mainly divided into two forms: centralized control and distributed control. In a large-scale confrontation environment, the control center of centralized control is easily targeted by the opponent. Once the control center is damaged, it will lead to the paralysis of the overall system function of our side. When performing confrontation tasks, UAV swarms often face complex situations with multiple targets and multiple tasks. In the absence of a control center, how to comprehensively judge the environmental situation and make reasonable autonomous decisions on target allocation and task planning is a major challenge for the large-scale application of UAV swarms.

[0004] In the distributed decision-making method, since each individual is a decision-maker and the information it obtains mainly depends on the information shared by its own aircraft and neighboring UAVs, the local information on which each UAV's decision depends is different, and it is difficult to ensure that there are no conflicts between the UAV decision results. For support tasks, support position conflicts often lead to risks such as ineffective support or UAV collisions. Therefore, how to resolve conflicts in the distributed decision results of all UAVs has become an important issue for UAV swarms when facing support tasks. Summary of the Invention

[0005] In view of this, the present invention proposes a UAV distributed decision-making method based on a dynamic task list, which arranges several larger support positions according to the value quantity of the support positions generated by the decision, ensures that UAVs support more ground targets, and finally, during the UAV support process, receives information shared by one-hop neighboring UAVs in real time and dynamically adjusts the local task list to avoid support position conflicts between multiple UAVs.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] A UAV distributed decision-making method based on a dynamic task list, comprising the following steps:

[0008] Step 1: Initialize the position of the UAV so that it covers the target points in the environment and maintains overall communication connectivity.

[0009] Step 2: Through the sensing function, the UAV obtains the position information of the ground target in real time and publishes the sensing information to the neighboring UAVs within the one-hop communication range.

[0010] Step 3: The UAV receives the sensing information shared by the neighboring UAVs within the one-hop communication range in real time.

[0011] Step 4: Grid the ground area, and determine the value of each position in the grid map based on the UAV position, sensing information, and sensing information from neighboring UAVs.

[0012] Step 5: According to the value size, screen the N target locations with the largest value, arrange them from largest to smallest, and construct a list of locations to be supported, where N is the number of UAVs in the scenario.

[0013] Step 6: Each UAV travels to the location with the highest value in the list of locations to be supported and makes real-time adjustments according to the update of the list of locations to be supported.

[0014] Step 7: During the process of the UAV going to support, it receives the list of locations to be supported from neighboring UAVs in real time, and combines the list of locations to be supported of the local UAV and neighboring UAVs to adjust the current list of locations to be supported of the local UAV.

[0015] Furthermore, the specific method of Step 4 is as follows:

[0016] 4a) Integrate the sensing information of the local UAV and the sensing information from neighboring UAVs to obtain all the sensing information obtained by the local UAV.

[0017] 4b) Grid the entire scenario according to the support radius of the UAV.

[0018] 4c) Count the number of targets that can be supported at each grid position, and accumulate the value of the supported targets to obtain the value of this position.

[0019] 4d) Based on the value map, combined with distance attenuation, obtain the value map of the local UAV support position; the specific method is as follows:

[0020] Assume that the maximum support distance of the UAV is maxL, and the distance of the UAV relative to each grid position is d, then the local support value of this grid position is:

[0021]

[0022] Where:

[0023]

[0024] V map is the initial map value quantity, V UAV is the value quantity of the local support position to be obtained.

[0025] Furthermore, the specific method of step 7 is as follows:

[0026] 7a) Compare the list of local positions to be supported with the list of positions to be supported by neighboring drones, and determine whether there is a conflict between the position to be supported that the local machine is going to and the positions to be supported that all neighbors are going to. If not, continue to go to this position to be supported;

[0027] 7b) If there is a conflict, determine the priority of the tasks selected by the local machine and neighboring drones through the drone numbers. If the local machine has the highest priority, continue to go to this position to be supported, and let other drones adjust based on the list of local positions to be supported;

[0028] 7c) If there are other drones with higher priority than the local machine, remove the first position to be supported in the list of local positions to be supported, shift the local positions to be supported to the next one, and continue to go;

[0029] 7d) During the process of the drone going to the position to be supported, continuously perform the judgment operations of steps 7a), 7b), and 7c), continuously update and adjust the support positions of the drones, and ensure that the positions where the drones go do not conflict.

[0030] The present invention has the following advantages compared with the prior art:

[0031] 1. The present invention introduces distance attenuation, not only relying on the value quantity map generated by local targets to be supported, but also adding distance reference, so that the drone can go to the effective position as close as possible;

[0032] 2. The decision generates multiple support positions and manages them in an ordered list. During the operation of the drone, the local support position is effectively adjusted in real time to avoid task position conflicts caused by local information decisions of multiple drones. Brief Description of the Drawings

[0033] Figure 1 is the overall flowchart of a drone distributed decision-making method based on a dynamic task list in an embodiment of the present invention;

[0034] Figure 2 is the coverage of ground targets by a drone swarm in a communication networking state;

[0035] Figure 3(a) is the value quantity map;

[0036] Figure 3(b) is the value quantity map relative to the local machine;

[0037] Figure 4 Schematic diagram of conflict resolution mechanism;

[0038] Figure 5 Schematic diagram of possible mission conflicts for two UAVs in a non - communication connection state. Specific implementation manners

[0039] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings.

[0040] A distributed decision - making method for UAVs based on a dynamic task list includes steps such as initialization, information sharing, relative value map establishment, task location list generation, real - time update and adjustment:

[0041] In the initialization step, mainly based on the operation range, the positions of the UAVs are initialized to cover the target points in the environment to a large extent and keep communication connectivity among the UAVs;

[0042] In the information sharing step, it is used for UAVs to publish their own information and receive information from one - hop neighbor UAVs. The shared information mainly includes: UAV number, UAV position, target point positions obtained by perception, task location list obtained by decision - making, etc. Based on this shared information, UAVs can obtain local scene information outside their own detection range and the operation status of other UAVs. Moreover, task sharing can avoid task conflicts in multi - UAV decision - making;

[0043] In the relative value map establishment step, mainly based on the results of UAV detection or detection of target points from neighbor UAVs, the ground target distribution is obtained, and a value map is generated to quantify the benefit magnitude generated by support at each position. The specific method is as follows:

[0044] (1) Integrate the perceived target position information of the local UAV and the perceived target positions of neighbor UAVs to obtain all the perceived target position information obtained by the local UAV;

[0045] (2) Grid the entire scene according to the UAV support radius;

[0046] (3) Count the number of targets that can be supported at each grid position, and accumulate the value of the supported targets to obtain the value of this position;

[0047] (4) Based on the value map, combined with distance attenuation, obtain a relative value map based on the local UAV position. The distance attenuation uses a Gaussian kernel function, and the attenuation degree first increases and then decreases with distance. The specific method is as follows:

[0048] Assume that the maximum action distance of the UAV is maxL, and the distance of the UAV relative to each grid position is d. Then the local support value of this grid position is:

[0049]

[0050] Wherein:

[0051]

[0052] V map is the initial map value, and V UAV is the value of the local support position to be obtained.

[0053] The task location list generation step mainly filters N target locations with larger value according to the size of the value, arranges them from large to small, and constructs a list of target locations to be supported. Here, N is the number of drones in the scene. Each drone goes to the location with the highest value in the list of target locations to be supported and adjusts in time according to the update of the target location list;

[0054] The real-time update and adjustment step mainly receives the list of target locations to be supported from the decisions of neighboring drones in real time during the process of the drone going to support, and adjusts the task target location list by combining the local support target location information and neighbor information. The specific method is as follows:

[0055] (1) Compare the local support target location list and the neighbor drone location list, and judge whether there is a conflict between the location to be supported that the local machine is going to and the locations to be supported that all neighbors are going to. If not, continue to go to this location;

[0056] (2) If there is a conflict, determine the priority of the tasks selected by the local machine and neighbor drones through the drone's own number. If the local machine has the highest priority, continue to go to the location to be supported;

[0057] (3) If there is another drone with a higher priority than the local machine, remove the first location in the local support list, shift the location to be supported to the next one, and go there;

[0058] (4) During the process of the drone going to the target location to be supported, continuously perform operations (1), (2), and (3) to continuously update and adjust the target location where the drone goes, ensuring that the locations where the drones go do not conflict.

[0059] The following is a more specific example:

[0060] Refer to Figure 1 , a distributed decision-making method for drones based on a dynamic task location list, specifically including the following steps:

[0061] Step 1, the ground station initializes the drone positions based on the entire operation area, assigns initial positions to be traveled to each drone, so that the drone cluster covers the target points in the environment to a large extent and maintains overall communication connectivity;

[0062] Step 2: The UAV obtains the position information of ground targets in real time through its sensing function and publishes the sensed information to neighboring UAVs within the communication range. As shown in Figure 2 , neighboring UAVs are the communicable UAVs within one-hop range. For example, the one-hop neighbors of UAV-1 are UAV-2 and UAV-3;

[0063] Step 3: The UAV receives in real time the sensed ground target information shared by neighboring UAVs and integrates it with the target information obtained by itself to obtain all the target information within its local range;

[0064] Step 4: The operation area is rasterized, and according to the UAV position, the sensed ground target information, and the ground target sensing information from neighboring UAVs, the value of each position in the raster map relative to the UAV itself is determined. As shown in Figure 3, the value map calculated through the shared target information of the UAV itself and neighboring UAVs and the value map relative to the UAV itself are shown. The specific operation method is as follows:

[0065] 4a) Integrate the sensed target position information of the UAV itself and the sensed target positions of neighboring UAVs to obtain all the sensed target position information obtained by the UAV itself;

[0066] 4b) Rasterize the entire scene according to the UAV support radius size, as shown in Figure 3(a), and count the number of targets that can be supported at each raster position, and accumulate the value of the supported targets to obtain the value of this position;

[0067] 4d) According to the value map, combined with distance attenuation, obtain the value map of the UAV support position relative to the UAV itself, as shown in Figure 3(b). Assume that the maximum action distance of the UAV is 1 km and the UAV support radius is 50 m. After distance attenuation according to the value map 3(a), the effect shown in Figure 3(b) is obtained. In this way, the UAV can go to a position closer to itself and with a higher value as much as possible, and to a certain extent, it can also avoid nearby UAVs making decisions to the position with the highest value in this local environment at the same time.

[0068] Step 5: According to the value size, screen N target locations with larger values, arrange them from large to small, and construct a list of target locations to be supported, where N is at least the number of UAVs in the scene;

[0069] Step 6: Each UAV goes to the position with the highest value in the support position list and adjusts in real time according to the update of the support position list;

[0070] Step 7: During the process of the UAV going to support, it receives in real time the list of support target positions decided by neighboring UAVs, combines the local support target position information and neighboring information, and adjusts the list of support target positions. Figure 4The following shows the conflict resolution mechanism when multiple UAVs head to the same location according to the task location list. The specific operation method is as follows:

[0071] 7a) Compare the local support target location list with the support location lists of neighboring UAVs to determine whether there is a conflict between the location to be supported by the local UAV and the locations that all neighbors are going to. If not, continue to head to that location;

[0072] 7b) If there is a conflict, determine the priority of the tasks selected by the local UAV and neighboring UAVs through the UAV's own number. The earlier the number, the higher the priority. If the local UAV has the highest priority, continue to head to the location to be supported, such as Figure 4 the action plan of UAV-1 in

[0073] 7c) If there are other UAVs with higher priority than the local UAV, remove the first location in the local support list, shift the location to be supported to the next one, and then head there. Such as Figure 4 the action plan of UAV-2 in

[0074] discard the location with the largest value in the first position of the support location list and head to the next support location;

[0075] Figure 5 The following shows the possible task conflict situations of two UAVs in a non-communication connection state under the condition of local information sharing. When the UAV swarm is in this state, since the distance between UAV-1 and UAV-2 is far beyond the communication range, in their respective local communication networks, relying on the known local information, they both decide to go to the same location (250, 200) with more ground target points near UAV-3 and UAV-4, resulting in a task conflict. When the UAVs head to the same location, the UAVs are always approaching each other and will surely enter each other's communication range. When they establish communication with each other, the task list can be adjusted according to the above strategy. Among them, UAV-2 can discard the optimal location (250, 200) and head to the sub-optimal location (400, 400).

[0076] In summary, in view of the problem that it is difficult to achieve information sharing among all UAVs when multiple UAVs perform tasks in a large range and it is impossible to make an overall decision based on global information, the present invention proposes a distributed decision-making method based on the local information of UAV one-hop neighbors, which can ensure that each UAV can independently execute tasks based on local information and avoid task conflicts with other UAVs, and can be applied to the environment where multiple UAVs cooperate to autonomously sense and execute support tasks.

Claims

1. A distributed decision-making method for drones based on a dynamic task list, characterized in that, it includes the following steps: Step 1, initialize the positions of the drones so that they cover the target points in the environment and maintain overall communication connectivity; Step 2, the drones obtain the position information of the ground targets in real time through the sensing function and publish the sensing information to the neighboring drones within the one-hop communication range; Step 3, the drones receive the sensing information shared by the neighboring drones within the one-hop communication range in real time; Step 4, rasterize the ground area, and determine the value of each position on the raster map according to the drone positions, sensing information, and sensing information from neighboring drones. The specific method is as follows: 4a) Integrate the sensing information of the local drone and the sensing information from neighboring drones to obtain all the sensing information obtained by the local drone; 4b) Rasterize the entire scene according to the support radius of the drones; 4c) Count the number of targets that can be supported at each raster position, and accumulate the value of the supported targets to obtain the value of this position; 4d) According to the value map, combined with distance attenuation, obtain the local support position value map; The specific method is as follows: Assume that the maximum support distance of the drone is maxL, and the distance of the drone relative to each raster position is d. Then the local support value of this raster position is: Where: V map is the initial map value, V UAV is the value of the local support position sought; Step 5, according to the magnitude of the value, screen the N target locations with the largest value, arrange them from largest to smallest, and construct a list of locations to be supported, where N is the number of drones in the scene; Step 6, each drone goes to the position with the highest value in the list of locations to be supported, and makes real-time adjustments according to the update of the list of locations to be supported; Step 7, during the process of the drones going to support, receive the list of locations to be supported from neighboring drones in real time, and combine the local and neighboring lists of locations to be supported to adjust the local current list of locations to be supported.

2. The distributed decision-making method for drones based on a dynamic task list according to claim 1, characterized in that, the specific method of Step 7 is: 7a) Compare the local list of locations to be supported and the list of locations to be supported by neighboring drones, and judge whether the location to be supported that the local drone is about to go to conflicts with the locations to be supported that all neighbors are about to go to. If not, continue to go to this location to be supported; 7b) If there is a conflict, determine the priority of the tasks selected by the local drone and neighboring drones through the drone numbers. If the local priority is the highest, continue to go to this location to be supported and let other drones adjust based on the local list of locations to be supported; 7c) If there are other drones with higher priority than the local drone, remove the first location to be supported in the local list of locations to be supported, shift the local list of locations to be supported to the next one, and continue to go; 7d) During the process of the drone going to the location to be supported, continuously perform the judgment operations in Steps 7a), 7b), and 7c), continuously update and adjust the support location of the drone to ensure that the locations where the drone goes do not conflict.

Images