Search

CN-122022211-A - Unmanned aerial vehicle cluster reversible task reconstruction method and system based on risk prediction

CN122022211ACN 122022211 ACN122022211 ACN 122022211ACN-122022211-A

Abstract

The invention discloses a reversible task reconstruction method and system for an unmanned aerial vehicle cluster based on risk prediction, and relates to the field of unmanned aerial vehicle cluster cooperative control and intelligent task scheduling. The method comprises the steps of constructing a task graph based on job task constraint, performing causal consistency constraint cross-modal coding on environment perception data and task semantic data to obtain stable risk representation, predicting dynamic risk distribution by combining unmanned plane state data and cluster link data to generate a risk driving reconstruction boundary, and completing task division, compensation node inheritance, task dependent edge reconstruction updating and subtask chain inheritance allocation. The method can improve task continuity, reconstruction accuracy, result inheritance effectiveness and cooperative robustness in a complex dynamic environment.

Inventors

  • XU RONGSHENG

Assignees

  • 苏州智蓝德科技有限公司

Dates

Publication Date
20260512
Application Date
20260416

Claims (10)

  1. 1. A method for reconstructing reversible tasks of unmanned aerial vehicle clusters based on risk prediction is characterized by comprising the steps of constructing a task graph based on preface constraint, space transfer constraint and resource coupling constraint of a job task , Performing causal consistency constraint cross-modal coding on the environment perception data and the task semantic data to obtain the task graph Stable risk representation corresponding to medium task node ; Based on the stability risk characterization The unmanned plane state data and the cluster link data predict comprehensive risk values of all positions in a future time domain of an operation area at all times to form dynamic risk distribution, determine a key risk factor propagation direction vector, and generate a risk driving reconstruction boundary according to a risk propagation enhancement trend, a risk gradient mutation position and a task dependent edge crossing a high risk area ; Driving a reconstruction boundary according to the risk Mapping the task graph The method comprises the steps of dividing the method into a locking node area, a movable node area and a compensation access area, wherein the completed result of task nodes in the locking node area is kept to be inherited effectively, the task nodes in the movable node area allow decoupling recombination, and the compensation access area is used for inserting compensation nodes; Generating compensation nodes for unfinished task segments corresponding to faults, disconnection or risk out-of-bounds The compensation node Inheriting task targets, residual time limit constraints, resource demand constraints, execution context states and result boundary information of the original task segment; Cutting off crossing the risk driven reconstruction boundary And the task dependency edges of the high-risk nodes and the low-risk nodes are connected, the task dependency edges in the locking node areas are reserved, and the compensation nodes are connected Establishing an update task dependency edge with an adjacent locking node or a movable node to generate an update subtask chain; The updated subtask chain is distributed to a target unmanned aerial vehicle for execution based on unmanned aerial vehicle residual energy, payload capacity, space reachability, link reachability and task inheritance mismatch cost, wherein the task inheritance mismatch cost characterizes the target unmanned aerial vehicle to the compensation node The degree of mismatch of the load connection, the track connection, the data context connection and the link continuous connection.
  2. 2. The unmanned aerial vehicle cluster reversible task reconstruction method based on risk prediction of claim 1, wherein the causal consistency constraint cross-modal coding comprises performing cross-modal alignment on multi-disturbance samples under the same task target to obtain a joint feature vector For the joint feature vector Performing causal decomposition to obtain stable risk feature vector And an unstable disturbance eigenvector Only the stable risk feature vector Mapping to the stable risk characterization , In which, in the process, In order to combine the feature vectors, In order to stabilize the risk feature vector, As the feature vector of the unstable disturbance, To stabilize the risk mapping function.
  3. 3. The unmanned aerial vehicle cluster reversible task reconstruction method based on risk prediction according to claim 1, wherein the position in the work area is At the moment of time Is a comprehensive risk value of (1) Determined according to the following formula: In which, in the process, In order to be a value of the risk of flight, In order to communicate the risk value of the communication, As a value of the energy risk, As a value of the risk of failure of the task, To be characterized by the stability risk The resulting semantic risk value is mapped to, 、 、 、 And And the weight coefficients of the corresponding risk items are respectively.
  4. 4. The unmanned aerial vehicle cluster reversible task reconstruction method based on risk prediction according to claim 3, wherein risk propagation enhances trend Determined according to the following formula and based on the risk propagation enhancement trend Generating the risk driven reconstruction boundary : In which, in the process, Is the position At the moment of time Is a risk-spread-enhancing trend value of (1), To integrate the time rate of change of the risk value, For the spatial gradient of the integrated risk value, For the modular length of the spatial gradient of the integrated risk value, Is the gradient adjustment coefficient when Exceeding a preset threshold When the task dependency edge crosses the high-risk area, determining the corresponding position of the task dependency edge as the risk-driven reconstruction boundary Is a candidate generation location of (1).
  5. 5. The unmanned aerial vehicle cluster reversible task reconstruction method based on risk prediction according to claim 1, wherein the updating task dependency edge set Determined according to the following formula: In which, in the process, In order to update the set of task dependent edges, To lock the set of task dependent edges maintained within the node region, For a reconstructed set of task dependent edges within the migratable node area, A task dependent edge set newly added after the node is accessed for compensating, wherein the task dependent edge set is used for compensating the node access The update task dependent edge in the (B) is connected with the compensation node With adjacent lock nodes or with a migratable node.
  6. 6. The unmanned aerial vehicle cluster reversible task reconstruction method based on risk prediction according to claim 1, wherein the compensation node The inherited execution context state comprises at least two of a stage data state, a perception buffer state, a communication session state and a control mode state of an original task segment, and the result boundary information comprises at least two of a completed task range, an incomplete task starting point and continuous execution interface information.
  7. 7. The unmanned aerial vehicle cluster reversible task reconstruction method based on risk prediction according to claim 1, wherein the task inherits mismatch costs Determined according to the following formula: , In the formula, The mismatch cost is inherited for the task, In order for the load to match the cost, In order to make a trace-continuation cost, For the data context to be missing a cost, For the cost of the link continuity loss, 、 、 And And (5) the weight coefficient of the corresponding cost item.
  8. 8. The unmanned aerial vehicle cluster reversible task reconstruction method based on risk prediction according to claim 7, wherein the unmanned aerial vehicle cluster reversible task reconstruction method based on risk prediction is characterized in that the unmanned aerial vehicle cluster reversible task reconstruction method based on risk prediction is the following The frame candidate unmanned aerial vehicle inherits the mismatch cost to the task of its compensation node C The candidate screening condition and the comprehensive allocation cost component item used for updating the subtask chain allocation only meet And is also provided with Is included in the assignable object set, and the first unmanned aerial vehicle is included in the assignable object set The frame candidate unmanned aerial vehicle determines comprehensive allocation cost according to the following formula : In which, in the process, Is the first The comprehensive allocation cost corresponding to the frame candidate unmanned aerial vehicle, In the light of the cost of time, In order for the cost of energy to be high, In the light of the risk penalty, In order to be at the cost of the link, Is the first The task corresponding to the rack candidate unmanned aerial vehicle inherits the mismatch cost, 、 、 、 And In order to correspond to the weight coefficients of the cost term, In order to mismatch the cost threshold value, Is the first Sensitivity of the rack candidate drone to key risk factors, Is the sensitivity threshold.
  9. 9. The unmanned aerial vehicle cluster reversible task reconstruction method based on risk prediction according to claim 4, wherein when When the risk driven reconstruction boundary is preferentially generated along the outside of the propagation direction of the key risk factors And preferentially distinguishing from the risk driven reconstruction boundaries Intersecting or reconstructing boundaries with the risk driven Is not greater than a preset distance threshold To suppress risk propagation to the lock node region.
  10. 10. Unmanned aerial vehicle cluster reversible task reconstruction system based on risk prediction, characterized by comprising: Task graph construction and stability risk characterization module for constructing task graph based on preface constraint, space transfer constraint and resource coupling constraint of job task Performing cross-modal coding of causal consistency constraint on environment perception data and task semantic data to obtain the task graph Stable risk representation corresponding to medium task node ; The boundary generation module is used for predicting comprehensive risk values of all positions in the future time domain of the operation area at all moments based on the stable risk representation, the unmanned plane state data and the cluster link data According to the comprehensive risk value Risk propagation enhancement trend, risk gradient abrupt change location, and task dependent edge generation risk driven reconstruction boundary across high risk areas ; A partitioning module for driving the reconstruction boundary according to the risk Mapping the task graph Dividing the system into a locking node area, a movable node area and a compensation access area; the compensation inheritance module is used for generating compensation nodes from unfinished task segments corresponding to faults, disunion or risk out-of-limits And causing the compensation node to Inheriting task targets, residual time limit constraints, resource demand constraints, execution context states and result boundary information of the original task segment; A task chain reconstruction module for cutting off the crossing of the risk driving reconstruction boundary And the task dependency edges of the high-risk nodes and the low-risk nodes are connected, the task dependency edges in the locking node areas are reserved, and the compensation nodes are connected Establishing an update task dependency edge with an adjacent locking node or a movable node to generate an update subtask chain; The inheritance allocation module is used for allocating the updated subtask chain to a target unmanned aerial vehicle for execution based on unmanned aerial vehicle residual energy, effective load capacity, space accessibility, link accessibility and task inheritance mismatch cost, wherein the task inheritance mismatch cost characterizes the target unmanned aerial vehicle to the compensation node The degree of mismatch of the load connection, the track connection, the data context connection and the link continuous connection.

Description

Unmanned aerial vehicle cluster reversible task reconstruction method and system based on risk prediction Technical Field The invention relates to the technical field of unmanned aerial vehicle cluster cooperative control and intelligent task scheduling, in particular to an unmanned aerial vehicle cluster reversible task reconstruction method and system based on risk prediction. Background With the development of unmanned aerial vehicle platforms, airborne sensing, trunking communication and intelligent decision-making technologies, unmanned aerial vehicle trunking is widely applied to scenes such as emergency reconnaissance, inspection and monitoring, target searching, material delivery, disaster assessment and complex regional operation. In the above scenario, multiple unmanned aerial vehicles are usually required to cooperatively execute a composite task with a preamble constraint, a space transfer constraint and a resource coupling constraint around the same job target, so how to maintain continuous execution of the task under the conditions of dynamic environmental changes, link fluctuation, single machine failure, disconnection and rapid local risk diffusion has become a key technical problem in unmanned aerial vehicle cluster application. In the prior art, unmanned aerial vehicle cluster task adjustment schemes are mostly focused on static task allocation, local path re-planning, task transfer after failure or risk assessment based on perception information. The scheme can generally correct an execution main body or an execution path after local abnormality occurs, but most of the scheme still stays on the technical level of 'risk assessment + rescheduling' or 'fault detection + reassignment', lacks a mechanism for carrying out bordering, partitioning and inheriting reconstruction around a task graph structure, and is difficult to consider the complete task result protection, incomplete task continuous connection and the overall structure stability of the task graph in a complex dynamic scene. Further, in complex work areas, risk sources often feature multiple sources, heterogeneous, and coupled, including both flight environmental risks, as well as communication risks, energy risks, and mission failure risks. In the prior art, the utilization of environment perception data and task semantic data is mainly based on simple fusion or correlation modeling, and the interference of pseudo-correlation semantics and transient disturbance on risk judgment is difficult to effectively inhibit, so that the stability of a risk prediction result is insufficient when the scene changes, and the accuracy of a subsequent task reconstruction decision is affected. In addition, when a certain unmanned aerial vehicle cannot continue to execute a current task segment due to failure, disconnection or risk out of range, the prior art generally directly redistributes an incomplete task to other unmanned aerial vehicles for execution. The simple reassignment mode often fails to keep task targets, residual time limit constraints, resource demand constraints, execution context states and result boundary information of the original task segment, and task repetition, task omission, boundary dislocation or data context interruption are easily caused, so that task continuity and overall cooperative efficiency are reduced. Meanwhile, the prior art generally lacks a risk-driven reconstruction boundary generation mechanism based on the risk propagation direction, the risk propagation enhancement trend and the situation that a task dependency edge spans a high-risk area, and cannot determine which task nodes should be locked and protected, which task nodes allow migration and reorganization, and which areas are suitable as access positions of compensation nodes. Therefore, when the risk is rapidly diffused or the link fragile region is dynamically changed, the conventional scheme is difficult to timely inhibit the risk from spreading to the completed task region, and is also difficult to realize reversible task reconstruction facing the task graph structure. Therefore, it is needed to provide a method and a system for reconstructing reversible tasks of unmanned aerial vehicle clusters based on risk prediction, so as to solve the problems of insufficient stability risk representation capability, unclear risk driving reconstruction boundary, weak task connection incomplete capability after failure and insufficient suitability of task inheritance allocation in the prior art. Disclosure of Invention The invention aims to provide a reversible task reconstruction method and a reversible task reconstruction system for an unmanned aerial vehicle cluster based on risk prediction, which are used for solving the problems that in the prior art, the stability of risk representation of the unmanned aerial vehicle cluster in a dynamic complex environment is insufficient, a task graph reconstruction boundary is undefined, unfinished tasks after failur