Multi-unmanned aerial vehicle three-dimensional group intelligent dynamic path planning method fused with multi-source environment perception

Abstract

The invention relates to the technical field of unmanned aerial vehicle path planning, and discloses a multi-unmanned aerial vehicle three-dimensional group intelligent dynamic path planning method fusing multi-source environment perception, which comprises an intelligent dynamic path planning system and comprises the following operation steps: S1, collecting and preprocessing multi-source environment information; s2, performing multi-source information fusion and three-dimensional environment modeling; s3, initializing an improved swarm intelligence algorithm and optimizing a path; s4, multi-machine collaboration and dynamic path adjustment; and S5, smoothly outputting the path. According to the multi-unmanned aerial vehicle three-dimensional group intelligent dynamic path planning method fused with multi-source environment perception, through multi-source data fusion and accurate three-dimensional environment modeling, environment information such as the terrain of an operation area can be comprehensively mastered in real time, so that an unmanned aerial vehicle group can still plan a feasible path in a complex and changeable scene, and the working efficiency is improved. By improving a swarm intelligence algorithm and a multi-aircraft cooperation mechanism, multi-aircraft conflicts are effectively avoided, and the operation efficiency and cooperation safety of the unmanned aerial vehicle group are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle path planning, in particular to a multi-unmanned aerial vehicle three-dimensional swarm intelligent dynamic path planning method fusing multi-source environment perception. BACKGROUND

[0002] With the rapid development of unmanned aerial vehicle technology, multi-unmanned aerial vehicle cooperative operation is increasingly widely used in fields such as logistics distribution, emergency rescue, power inspection, aerial photography and surveying and mapping. Path planning, as the core technology of multi-unmanned aerial vehicle cooperative operation, directly determines the operation efficiency, flight safety and task completion quality.

[0003] However, in the traditional technology, multi-unmanned aerial vehicles have low path planning accuracy, weak dynamic adaptability and frequent multi-vehicle conflicts due to insufficient environment perception and poor multi-vehicle cooperation in complex three-dimensional environments, which makes it difficult to meet the high-precision, high-real-time and high-safety requirements of multi-unmanned aerial vehicle swarm intelligent dynamic path planning in complex three-dimensional environments. At the same time, the traditional swarm intelligent algorithm has slow convergence speed and insufficient multi-objective optimization capability in three-dimensional path planning, and the path conflicts are frequent, the operation efficiency is low, and the cooperative safety is poor due to the lag of information interaction and imperfect decision mechanism when multi-unmanned aerial vehicles cooperate. SUMMARY

[0004] (I) Technical problems solved In view of the deficiencies of the prior art, the present application provides a multi-unmanned aerial vehicle three-dimensional swarm intelligent dynamic path planning method fusing multi-source environment perception, which has the advantages of comprehensive and accurate environment perception, efficient and optimal path planning, orderly multi-vehicle cooperation and flexible dynamic response, and solves the problems of insufficient multi-source environment perception fusion, optimization accuracy and efficiency of swarm intelligent algorithm, and multi-vehicle dynamic cooperation of existing path planning methods.

[0005] (II) Technical solutions In order to achieve the purposes of comprehensive and accurate environment perception, efficient and optimal path planning, orderly multi-vehicle cooperation and flexible dynamic response, the present application provides the following technical solutions: a multi-unmanned aerial vehicle three-dimensional swarm intelligent dynamic path planning method fusing multi-source environment perception, comprising an intelligent dynamic path planning system, and the operation steps are as follows: S1: Multi-source environment information acquisition and preprocessing; a machine-mounted perception system is formed by a visual sensor, a laser radar, a millimeter wave radar, a GPS / Beidou dual-mode positioning module and a meteorological sensor carried by an unmanned aerial vehicle, combined with ground base station microwave communication data and satellite remote sensing information, to synchronously collect terrain elevation, obstacle distribution, weather conditions, electromagnetic interference, airspace restrictions and unmanned aerial vehicle state data; a filtering algorithm is used to denoise, time-space align and remove redundancy of multi-source data, to generate a standardized data set; S2: Multi-source information fusion and three-dimensional environment modeling; a weighted adaptive fusion algorithm is used to realize the accurate fusion of multi-source heterogeneous data by combining Kalman filtering, D-S evidence theory or Bayesian estimation, and dynamically adjusting the weight of each data source; an octree structure-based dynamic three-dimensional environment model is constructed to support real-time updating of obstacle state, terrain parameters and weather conditions, and the model resolution can be adaptively adjusted within the range of 0.1m-1m; S3: Improved swarm intelligence algorithm initialization and path optimization; based on the particle swarm optimization algorithm, chaos initialization strategy, adaptive inertia weight and dynamic learning factor are introduced to optimize the global search and local optimization ability of the algorithm; the starting point, ending point, flight constraint conditions and multi-objective optimization function of multiple unmanned aerial vehicles are set to generate an initial three-dimensional path by iteratively updating the particle position; S4: Multi-machine cooperation and dynamic path adjustment; multi-unmanned aerial vehicle data sharing is achieved through 5G+WiFi6 dual-mode communication or wireless self-organizing network, and real-time monitoring of environmental changes and flight status is achieved based on a distributed collaborative decision-making mechanism; when the detected environmental change exceeds the threshold or the distance between multiple machines is less than the safety threshold, local or global path re-planning is triggered to dynamically adjust the path parameters; S5: Path smoothing output; B-spline curves are used to smooth the optimized path to eliminate sharp corners and ensure stable unmanned aerial vehicle flight attitude, outputting three-dimensional coordinate sequences, speed suggestion values and energy consumption estimation data that can be directly called by the flight control system.

[0006] Preferably, the intelligent dynamic path planning system comprises a multi-source environment information acquisition module, a data preprocessing module, a multi-source information fusion module, a three-dimensional environment modeling module, an improved swarm intelligence algorithm module, a multi-machine cooperation and dynamic adjustment module, a path smoothing output module, and an unmanned aerial vehicle flight control interface module.

[0007] Preferably, the sampling frequency of the multi-source sensor in step S1 is synchronized to 10Hz-20Hz, wherein the visual sensor realizes obstacle recognition and positioning through the YOLOv8 algorithm, the laser radar ranging accuracy is within ±0.3m, the GPS / Beidou positioning module positioning error is ≤±0.5m, the millimeter wave radar has obstacle detection capability in bad weather such as rain, snow and fog, and the pre-processing process adopts a combination algorithm of Kalman filtering and wavelet transform, or a combination scheme of median filtering and Kalman filtering, and the data processing delay after noise suppression is controlled within 50ms.

[0008] Preferably, the weight coefficient updating period of the weighted adaptive fusion algorithm in step S2 is 50 ms, the error of the fused environmental data is controlled within ±0.3 m, the three-dimensional environmental model supports the representation of complex terrain with an elevation difference of 0-500 m, and the data updating frequency is synchronized with the sensor acquisition frequency.

[0009] Preferably, the chaotic initialization strategy of the improved swarm intelligence algorithm in step S2 is implemented based on a Logistic mapping, the adaptive inertia weight value ranges from 0.4 to 0.9 and is dynamically adjusted according to the iteration number or the environmental complexity, the learning factor adaptively varies between 1.5 and 2.5, the population size is set to 50-100, and the maximum iteration number is 50-100.

[0010] Preferably, the flight constraint conditions in step S3 include a maximum flight speed ≤ 30 m / s, a maximum climb / descent rate ≤ 5 m / s, and a minimum turn radius ≥ 5 m; the multi-vehicle safety distance can be dynamically adjusted, which is ≥ 5 m when the speed is ≤ 10 m / s, is set to 0.8 times the speed value when the speed is > 10 m / s, and is not less than 10 m.

[0011] Preferably, the multi-objective optimization function in step S3 is constructed by weighted summation, the weight coefficients satisfy a total sum of 1, and the optimization indexes include path length (weight 0.3), flight energy consumption (weight 0.4), path smoothness, obstacle safety distance, and multi-vehicle conflict coefficient (total weight 0.3), and the smaller the objective function value, the higher the path optimality.

[0012] Preferably, the communication link in step S4 adopts a 5G+WiFi6 dual-mode or a 2.4 GHz wireless self-organizing network, the communication delay is ≤ 10 ms, the multi-unmanned vehicles share position, speed, and path information through a distributed peer-to-peer network architecture, and the conflict detection response time is ≤ 100 ms.

[0013] Preferably, the environmental change threshold in step S4 is set as: the obstacle movement distance exceeds 1 m, the wind speed mutation exceeds 5 m / s, the environmental model difference degree is > 0.2, or a new obstacle appears, the path re-planning response time is not more than 200 ms, and the smooth path curvature change rate is ≤ 0.5 rad / m.

[0014] (Three) Beneficial Effects Compared with the prior art, the present application provides a multi-unmanned vehicle three-dimensional swarm intelligence dynamic path planning method fusing multi-source environmental perception, which has the following beneficial effects: 1. The multi-unmanned vehicle three-dimensional swarm intelligence dynamic path planning method fusing multi-source environmental perception can comprehensively and real-timely master the environmental information such as the terrain, obstacles, and weather in the operation area through multi-source data fusion and accurate three-dimensional environmental modeling, greatly improves the environmental adaptability of path planning, and enables the unmanned vehicle swarm to still plan a feasible path in a complex and variable scene.

[0015] 2. The method of the present application, through the improved swarm intelligence algorithm and multi-vehicle coordination mechanism, realizes multi-objective optimization and dynamic adjustment of the path, effectively avoids multi-vehicle conflict, and significantly improves the operation efficiency and coordination safety of the UAV swarm while ensuring the optimal path length and the lowest energy consumption. BRIEF DESCRIPTION OF DRAWINGS

[0016] Fig. 1 The method flowchart of the present application; Fig. 2 The system operation architecture diagram of the present application; Fig. 3 The operation flowchart of the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0018] Please refer to Figs. 1-3 The method of the present application, which fuses multi-source environment perception, includes an intelligent dynamic path planning system, and the operation steps are as follows: S1: Multi-source environment information acquisition and preprocessing; an airborne perception system is formed by a visual sensor, a laser radar, a millimeter wave radar, a GPS / Beidou dual-mode positioning module, and a meteorological sensor carried by a UAV, combined with microwave communication data of a ground base station and satellite remote sensing information, to synchronously acquire terrain elevation, obstacle distribution, weather conditions, electromagnetic interference, airspace restrictions, and UAV self-state data; a filtering algorithm is used to denoise, time-space align, and remove redundancies of the multi-source data, to generate a standardized data set; Further, in the data preprocessing process, according to the characteristics of different types of sensor data, an adaptive filtering strategy is adopted to focus on optimizing edge detail retention for visual sensor data, to strengthen density uniformity processing for laser radar point cloud data, and to improve time sequence stability for meteorological sensor data. After processing, the data delay is controlled within 50 ms, the spatial coordinate error of multi-source data is less than 0.2 m, and the time synchronization accuracy reaches 10 ms, laying a foundation of high precision and high timeliness for subsequent multi-source information fusion and three-dimensional environment modeling.

[0019] S2: Multi-source information fusion and three-dimensional environment modeling; a weighted adaptive fusion algorithm is used to realize the accurate fusion of multi-source heterogeneous data by combining Kalman filter, D-S evidence theory or Bayesian estimation, and dynamically adjusting the weight of each data source; based on octree structure, a dynamic three-dimensional environment model is constructed to support real-time updating of obstacle state, terrain parameters and weather conditions, and the model resolution can be adaptively adjusted in the range of 0.1m-1m; Further, in the multi-source information fusion stage, a dynamic weight learning mechanism is established according to the data reliability difference in different scenarios: when the visual sensor is disturbed by light, the data weight of laser radar and millimeter wave radar is automatically increased; in the complex weather area, the fusion proportion of meteorological sensor and satellite remote sensing information is strengthened, based on the fused data, the octree three-dimensional environment model can complete a local update within 100ms, the state prediction accuracy of moving obstacles is more than 95% in the standard test environment, and the modeling error of terrain elevation is controlled within 0.15m, which provides high-precision and high-dynamic three-dimensional environment support for subsequent path planning.

[0020] S3: Improved swarm intelligence algorithm initialization and path optimization; based on particle swarm optimization algorithm, chaos initialization strategy, adaptive inertia weight and dynamic learning factor are introduced to optimize the global search and local optimization ability of the algorithm; the starting point, ending point, flight constraint conditions and multi-objective optimization function of multiple UAVs are set to generate the initial three-dimensional path by iteratively updating the particle position; Further, in the algorithm iteration process, a hierarchical particle swarm topology is designed for multi-UAV cooperative scenarios: the UAV group is divided into multiple subgroups, full connection topology is used within the subgroup to ensure local optimization accuracy, and global information sharing is realized between subgroups through the leader particle, which is the particle with the best fitness function value in the subgroup, and its position information is used to guide the search direction of other subgroups, thereby realizing global cooperation. At the same time, a multi-objective Pareto dominance mechanism is introduced to dynamically balance the path length, energy consumption, flight time and safety margin—when the airspace is strictly limited, the safety margin target weight is prioritized; in long-distance logistics operations, the path length and energy consumption are optimized, and the convergence speed of the improved algorithm is improved by 30% compared with the traditional particle swarm algorithm, the optimization success rate in complex three-dimensional environment is more than 98%, the deviation between the generated initial path and the approximate optimal path verified by a large number of iterations is usually less than 5%, which lays a high-efficiency and high-quality path foundation for subsequent multi-UAV cooperative adjustment.

[0021] S4: Multi-UAV cooperation and dynamic path adjustment; multi-UAV data sharing is realized through 5G+WiFi6 dual-mode communication or wireless self-organizing network, and real-time monitoring of environmental changes and flight status is realized based on distributed cooperative decision mechanism; when the detected environmental change exceeds the threshold or the distance between multiple UAVs is less than the safety threshold, local or global path re-planning is triggered to dynamically adjust the path parameters; Further, in the dynamic path adjustment process, a hierarchical response mechanism is constructed: when detecting small-range obstacle movement (such as a single tree falling), only local path re-planning is triggered, and path adjustment is completed within 50 ms through the B-spline curve local correction algorithm; when encountering large-scale weather mutation (such as strong wind area diffusion), the global re-planning process is started, and based on the fast convergence characteristics of the improved swarm intelligence algorithm, a new global optimal path is generated within 200 ms, at the same time, the multi-machine realizes conflict avoidance through real-time state feedback, when the predicted distance between two machines is less than 1.5 times the safety threshold, the collision avoidance algorithm is automatically activated, through the coordinated adjustment of speed and heading, it is ensured that the safe distance between multi-machines is always above the minimum safety threshold.

[0022] S5: Path smoothing output; B-spline curve is used to smooth the optimized path, eliminate sharp corners and ensure the stability of the unmanned aerial vehicle flight attitude, output three-dimensional coordinate sequence, speed suggestion value and energy consumption estimation data which can be directly called by the flight control system; Further, in the B-spline curve smoothing process, curvature constraint and energy consumption optimization factor are introduced: according to the dynamic characteristics and flight mechanics model of the unmanned aerial vehicle, the path curvature is limited within the maximum turning curvature range of the unmanned aerial vehicle, and the energy consumption change rate of the path segment is kept stable, after smoothing, the curvature continuity of the path reaches 100%, the attitude angle change rate in the flight process of the unmanned aerial vehicle is reduced by 40%, and the energy consumption fluctuation amplitude is controlled within 5%, the output flight control data also has multi-mode adaptability, which can automatically convert data format according to different types of unmanned aerial vehicle flight control protocols (such as MavLink, custom protocol), realize seamless connection with various unmanned aerial vehicle flight control systems, in actual operation, the path tracking accuracy is improved to ±0.3 m, which fully guarantees that the multi-unmanned aerial vehicle group can complete the operation task stably, efficiently and lowly in three-dimensional space according to the planned path.

[0023] In the case implementation, the intelligent dynamic path planning system includes a multi-source environment information acquisition module, a data preprocessing module, a multi-source information fusion module, a three-dimensional environment modeling module, an improved swarm intelligence algorithm module, a multi-machine cooperation and dynamic adjustment module, a path smoothing output module and an unmanned aerial vehicle flight control interface module; The system obtains multi-dimensional data such as airborne sensors, ground base stations, satellite remote sensing, etc. through a multi-source environmental information acquisition module; a data preprocessing module filters and denoises the data, aligns the time and space, ensures the validity of the data, and the processing delay is ≤50 ms; a multi-source information fusion module uses a weighted adaptive fusion algorithm to accurately fuse multi-source data; a three-dimensional environmental modeling module constructs a dynamic three-dimensional model based on octree, with a resolution of 0.1 m-1 m adaptive adjustment; an improved swarm intelligence algorithm module introduces chaos initialization and other strategies to realize multi-objective optimization of path length, energy consumption and safety, and the algorithm convergence speed is improved by 30%; a multi-machine cooperation and dynamic adjustment module realizes data sharing through 5G+WiFi6 dual-mode communication, triggers re-planning within 200 ms when the environment changes or multi-machine conflicts occur, and the multi-machine conflict occurrence rate is reduced to below 0.1%; a path smoothing output module processes the path through B-spline curve, and outputs three-dimensional coordinates, speed and energy consumption data suitable for flight control; the unmanned aerial vehicle flight control interface module is compatible with multiple brands of flight control protocols, with a compatibility of more than 95%, thereby realizing precise, efficient and safe three-dimensional swarm intelligence dynamic path planning of multiple unmanned aerial vehicles in complex environments.

[0024] In the case implementation, the sampling frequency of the multi-source sensor in step S1 is 10Hz-20Hz, wherein the visual sensor realizes obstacle recognition and positioning through YOLOv8 algorithm, the laser radar ranging accuracy is within ±0.3m, the GPS / Beidou positioning module positioning error is ≤±0.5m, the millimeter wave radar has obstacle detection capability in bad weather such as rain, snow and fog, and the pre-processing process adopts the algorithm combining Kalman filter and wavelet transform, or the combination scheme of median filter and Kalman filter, and the data processing delay after noise suppression is controlled within 50ms; Further, in the multi-source data time and space alignment link, based on the time stamp of the sensor and the space coordinate system conversion matrix, a multi-source data synchronous calibration model is constructed: through the precise time service function of GPS / Beidou, the time deviation of each sensor data is corrected to within 1ms; the space registration is carried out by using the point cloud features of laser radar and the image features of visual sensor, so that the spatial coordinate deviation of multi-source data is less than 0.1m, and the standardized data set after preprocessing realizes high unification in feature dimension and data accuracy, wherein the accuracy of obstacle recognition is improved to 99%, and the measurement accuracy of terrain elevation is optimized to ±0.2m, which provides high-quality input data for subsequent multi-source information fusion and three-dimensional environmental modeling, and effectively supports the accurate perception ability of the whole path planning system in complex scenes.

[0025] In the case implementation, the weight coefficient update period of the weighted adaptive fusion algorithm in step S2 is 50 ms, the error of the fused environmental data is controlled within ±0.3 m, the three-dimensional environmental model supports the representation of complex terrain with an elevation difference of 0-500 m, and the data update frequency is synchronized with the sensor acquisition frequency; Further, in the dynamic updating process of the three-dimensional environmental modeling, a layered detail level (LOD) strategy is adopted: a fine modeling with a resolution of 0.1 m is adopted for the core area of the unmanned aerial vehicle operation, and the resolution is adaptively reduced to 1 m for the non-key area at a long distance, which reduces the model data volume by 60% while ensuring the accuracy of environmental representation, improves the data processing and transmission efficiency, and based on the three-dimensional environmental model, the prediction accuracy of the motion trend of the dynamic obstacle reaches 97%, and the identification response time of the terrain mutation area is less than 80 ms, which provides real-time and accurate environmental constraints for improving the path optimization of the swarm intelligence algorithm, and ensures that multiple unmanned aerial vehicles can still plan a safe and optimal three-dimensional path in a complex terrain and dynamic obstacle environment.

[0026] In the case implementation, the chaotic initialization strategy of the improved swarm intelligence algorithm in step S2 is realized based on the Logistic mapping, the adaptive inertia weight value range is 0.4-0.9, and is dynamically adjusted according to the iteration number or environmental complexity; the learning factor adaptively changes between 1.5-2.5, the population size is set to 50-100, and the maximum iteration number is 50-100; Further, in the algorithm iteration optimization process, a double-target-oriented termination condition is designed in combination with the multi-unmanned aerial vehicle cooperation scene: when the convergence accuracy of the multi-objective optimization function (path length, energy consumption, safety margin) of the path reaches 1e-4, or the iteration number reaches 80% of the maximum iteration number and there is no obvious optimization for 10 consecutive iterations, the algorithm automatically terminates. According to the test, the path optimization time of the improved swarm intelligence algorithm in the complex three-dimensional environment is controlled within 2 s, the deviation of the generated initial path from the theoretical optimal path is less than 3%, and when the population size is 80 and the maximum iteration number is 80, the convergence speed and optimization accuracy of the algorithm reach the best balance, which provides an efficient and high-quality initial solution for the multi-machine cooperation and dynamic path adjustment stage.

[0027] In the case implementation, the flight constraint conditions in step S3 include that the maximum flight speed is ≤30 m / s, the maximum climb / descent rate is ≤5 m / s, and the minimum turn radius is ≥5 m; the multi-machine safety distance can be dynamically adjusted, which is ≥5 m when the speed is ≤10 m / s, and is set to 0.8 times the speed value when the speed is >10 m / s, and is not less than 10 m; Further, in the multi-machine cooperative working scene, a real-time conflict detection and early warning model is constructed based on the above flight constraint conditions: through the real-time data of the positions, speeds and heading angles of multiple unmanned aerial vehicles, combined with the safety distance rules in the constraint conditions, the potential conflict risk is predicted 2 seconds in advance, when it is detected that the unmanned aerial vehicle is about to break through the minimum turning radius or the safety distance threshold, the path adjustment instruction is automatically triggered, under the premise of ensuring the stability of the flight attitude (such as the attitude angle change rate is not more than 10° / s during turning), the multi-machine is always kept within the constraint condition range by fine-tuning the speed and heading, in actual cases, this mechanism reduces the multi-machine conflict rate to below 0.1%, fully ensuring the safety and orderliness of the multi-unmanned aerial vehicle group working in three-dimensional space.

[0028] In the case implementation, the multi-objective optimization function in step S3 is constructed by weighted summation, the weight coefficients satisfy the sum of 1, the optimization indexes include path length (weight 0.3), flight energy consumption (weight 0.4), path smoothness, obstacle safety distance and multi-machine conflict coefficient (total weight 0.3), the smaller the target function value, the higher the path optimality; Further, through the multi-objective optimization function, the group intelligence algorithm can comprehensively weigh the multi-dimensional indexes such as path length, energy consumption, smoothness, safety and the like in the iterative optimization process, among them, the flight energy consumption weight is the highest (0.4), which reflects the importance of the endurance of the unmanned aerial vehicle; the path length (0.3) ensures the working efficiency; the path smoothness, obstacle safety distance and multi-machine conflict coefficient (total weight 0.3) ensure the stability and safety of the flight, this weighted summation construction method makes the smaller the target function value, the higher the optimality of the path in multiple dimensions, the finally output path scheme is not only high efficiency and energy saving, but also can ensure the safe cooperative work of the multi-unmanned aerial vehicle group in complex environment, effectively improving the practicability and reliability of the whole path planning system.

[0029] In the case implementation, the communication link in step S4 adopts 5G+WiFi6 dual mode or 2.4GHz wireless self-organizing network, the communication delay is ≤10ms, the multiple unmanned aerial vehicles share the position, speed and path information through the distributed peer-to-peer network architecture, the conflict detection response time is ≤100ms; Furthermore, in the instruction transmission and execution phase of dynamic path adjustment, a hierarchical instruction priority mechanism was constructed based on this communication architecture: path adjustment instructions involving safety conflicts were set to the highest priority, ensuring that the entire process from detection to instruction issuance was completed within 10ms; for path optimization instructions with normal environmental changes, an adaptive scheduling strategy was adopted to ensure execution efficiency. In actual testing, when encountering sudden obstacles (such as temporarily appearing construction equipment), the total time taken for multiple UAVs to complete local path replanning and execute the new path from detection did not exceed 200ms, and the flight stability after path adjustment was not affected. This fully verified the efficiency and reliability of the communication and decision-making mechanism in dynamic scenarios, and strongly supported the collaborative adaptability of multiple UAV swarms in complex operating environments.

[0030] In the implementation of the case, the environmental change threshold in step S4 is set as follows: the obstacle moves more than 1m, the wind speed changes more than 5m / s, the environmental model difference is >0.2 or a new obstacle appears, the path replanning response time does not exceed 200ms, and the smoothed path curvature change rate is ≤0.5rad / m. Furthermore, in the algorithm adaptation and effect verification of path replanning, targeted optimization strategies were designed for different types of environmental changes: when obstacle movement is detected, the focus is on optimizing the obstacle avoidance efficiency of the local path; when encountering sudden changes in wind speed, the focus is on balancing the energy consumption and stability of the path. After testing in actual scenarios, after environmental changes trigger replanning, the smoothness of the connection between the new path and the original path remains consistent, the attitude fluctuation amplitude during the flight of the UAV is less than 2°, and the loss in length and energy consumption of the replanned path is controlled within 5%, which fully ensures the continuity and efficiency of multi-UAV swarm operations in dynamic environments.

[0031] In summary, this intelligent dynamic path planning method for multi-UAV 3D swarms, which integrates multi-source environmental perception, can comprehensively and in real-time grasp environmental information such as terrain, obstacles, and weather in the work area through multi-source data fusion and accurate 3D environment modeling. This significantly improves the environmental adaptability of path planning, enabling UAV swarms to plan feasible paths even in complex and ever-changing scenarios. It solves the problems of low path planning accuracy, weak dynamic adaptability, and frequent multi-UAV conflicts caused by insufficient environmental perception and poor multi-UAV coordination in complex 3D environments in traditional technologies.

[0032] Furthermore, by improving the swarm intelligence algorithm and multi-drone collaboration mechanism, multi-objective optimization and dynamic adjustment of the path were achieved. While ensuring the optimal path length and the lowest energy consumption, multi-drone conflicts were effectively avoided, significantly improving the operational efficiency and collaborative safety of the drone swarm. This solved the problems of slow convergence speed and insufficient multi-objective optimization capability of traditional swarm intelligence algorithms in 3D path planning, as well as the problems of multiple path conflicts, low operational efficiency, and poor collaborative safety caused by information lag and imperfect decision-making mechanisms when multiple drones are working together.

[0033] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0034] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-unmanned aerial vehicle three-dimensional swarm intelligence dynamic path planning method fusing multi-source environment perception, comprising an intelligent dynamic path planning system, characterized in that: The operation steps are as follows: S1: multi-source environmental information collection and preprocessing; the airborne sensing system is composed of a visual sensor, a laser radar, a millimeter wave radar, a GPS / Beidou dual-mode positioning module and a weather sensor carried by a UAV, combined with ground station microwave communication data and satellite remote sensing information, to synchronously collect terrain elevation, obstacle distribution, weather conditions, electromagnetic interference, airspace restrictions and UAV self-state data; a filtering algorithm is used to denoise, time-space align and remove redundancies of the multi-source data, to generate a standardized data set; S2: multi-source information fusion and three-dimensional environment modeling; a weighted adaptive fusion algorithm is used to realize accurate fusion of multi-source heterogeneous data by combining Kalman filtering, D-S evidence theory or Bayesian estimation, and dynamically adjusting the weight of each data source; a dynamic three-dimensional environment model is constructed based on an octree structure, supporting real-time updating of obstacle state, terrain parameters and weather conditions, and the model resolution can be adaptively adjusted within the range of 0.1m-1m; S3: improved swarm intelligence algorithm initialization and path optimization; based on the particle swarm optimization algorithm, the chaos initialization strategy, adaptive inertia weight and dynamic learning factor are introduced to optimize the global search and local optimization ability of the algorithm; the starting point, ending point, flight constraint conditions and multi-objective optimization function of multiple UAVs are set to iteratively update the particle position to generate an initial three-dimensional path; S4: multi-UAV cooperation and dynamic path adjustment; multi-UAV data sharing is realized through 5G+WiFi6 dual-mode communication or wireless self-organizing network, and real-time monitoring of environmental changes and flight states is realized based on a distributed collaborative decision-making mechanism; when the detected environmental change exceeds the threshold or the distance between multiple UAVs is less than the safety threshold, local or global path re-planning is triggered to dynamically adjust the path parameters; S5: path smoothing output; the optimized path is smoothed by using a B-spline curve to eliminate sharp corners and ensure stable UAV flight attitude, and a three-dimensional coordinate sequence, speed suggestion value and energy consumption estimation data that can be directly called by the flight control system are output. 2.The method of claim 1, wherein the method further comprises: The intelligent dynamic path planning system comprises a multi-source environmental information collection module, a data preprocessing module, a multi-source information fusion module, a three-dimensional environment modeling module, an improved swarm intelligence algorithm module, a multi-UAV cooperation and dynamic adjustment module, a path smoothing output module and a UAV flight control interface module. 3.The method of claim 1, wherein the method further comprises: The sampling frequency of the multi-source sensor in step S1 is synchronized to 10Hz-20Hz, wherein the visual sensor realizes obstacle recognition and positioning through the YOLOv8 algorithm, the laser radar has a ranging accuracy within ±0.3m, the GPS / Beidou positioning module has a positioning error of ≤±0.5m, the millimeter wave radar has obstacle detection capability in adverse weather such as rain, snow and fog, and the preprocessing process adopts an algorithm combining Kalman filtering and wavelet transform, or a combination scheme of median filtering and Kalman filtering, and the data processing delay after noise suppression is controlled within 50ms; the preprocessing process in step S1 adopts an algorithm combining Kalman filtering and wavelet transform, or a combination scheme of median filtering and Kalman filtering, and the data processing delay after noise suppression is controlled within 50ms.

4. The method of claim 1, wherein the method further comprises: The weight coefficient update period of the weighted adaptive fusion algorithm in step S2 is 50 ms, the error of the fused environmental data is controlled within ±0.3 m, the three-dimensional environmental model supports the representation of complex terrain with an elevation difference of 0-500 m, and the data update frequency is synchronized with the sensor acquisition frequency.

5. The method of claim 1, wherein the method further comprises: The improved swarm intelligence algorithm in step S2 is based on the Logistic mapping for the chaotic initialization strategy, and the adaptive inertia weight value ranges from 0.4 to 0.9 and is dynamically adjusted according to the iteration number or environmental complexity. The learning factor is adaptively changed between 1.5 and 2.5, the population size is set to 50-100, and the maximum iteration number is 50-100. 6.The method of claim 1, wherein the method further comprises: The flight constraint conditions in step S3 include a maximum flight speed ≤30 m / s, a maximum climb / descent rate ≤5 m / s, and a minimum turn radius ≥5 m; the multi-vehicle safety distance can be dynamically adjusted, which is ≥5 m when the speed is ≤10 m / s, and is set to 0.8 times the speed value when the speed is >10 m / s, and is not less than 10 m.

7. The method of claim 1, wherein the method further comprises: The multi-objective optimization function in step S3 is constructed by weighted summation, the weight coefficients satisfy the sum of 1, and the optimization indexes include path length (weight 0.3), flight energy consumption (weight 0.4), path smoothness, obstacle safety distance, and multi-vehicle conflict coefficient (total weight 0.3), and the smaller the objective function value is, the higher the path optimality is. 8.The method of claim 1, wherein the method further comprises: The communication link in step S4 adopts 5G+WiFi6 dual-mode or 2.4 GHz wireless self-organizing network, the communication delay is ≤10 ms, multiple unmanned vehicles share position, speed and path information through a distributed peer-to-peer network architecture, and the conflict detection response time is ≤100 ms. 9.The method of claim 1, wherein the method further comprises: The environmental change threshold in step S4 is set as follows: the obstacle moving distance exceeds 1 m, the wind speed suddenly changes by more than 5 m / s, the environmental model difference is >0.2, or a new obstacle appears, the path re-planning response time is not more than 200 ms, and the curvature change rate of the smoothed path is ≤0.5 rad / m.

Images