Search

CN-121995961-A - Unmanned plane cluster flight control cooperative autonomous obstacle avoidance method and system

CN121995961ACN 121995961 ACN121995961 ACN 121995961ACN-121995961-A

Abstract

The invention discloses a unmanned aerial vehicle cluster flight control collaborative autonomous obstacle avoidance method and system, which belong to the technical field of unmanned aerial vehicle cluster flight control and comprise the steps that each unmanned aerial vehicle in a cluster collects environment and cluster state information based on a dynamic perception domain allocation rule, data real-time sharing is realized through distributed communication, multi-source fusion is carried out on the shared data by adopting a differential processing strategy, redundancy and noise are eliminated, a unified environment-cluster fusion state model is generated, a high risk collision area is marked, the obstacle avoidance strategy is converted into a hierarchical flight control instruction and executed, and instruction deviation is adjusted through feedback verification. The sensing range/sampling frequency is dynamically adjusted through the adjacent machine distance and the communication quality, a non-overlapping and fully covered sensing network is formed, the problem of coexistence of sensing redundancy and blind areas is solved, an emergency degree-completion time limit-importance three-dimensional priority hierarchical decision mechanism is established, and balance of obstacle avoidance safety and task efficiency is achieved by combining dynamic weight adjustment.

Inventors

  • ZHAO GANG
  • ZHANG HANG
  • ZHONG JINGYANG
  • KANG ZHAOZHAO

Assignees

  • 西安保通防务科技有限公司

Dates

Publication Date
20260508
Application Date
20260403

Claims (10)

  1. 1. The unmanned plane cluster flight control cooperative autonomous obstacle avoidance method is characterized by comprising the following steps of: s1, each unmanned aerial vehicle in a cluster collects environment and cluster state information based on a dynamic perception domain allocation rule, and data real-time sharing is achieved through distributed communication; s2, carrying out multi-source fusion on the shared data by adopting a differential processing strategy, eliminating redundancy and noise, and generating a unified environment-cluster fusion state model; s3, identifying the type of the obstacle based on the fusion state model, prejudging the dynamic obstacle track by combining the historical motion data and the environmental constraint, and marking a high-risk collision area; S4, generating an obstacle avoidance strategy by adopting a layered decision mechanism according to the task priority of each unmanned aerial vehicle, wherein the unmanned aerial vehicle with high priority occupies a safety channel preferentially, the unmanned aerial vehicles with low priority cooperatively avoid, and meanwhile, the constraint conditions that the topology maintenance coefficient is more than or equal to 0.8 and the formation deviation is less than or equal to 5% are introduced to keep the topology stability of the cluster; s5, converting the obstacle avoidance strategy into a hierarchical flight control instruction and executing the hierarchical flight control instruction, and adjusting instruction deviation through feedback verification; S6, monitoring the cluster state in real time, triggering the unmanned aerial vehicle to exit, the perception domain complement and the obstacle avoidance strategy reconstruction when the single machine fault or the communication interruption is detected, forming dynamic reconstruction, and ensuring the continuity of the obstacle avoidance capability of the cluster.
  2. 2. The unmanned aerial vehicle cluster flight control collaborative autonomous obstacle avoidance method according to claim 1, wherein the dynamic perception domain allocation rule dynamically adjusts the perception range and sampling frequency of each unmanned aerial vehicle according to the adjacent machine distance and communication quality, wherein the preset safety distance is 5m of a medium unmanned aerial vehicle, 3m of a small unmanned aerial vehicle, the communication quality preset threshold is the data transmission success rate of 95%, when the adjacent machine distance is smaller than the preset safety distance, the perception range is reduced according to R=R0× (d/d 0), when the communication quality is lower than the preset threshold, the sampling frequency is increased according to F=10+0.1× (100-R s ), the perception range is limited to 0.5-50m, the sampling frequency is limited to 5-15Hz, and a non-overlapping and fully-covered perception network is formed, wherein R is the perception radius, R0 is the basic radius, d is the adjacent machine distance, d0 is the safety distance, F is the sampling frequency, and R s is the communication success rate.
  3. 3. The unmanned aerial vehicle cluster flight control collaborative autonomous obstacle avoidance method according to claim 2, wherein the multi-source fusion differentiated handling policy comprises: Performing position calibration on static obstacle data by adopting an ICP iterative closest point cloud registration technology, wherein the position error is less than or equal to 0.1m; extracting motion characteristics of dynamic obstacle data by adopting a time sequence filtering algorithm, wherein the motion parameter error is less than or equal to 5%; And adopting consistency check to remove abnormal values on the cluster state data.
  4. 4. The unmanned aerial vehicle cluster flight control collaborative autonomous obstacle avoidance method according to claim 3, wherein the dynamic obstacle track prejudging combines historical motion data with real-time environment constraint parameters, the environment constraint parameters comprise wind speeds of 0-15m/s and terrain gradients of 0-30 degrees, the three-dimensional track coordinates of the dynamic obstacle prejudged by the LSTM network are corrected through posterior correction factors, the corrected track coordinates are P ́ =Px (1+K1+K2), P is the original track coordinates prejudged by the LSTM, the wind speed correction factors are K1=0.05×wind speed values, the terrain gradient correction factors are K2=0.03×gradient values, and the prejudging error is not more than 0.3m, so that 2-5 seconds of early warning is realized.
  5. 5. The unmanned aerial vehicle cluster flight control collaborative autonomous obstacle avoidance method according to claim 4, wherein the task priority of the hierarchical decision mechanism is determined based on three-dimensional dimension weighting of the degree of emergency E, the completion time limit T and the importance I, the priorities P=alpha×E+beta×T+gamma×I, alpha+beta+gamma=1, alpha, beta and gamma are weight coefficients of E, T, I respectively, the environmental risk level is divided according to the distance between the obstacle and the unmanned aerial vehicle, the high risk is within 50% of the safety distance, the medium risk is 50% -80%, the low risk is above 80%, the alpha is increased by 30% in the high risk area, the beta/gamma is reduced in proportion, balance between obstacle avoidance safety and task efficiency is realized, the path offset of the high priority task is less than or equal to 10%, and the value of E, T, I is 1-5 minutes; wherein, emergency degree E: The method comprises the steps of 1 dividing normal operation, 2 dividing daily inspection, 3 dividing emergency inspection, 4 dividing emergency rescue and 5 dividing extreme disaster rescue; Completion time limit T: 1 is divided into more than 8 hours, 2 is divided into 6-8 hours, 3 is divided into 4-6 hours, 4 is divided into 2-4 hours, and 5 is divided into less than or equal to 2 hours; Importance I: the method comprises the steps of dividing 1 into redundant tasks, dividing 2 into auxiliary tasks, dividing 3 into general core tasks, dividing 4 into important core tasks and dividing 5 into key core tasks.
  6. 6. The unmanned aerial vehicle cluster flight control collaborative autonomous obstacle avoidance method according to claim 5, wherein the feedback verification mechanism of the flight control instruction calculates position deviation, speed and attitude deviation values by comparing actual execution data with expected results, a preset deviation threshold is that the position deviation is >0.5m, the speed deviation is >0.3m/s and the attitude deviation is >3 degrees, when the deviation exceeds the preset threshold, the instruction fine adjustment is triggered to a safety area direction, the safety area direction is a three-dimensional shortest vector direction far away from a high risk collision area and accords with the constraint direction of cluster topology maintenance coefficient not less than 0.8, the position deviation fine adjustment step length is 10% of an original instruction, the attitude deviation fine adjustment step length is 20% of the original instruction, the speed deviation fine adjustment step length is 15% of the original instruction, the execution precision is improved, and the final execution error is not more than 0.2m.
  7. 7. The unmanned aerial vehicle cluster flight control collaborative autonomous obstacle avoidance method according to claim 6, wherein the dynamic reconstruction rapidly achieves a reconstruction protocol through a distributed consensus mechanism based on Bayesian fault tolerance, the Bayesian fault tolerance mechanism adaptation rule is that the number N of cluster nodes is not less than 3f+1, N is that of cluster nodes, f is that of fault nodes, the judging period of the number f of fault nodes is that data abnormality is detected continuously 3 times, namely, the fault nodes are judged, the consensus achievement condition is that more than 2/3 nodes vote, the fault unmanned aerial vehicle breaks away from formation to the outside of the cluster according to a preset safety path, the residual unmanned aerial vehicle complements a perception blind area according to a dynamic perception domain allocation rule and re-optimizes an obstacle avoidance decision, the reconstruction response time is not more than 200ms, and the cluster obstacle avoidance success rate is not less than 95% under the fault state.
  8. 8. An unmanned aerial vehicle cluster flight control cooperative autonomous obstacle avoidance system applying the unmanned aerial vehicle cluster flight control cooperative autonomous obstacle avoidance method of any one of claims 1-7, comprising: The cooperative sensing module is used for dynamically distributing a sensing domain based on the distance between the adjacent machines and the communication quality, and collecting and sharing environment and cluster state data; the information fusion module is used for processing the perceived data by adopting a differential processing strategy to generate an environment-cluster fusion state model; The obstacle recognition and prejudgment module is used for recognizing the type of the obstacle, prejudging the dynamic obstacle track by combining the environmental constraint and marking a high-risk area; The layering decision module is used for generating an obstacle avoidance strategy based on the three-dimensional priority of the task and keeping the cluster topology stable; the flight control execution module is used for converting the obstacle avoidance strategy into a hierarchical flight control instruction and executing the hierarchical flight control instruction, and adjusting deviation through feedback verification; The dynamic reconstruction module is used for detecting single machine faults, triggering fault exit, perception domain completion and strategy reconstruction; each module forms closed loop cooperation through a distributed communication network to realize autonomous obstacle avoidance of the cluster.
  9. 9. The unmanned aerial vehicle cluster flight control collaborative autonomous obstacle avoidance system according to claim 8, wherein the collaborative awareness module integrates a laser radar, a visual camera and a millimeter wave radar, the laser radar is responsible for ranging from 0.5 m to 50m, the millimeter wave radar is responsible for short-distance detection from 0m to 30m, the visual camera is responsible for obstacle visual characteristic acquisition, the awareness range is supported to be 0.5 m to 50m and adjustable, the sampling frequency is 5Hz to 15Hz, data sharing is achieved through an ROS2/DDS distributed communication protocol, and hardware of the collaborative awareness module is an STM32H743IIK6 MCU processor, YDLIDAR G laser radar, HB100 millimeter wave radar and an IMX219 visual camera.
  10. 10. The unmanned aerial vehicle cluster flight control collaborative autonomous obstacle avoidance system according to claim 9, wherein the obstacle recognition and prediction module recognizes the obstacle type by adopting YOLOv feature matching algorithm, YOLOv feature library extracts geometric, texture and motion features, the labeled sample size is more than or equal to 10000, a dynamic obstacle track prediction model is built by combining 3 layers of LSTM network with wind speed and terrain gradient parameters, the 3 layers of LSTM network is an input layer 128 neuron, a hidden layer 64 neuron and an output layer 3 neuron, the activation function is Tanh, and the prediction error is not more than 0.3m.

Description

Unmanned plane cluster flight control cooperative autonomous obstacle avoidance method and system Technical Field The invention relates to the technical field of unmanned aerial vehicle cluster flight control, in particular to an unmanned aerial vehicle cluster flight control cooperative autonomous obstacle avoidance method and system. Background Along with the wide application of unmanned aerial vehicle cluster technology in scenes such as logistics distribution, disaster relief, industrial inspection, etc., the cooperation keeps away the barrier and becomes the core technology of guaranteeing cluster safety operation. In the prior art, the distributed cooperative control architecture realizes multi-machine decision cooperation through a blockchain ledger, a DMPC framework and the like, but has the problem of obstacle avoidance response lag caused by information interaction delay, the multi-mode perception fusion technology integrates laser radar and vision sensor data, but does not form a dynamic perception resource allocation mechanism, so that local area perception redundancy or blind areas are caused, and a differential fusion strategy is not designed for static/dynamic obstacles, the path planning methods such as a manual potential field method, an improved ant colony algorithm and the like can realize static obstacle avoidance, but have insufficient track prejudging capability for the dynamic obstacles, do not combine with real-time environment constraints (such as wind speed and terrain gradient) to carry out track correction, and the obstacle avoidance strategy and task priority lack of dynamic adaptation, and meanwhile, the existing system has the defect of a quick reconstruction mechanism when a single machine fails, the cluster obstacle avoidance robustness is insufficient, and feedback verification logic is simple in the flight control execution process, and instruction deviation under a complex environment is difficult to deal with. In view of the above, the present invention is specifically proposed to solve the above-mentioned technical problems. Disclosure of Invention The invention aims to provide a flight control cooperative autonomous obstacle avoidance method and a flight control cooperative autonomous obstacle avoidance system for an unmanned aerial vehicle cluster, which are used for solving the technical problems that in the prior art, collision risks easily occur and task execution efficiency is reduced in a complex dynamic environment of the unmanned aerial vehicle cluster. In order to achieve the above purpose, the present invention provides the following technical solutions: the first object of the invention is to provide a unmanned aerial vehicle cluster flight control cooperative autonomous obstacle avoidance method, which comprises the following steps: s1, each unmanned aerial vehicle in a cluster collects environment and cluster state information based on a dynamic perception domain allocation rule, and data real-time sharing is achieved through distributed communication; s2, carrying out multi-source fusion on the shared data by adopting a differential processing strategy, eliminating redundancy and noise, and generating a unified environment-cluster fusion state model; s3, identifying the type of the obstacle based on the fusion state model, prejudging the dynamic obstacle track by combining the historical motion data and the environmental constraint, and marking a high-risk collision area; S4, generating an obstacle avoidance strategy by adopting a layered decision mechanism according to the task priority of each unmanned aerial vehicle, wherein the unmanned aerial vehicle with high priority occupies a safety channel preferentially, the unmanned aerial vehicles with low priority cooperatively avoid, and meanwhile, the constraint conditions that the topology maintenance coefficient is more than or equal to 0.8 and the formation deviation is less than or equal to 5% are introduced to keep the topology stability of the cluster; s5, converting the obstacle avoidance strategy into a hierarchical flight control instruction and executing the hierarchical flight control instruction, and adjusting instruction deviation through feedback verification; S6, monitoring the cluster state in real time, triggering the unmanned aerial vehicle to exit, the perception domain complement and the obstacle avoidance strategy reconstruction when the single machine fault or the communication interruption is detected, forming dynamic reconstruction, and ensuring the continuity of the obstacle avoidance capability of the cluster. Further, a dynamic perception domain allocation rule dynamically adjusts the perception range and sampling frequency of each unmanned aerial vehicle according to the adjacent machine distance and the communication quality, wherein the preset safety distance is 5m of the medium unmanned aerial vehicle and 3m of the small unmanned aerial vehicle, the preset communication quali