A low-altitude aircraft flight behavior modeling method based on avionics data time sequence decoupling

Abstract

The present application relates to the technical field of avionics data processing, and particularly relates to a low-altitude aircraft flight behavior modeling method based on avionics data time sequence decoupling, comprising the following steps: S1, acquiring multi-source avionics data, identifying a compound action state formed by active manipulation, environmental compensation and navigation correction, and constructing a behavior coupling shielding atlas; S2, splitting continuous flight behavior occupation rights based on the behavior coupling shielding atlas, backtracking original triggering behaviors after compensation shielding, and generating a behavior responsibility decoupling chain; S3, identifying an implicit behavior migration path based on the behavior responsibility decoupling chain, and constructing a flight behavior hidden factor model. The present application realizes real triggering identification of complex flight behaviors, separation of compensation behaviors and analysis of implicit behavior evolution mechanism, and improves the accuracy, interpretability and complex scene adaptability of low-altitude aircraft flight behavior analysis.

Description

Technical Field

[0001] This invention relates to the field of avionics data processing technology, and in particular to a method for modeling the flight behavior of low-altitude aircraft based on temporal decoupling of avionics data. Background Technology

[0002] With the rapid development of the low-altitude economy, drone logistics, urban air traffic, and autonomous flight technology, the application scenarios of low-altitude aircraft in complex low-altitude environments are constantly increasing. During flight, low-altitude aircraft continuously generate multi-source avionics data, including flight control data, attitude change data, navigation and positioning data, power response data, and environmental perception data. Analyzing this avionics data enables flight status identification, flight behavior analysis, and abnormal behavior detection, which is crucial for ensuring flight safety, optimizing autonomous flight, and improving adaptability to complex environments. Therefore, accurately analyzing the real flight behavior of low-altitude aircraft using multi-source avionics data has become an important research direction in the field of low-altitude flight behavior analysis.

[0003] However, most existing technologies only perform single correlation analysis of flight behavior based on flight trajectory, attitude changes, or control commands. They typically treat continuous flight behavior as a linear process, making it difficult to distinguish the coupling relationships between active control behaviors, environmental disturbance compensation behaviors, and navigation drift correction behaviors. When low-altitude aircraft are in complex scenarios such as wind resistance correction, obstacle avoidance, low-altitude hovering, or navigating narrow areas, different driving sources can easily overlap, cancel each other out, and exhibit pseudo-continuous transmission phenomena. This causes the actual triggering behavior to be masked by environmental compensation or navigation correction actions, making it difficult for existing technologies to accurately identify the true triggering source of flight behavior. At the same time, existing technologies lack the ability to analyze implicit transfer relationships across flight behavior stages, failing to effectively explain the formation causes and potential evolution mechanisms of complex flight behaviors. This results in poor interpretability of flight behavior analysis results and insufficient adaptability to complex scenarios. Summary of the Invention

[0004] This invention provides a method for modeling the flight behavior of low-altitude aircraft based on temporal decoupling of avionics data. By temporally decoupling the behavioral driving sources in multi-source avionics data, the coupling and masking relationships between different driving sources under complex action states are identified, and a behavioral coupling and masking map is constructed. Furthermore, through behavioral occupancy right decomposition, behavioral continuity relationship severing, and responsibility backtracking analysis, a behavioral responsibility decoupling chain is generated. Finally, by cross-stage reorganization of the behavioral responsibility decoupling chain, implicit behavioral migration paths are identified, and a flight behavior latent cause model is constructed. This enables accurate modeling of the real triggering mechanism, compensation masking behavior, and implicit behavioral evolution path of complex flight behavior of low-altitude aircraft, improving the accuracy, interpretability, and adaptability to complex scenarios of low-altitude aircraft flight behavior analysis.

[0005] A method for modeling the flight behavior of low-altitude aircraft based on time-series decoupling of avionics data includes the following steps:

[0006] S1. Acquire multi-source avionics data generated by the low-altitude aircraft during continuous flight. Separate the behavioral driving sources of the flight control data, attitude change data, navigation and positioning data, power response data and environmental perception data in the multi-source avionics data. Identify the composite action state formed by the combined effects of active aircraft control, environmental disturbance compensation and navigation drift correction in the same time period. Construct the corresponding behavioral coupling masking map based on the mutual coverage, mutual cancellation and pseudo-continuous transmission relationships between different driving sources in the composite action state.

[0007] S2, based on the behavior coupling masking map, the occupancy rights of the continuous flight behavior generated by the low-altitude aircraft during flight are decomposed, the dominant contribution range of each driving source in attitude adjustment, path deviation, altitude change and heading correction is identified, and the behavior continuity relationship formed by pseudo-inheritance of environmental compensation actions is cut off. At the same time, reverse responsibility backtracking is performed on the cut-off flight behavior segments to identify the original triggering behavior hidden after the compensation action and generate the corresponding behavior responsibility decoupling chain. The behavior responsibility decoupling chain includes the original triggering behavior segment, the disturbance coverage behavior segment, the compensation takeover behavior segment and the drift correction behavior segment.

[0008] S3. Based on the aforementioned behavior responsibility decoupling chain, the original triggering behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment, and drift correction behavior segment in different flight phases are recombined across phases to identify implicit behavior migration paths formed by low-altitude aircraft during obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering, and narrow area traversal. Based on these implicit behavior migration paths, a latent cause model of low-altitude aircraft flight behavior is constructed.

[0009] Optionally, S1 includes:

[0010] S11: Acquire multi-source avionics data output from the flight control system, inertial navigation system, satellite navigation system, power system, and environmental perception system of the low-altitude aircraft, and perform timestamp correction, sampling frequency unification, and abnormal missing measurement supplementation according to a unified flight time axis to form a multi-source avionics synchronization sequence. The multi-source avionics synchronization sequence includes flight control sequence, attitude change sequence, navigation and positioning sequence, power response sequence, and environmental perception sequence.

[0011] S12, based on the multi-source avionics synchronization sequence, extract active control features, environmental disturbance features and navigation drift features respectively, and calculate the contribution weight of each type of driving source to the change of flight state within the same time window. When at least two types of contribution weights exceed the corresponding judgment threshold at the same time, mark the time window as a composite action state.

[0012] S13, for the composite action state, compare the sequential triggering relationship and contribution direction of different driving sources in attitude adjustment, heading deviation, altitude change and dynamic response, identify mutual coverage, mutual cancellation and pseudo-continuous transmission relationship, and construct a behavior coupling occlusion map with driving source as node and occlusion relationship as edge. The behavior coupling occlusion map is used to output the overlapping occupancy relationship, the actual triggering order and the occluded behavior segment between each driving source in the composite action state.

[0013] Optionally, S12 includes:

[0014] S121, Based on the multi-source avionics synchronization sequence, a continuous time window is constructed according to a preset time length and sliding step size, and local changes in the flight control sequence, attitude change sequence, navigation and positioning sequence, power response sequence and environmental perception sequence are extracted in each time window to form a window state change set;

[0015] S122, within each time window, active control features are extracted based on flight control sequence and dynamic response sequence, environmental disturbance features are extracted based on environmental perception sequence and attitude change sequence, and navigation drift features are extracted based on the deviation between navigation positioning sequence and inertial navigation attitude change sequence, generating active control contribution, environmental disturbance contribution and navigation drift contribution respectively.

[0016] S123, normalize the contribution of active manipulation, environmental disturbance and navigation drift to obtain the contribution weight of each type of driving source in the same time window. When at least two types of contribution weights in the same time window exceed their corresponding judgment thresholds, the time window is marked as a composite action state; otherwise, it is marked as a single driving state.

[0017] Optionally, S13 includes:

[0018] S131, for the time window marked as a composite action state, extract the first significant change moment in attitude adjustment, heading deviation, altitude change and dynamic response of the three types of driving sources: active control, environmental disturbance and navigation drift, and determine the sequential triggering relationship between different driving sources according to the first significant change moment;

[0019] S132, based on the first significant change time, contribution direction, and change amplitude, determine the occlusion relationship between any two types of driving sources, specifically including:

[0020] When the time difference between the first significant changes of two types of driving sources in the same flight state quantity is less than or equal to the trigger synchronization threshold, and the contribution directions are the same, it is determined that there is a mutual overlap relationship between the two types of driving sources.

[0021] When the two types of driving sources contribute in opposite directions to the same flight state quantity, and the change amplitude between the two types of driving sources is greater than the corresponding cancellation determination threshold, it is determined that there is a mutual cancellation relationship between the two types of driving sources.

[0022] When the first significant change of the latter driving source is later than the first significant change of the former driving source, and the direction of the state change of the latter driving source before its own triggering is consistent with the direction of contribution of the former driving source, and at the same time the change in control command, change in environmental perception, or change in navigation drift corresponding to the latter driving source does not reach its significant change threshold, it is determined that there is a pseudo-continuous transmission relationship between the two types of driving sources.

[0023] S133, based on the determination results of mutual coverage relationship, mutual cancellation relationship and pseudo-continuous transmission relationship, takes active manipulation, environmental disturbance and navigation drift as nodes, and the driving source pair that satisfies any kind of occlusion relationship as edge, and writes the relationship type, corresponding flight state quantity, triggering order, change amplitude and occluded segment into the edge attribute to generate a behavior coupling occlusion map.

[0024] Optionally, S2 includes:

[0025] S21, based on the behavior coupling occlusion graph, read the edge attributes corresponding to each occlusion relationship edge, and combine them with the first... The contribution weights of each driving source to attitude adjustment, path deviation, altitude change and heading correction within each time window are calculated. The behavioral occupancy weight of each driving source on the corresponding flight state quantity is calculated. The driving source with the largest behavioral occupancy weight is determined as the dominant driving source of the flight state quantity in the current time window. The time windows in which the same dominant driving source is continuously maintained are merged into the dominant contribution interval.

[0026] S22, based on the behavior occupancy right and the pseudo-continuous transmission relationship in the behavior coupling occupancy map, determine whether there is a behavior continuity relationship formed by the pseudo-inheritance of environmental compensation actions between adjacent time windows, and generate the cut-off flight behavior segment.

[0027] S23, for the severed flight behavior segment, the preceding driving source with mutual coverage, mutual cancellation or pseudo-continuous transmission relationship with the segment is retrieved in reverse along the occlusion relationship edge in the behavior coupling occlusion map, and the responsibility backtracking score is calculated according to the triggering time, behavior occupancy right and occlusion relationship type of the preceding driving source. The behavior segment corresponding to the preceding driving source with the highest responsibility backtracking score is determined as the original triggering behavior hidden after the compensation action, and the behavior responsibility decoupling chain is generated in the order of original triggering behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment and drift correction behavior segment.

[0028] Optionally, S22 includes:

[0029] S221, based on the first The first time window and the first The behavior occupancy rights of the first time window are determined, and the first action within the adjacent time window is determined respectively. The dominant driving source of the flight-like state quantity is identified, and the direction of state change, changes in the dominant driving source, and changes in the behavior occupancy right are compared between adjacent time windows to form the behavior inheritance result between adjacent windows.

[0030] S222, the first A flight state change within a time window that is dominated by environmental disturbances and whose behavior occupies the threshold for compensation takeover is defined as a compensation takeover behavior. The compensation takeover action is formed within a time window, and the direction of its state change is the same as that of the first... The first time window is consistent, and at the same time the second When the change in the environmental trigger input corresponding to a time window does not reach the significant change threshold, and there is a pseudo-continuous transmission relationship edge pointing to the source of environmental disturbance in the behavior coupling occlusion graph, it is determined that there is row continuity between adjacent time windows formed by pseudo-inheritance of environmental compensation actions.

[0031] S223, when it is determined that there is a pseudo-inheritance relationship of environmental compensation actions between adjacent time windows, the first... The starting time of each time window is used as the cutting boundary. The original continuous flight behavior is split into pre-cut behavior segments and post-cut behavior segments. The time windows that continuously satisfy the pseudo-inheritance judgment condition are merged to generate the cut flight behavior segments.

[0032] Optionally, S23 includes:

[0033] S231, for the severed flight behavior segment, determine its corresponding severed driving source. Cut-off start time and corresponding flight state quantities and with the cut-off drive source With the endpoint as the target, the reverse search is performed along the occlusion relationship edges in the coupling occlusion graph to filter out the preceding driving sources that were triggered before the cutoff start time and have mutual coverage, mutual cancellation or pseudo-continuous transmission relationship with the cutoff driving source, forming a candidate preceding driving source set.

[0034] S232, for each candidate preceding driving source in the candidate preceding driving source set, read its first significant change time, the behavior occupancy right in the corresponding time window, and the occlusion relationship type with the cut-off driving source, and calculate the responsibility backtracking score according to the time proximity, behavior occupancy right size and occlusion relationship correction coefficient.

[0035] S233, the candidate preceding driving source with the highest responsibility backtracking score is determined as the original trigger driving source hidden after the compensation action, and the behavior segment corresponding to the original trigger driving source is extracted as the original trigger behavior segment. According to the occlusion relationship edge between the original trigger driving source and the cut-off driving source in the behavior coupling occlusion map, the disturbance coverage behavior segment, the compensation takeover behavior segment, and the drift correction behavior segment are extracted in sequence, and a behavior responsibility decoupling chain is generated according to the triggering time sequence and responsibility transmission sequence.

[0036] Optionally, S3 includes:

[0037] S31, based on the behavior responsibility decoupling chain, the original triggering behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment and drift correction behavior segment are grouped according to the flight stage, and the behavior segments with the same original triggering drive source, the same masking relationship type or the same state change direction in adjacent flight stages are connected across stages to generate a cross-stage behavior recombination sequence.

[0038] S32, Based on the cross-stage behavior recombination sequence, calculate the behavior migration matching degree of the low-altitude aircraft in obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering and narrow area passage scenarios respectively. When the behavior migration matching degree of the scenario reaches the migration judgment threshold of the scenario, the corresponding cross-stage behavior recombination sequence is identified as the implicit behavior migration path in the scenario.

[0039] S33. Based on the implicit behavior migration path, the original trigger behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment and drift correction behavior segment are mapped as latent cause nodes respectively, and the cross-stage connection relationship, occlusion relationship type and behavior migration matching result are mapped as latent cause edges to construct a latent cause model of low-altitude aircraft flight behavior.

[0040] Optionally, S32 includes:

[0041] S321. Based on the cross-stage behavior reorganization sequence, extract the behavior type, original triggering source, occlusion relationship type, state change direction, compensation takeover mark, drift correction mark and flight stage sequence of each behavior segment, and map them into scene migration feature vectors. Among them, obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering and narrow area passage are respectively configured with corresponding standard migration feature templates.

[0042] S322, the scene migration feature vector is matched with the standard migration feature templates corresponding to obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering and narrow area passage, respectively, and the behavior migration matching degree under each scenario is calculated.

[0043] S323, compare the behavior transfer matching degree in each scenario with the corresponding transfer judgment threshold, when the first... The cross-stage behavioral recombination sequence in the first When the behavior transfer matching degree in a low-altitude flight scenario reaches the corresponding transfer judgment threshold, the cross-stage behavior recombination sequence is marked as the first... The implicit behavior transfer path in low-altitude flight scenarios is recorded, along with its path label, start and end flight stages, constituent behavior segments, and matching scenarios.

[0044] Optionally, S33 includes:

[0045] S331, based on the implicit behavior migration path, extract the original trigger behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment and drift correction behavior segment in the path, and map them into original trigger hidden cause nodes, disturbance coverage hidden cause nodes, compensation takeover hidden cause nodes and drift correction hidden cause nodes according to the behavior segment type. At the same time, write the flight stage, corresponding driving source, state change direction, behavior occupancy right and start and end time of each hidden cause node into the node attributes to form a set of hidden cause nodes.

[0046] S332, in the implicit behavior migration path, according to the temporal adjacency relationship, cross-stage connection relationship, occlusion relationship type and behavior migration matching result between behavior segments, any adjacent latent cause nodes are connected. When there is a cross-stage behavior connection edge, occlusion relationship edge or the same implicit behavior migration path belonging relationship between the behavior segments corresponding to two latent cause nodes, a latent cause edge is generated between the two latent cause nodes, and the connection source, occlusion relationship type, migration scene category and behavior migration matching degree are written into the edge attribute.

[0047] S333, Based on the set of latent cause nodes and the set of latent cause edges, a latent cause model of low-altitude aircraft flight behavior is constructed, and the cross-stage connection relationship, occlusion relationship type and behavior migration matching result corresponding to each latent cause edge are written into the latent cause model of low-altitude aircraft flight behavior as a set of edge attributes.

[0048] The beneficial effects of this invention are:

[0049] This invention, by unifying the temporal alignment and separating the behavioral driving sources of flight control data, attitude change data, navigation and positioning data, power response data, and environmental perception data generated by low-altitude aircraft during continuous flight, can effectively identify the composite action state formed between active manipulation, environmental disturbance compensation, and navigation drift correction. Furthermore, it constructs a behavioral coupling masking map, thereby solving the problem in the prior art where multiple types of driving behaviors overlap and cancel each other out, making it difficult to identify the actual flight triggering behavior, and improving the accuracy and temporal correlation capability of low-altitude aircraft flight behavior analysis.

[0050] This invention, through the splitting of occupancy rights, the severing of pseudo-inheritance behaviors, and the analysis of responsibility backtracking in continuous flight behavior, can reversely identify the original triggering behavior that was obscured from environmental compensation actions, and generate a behavior responsibility decoupling chain including the original triggering behavior segment, the disturbance coverage behavior segment, the compensation takeover behavior segment, and the drift correction behavior segment. This solves the problem of confusion between compensation behavior and real control behavior in the existing flight behavior analysis process, and improves the ability to track behavior responsibility and identify abnormal behavior in complex flight states of low-altitude aircraft.

[0051] This invention restructures the behavioral responsibility decoupling chain across different flight phases and constructs a latent cause model of low-altitude aircraft flight behavior by combining implicit behavioral migration paths. This model can identify the implicit behavioral evolution mechanisms of low-altitude aircraft in complex scenarios such as obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering, and narrow area navigation. This solves the problem that existing technologies are unable to deeply model the causes of complex behaviors of low-altitude aircraft, and improves the interpretability, scenario adaptability, and implicit risk identification capabilities of low-altitude aircraft flight behavior analysis. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a schematic diagram of the modeling method flow according to an embodiment of the present invention;

[0054] Figure 2 This is a schematic diagram of the behavior coupling masking map construction in an embodiment of the present invention. Detailed Implementation

[0055] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Those skilled in the art may employ other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0056] like Figures 1-2 As shown, a method for modeling the flight behavior of low-altitude aircraft based on temporal decoupling of avionics data includes the following steps:

[0057] S1. Acquire multi-source avionics data generated by the low-altitude aircraft during continuous flight. Separate the behavioral driving sources from the flight control data, attitude change data, navigation and positioning data, power response data, and environmental perception data in the multi-source avionics data. Identify the composite action state formed by the combined effects of active aircraft control, environmental disturbance compensation, and navigation drift correction within the same time period. Based on the mutual coverage, mutual cancellation, and pseudo-continuous transmission relationships between different driving sources in the composite action state, construct the corresponding behavioral coupling masking map to characterize the overlapping and occupancy relationships of multiple driving sources in the same flight behavior and the actual behavior triggering order.

[0058] S2, based on the behavior coupling masking map, performs occupancy rights decomposition on the continuous flight behavior generated by the low-altitude aircraft during flight, identifies the dominant contribution range of each driving source in attitude adjustment, path deviation, altitude change and heading correction, and cuts off the behavior continuity relationship formed by pseudo-inheritance of environmental compensation actions. At the same time, reverse responsibility backtracking is performed on the cut-off flight behavior segments to identify the original triggering behavior hidden after the compensation action and generate the corresponding behavior responsibility decoupling chain. The behavior responsibility decoupling chain includes the original triggering behavior segment, the disturbance coverage behavior segment, the compensation takeover behavior segment and the drift correction behavior segment.

[0059] S3, based on the behavior responsibility decoupling chain, reorganizes the original trigger behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment, and drift correction behavior segment across different flight phases. It identifies the implicit behavior migration paths formed by low-altitude aircraft during obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering, and narrow area navigation. Based on the implicit behavior migration paths, it constructs a latent cause model of low-altitude aircraft flight behavior. The latent cause model of flight behavior is used to distinguish whether the current flight behavior of low-altitude aircraft belongs to real control behavior, environmental inducement behavior, compensation cover-up behavior, or navigation drift pseudo-behavior.

[0060] S1 includes:

[0061] S11 acquires multi-source avionics data from the low-altitude aircraft's flight control system, inertial navigation system, satellite navigation system, power system, and environmental perception system. This data is then corrected for timestamps, sampled for frequency uniformity, and missing data is supplemented according to a unified flight timeline to form a multi-source avionics synchronization sequence. This sequence includes flight control sequence, attitude change sequence, navigation and positioning sequence, power response sequence, and environmental perception sequence, specifically:

[0062] (1) Raw avionics data are obtained from the flight control system, inertial navigation system, satellite navigation system, power system, and environmental perception system of the low-altitude aircraft to form a raw multi-source avionics dataset. Each type of data carries the original timestamp, sample value, and data source identifier, as follows:

[0063] ;

[0064] in, For the first Avionics data, , Indicates the flight control system. Indicates inertial navigation system, Indicates satellite navigation system, Indicates the power system, This represents an environmental perception system. For the first In the avionics data of the first The original timestamps of each sampling point For the corresponding sampled values, For data source identification, For the first The number of sampling points for avionics data;

[0065] (2) Using the flight control system time or the airborne unified time synchronization module time as the reference time, offset correction and drift correction are performed on the original timestamps of various avionics data, and the sampling points of different systems are mapped to the unified flight time axis to obtain the corrected timestamp, which is expressed as:

[0066] ;

[0067] in, This is the corrected timestamp. For the first Avionics data with a fixed time offset relative to a reference time. For the first Clock drift coefficients for avionics data;

[0068] (3) Set a uniform sampling period based on the target modeling accuracy, construct a uniform flight time axis, and resample the corrected avionics data to this time axis. For data not directly collected at the target time, interpolate and fill in the gaps based on adjacent sampling points to form various synchronization subsequences, as follows:

[0069] ;

[0070] ;

[0071] in, To unify the first on the flight timeline At that moment, The reference time for the start of the flight. To standardize the sampling period, To unify the total number of sampling points on the time axis, For the first Avionics data at a unified time Synchronous sampling values ​​below, , They are located at Two adjacent correction timestamps , These are the corresponding sampled values;

[0072] This ultimately forms a multi-source avionics synchronization sequence:

[0073] ;

[0074] in, To unify the time The following multi-source avionics synchronization sequence, This is the flight control sequence. This is a sequence of attitude changes. For navigation and positioning sequences, For dynamic response sequence, This is an environmental perception sequence;

[0075] S12, based on the multi-source avionics synchronization sequence, extract active control features, environmental disturbance features and navigation drift features respectively, and calculate the contribution weight of each type of driving source to the change of flight state within the same time window. When at least two types of contribution weights exceed the corresponding judgment threshold at the same time, the time window is marked as a composite action state.

[0076] S13, for the complex action state, compare the sequential triggering relationship and contribution direction of different driving sources in attitude adjustment, heading deviation, altitude change and dynamic response, identify mutual coverage, mutual cancellation and pseudo-continuous transmission relationship, and construct behavior coupling occlusion map with driving source as node and occlusion relationship as edge. The behavior coupling occlusion map is used to output the overlapping occupancy relationship, the actual triggering order and the occluded behavior segment between each driving source in the complex action state.

[0077] S12 includes:

[0078] S121, based on multi-source avionics synchronization sequences, constructs continuous time windows according to preset time lengths and sliding step sizes, and extracts local changes in flight control sequences, attitude change sequences, navigation and positioning sequences, dynamic response sequences, and environmental perception sequences within each time window to form a window state change set, represented as:

[0079] ;

[0080] in, For the first A time window, To unify the time The following multi-source avionics synchronization sequence, For the first The start time of each time window The length of the time window;

[0081] S122, within each time window, active control features are extracted based on the flight control sequence and dynamic response sequence; environmental disturbance features are extracted based on the environmental perception sequence and attitude change sequence; and navigation drift features are extracted based on the deviation between the navigation and positioning sequence and the inertial navigation attitude change sequence. Active control contribution, environmental disturbance contribution, and navigation drift contribution are generated respectively, and expressed as follows:

[0082] ;

[0083] ;

[0084] ;

[0085] in, , , The first The contributions of active manipulation, environmental disturbances, and navigation drift within a given time window. This refers to the changes in flight control parameters. For the change in dynamic response, This refers to the perceived changes in the environment. This refers to attitude changes that are not explained by changes in flight control. For navigation and positioning results, The position estimation result is obtained by inertial navigation integration. , , , , , These are the corresponding weight coefficients;

[0086] S123, normalize the contributions of active manipulation, environmental disturbance, and navigation drift to obtain the contribution weights of each type of driving source within the same time window. When at least two types of contribution weights exceed their corresponding judgment thresholds within the same time window, the time window is marked as a composite action state; otherwise, it is marked as a single driving state, as shown below:

[0087] ;

[0088] ;

[0089] ;

[0090] in, For the first Class-driven source in the first Contribution weight within each time window , To actively manipulate, For environmental disturbance, For navigation drift, For the first The contribution of class-driven sources. To prevent extremely small constants with a denominator of zero, For the first The result of determining whether the class-driven source exceeds the threshold. For the first The threshold for determining the source of the class driver. For the first The status marking results for each time window;

[0091] ;

[0092] ;

[0093] in, , The first The average contribution weight and standard deviation of the contribution weight of class-driven sources in historical normal flight samples. For the first Fluctuation amplification factor from class-driven sources, For the current flight phase Correction amount for environmental disturbances caused by class-driven sources. This is the environmental disturbance correction factor. The intensity of environmental disturbance during the current flight phase. This represents the baseline environmental disturbance intensity under historical normal flight conditions.

[0094] S13 includes:

[0095] S131, for the time window marked as a composite action state, extract the first significant change moments in attitude adjustment, heading deviation, altitude change, and dynamic response from the three types of driving sources: active control, environmental disturbance, and navigation drift. Then, determine the sequential triggering relationship between different driving sources according to the first significant change moment, as shown below:

[0096] ;

[0097] in, For the first Class-driven source in the first The moment of the first significant change in flight-like state quantities , For posture adjustment, For heading deviation, Due to high degree of variation, For dynamic response, For the first Class-driven origins at a unified moment Next to the The contribution of changes in flight-like state quantities For the first Class-driven source in the first Significant change thresholds in flight-like state quantities;

[0098] ;

[0099] in, , These are the first in the historical normal flight samples. Class-driven source in the first The changes in flight-like state quantities contribute to the mean and standard deviation. The sensitivity coefficient is significantly changed;

[0100] S132, based on the time of the first significant change, the direction of contribution, and the magnitude of the change, determines the occlusion relationship between any two types of driving sources, specifically including:

[0101] When the time difference between the first significant changes of two types of driving sources in the same flight state variable is less than or equal to the triggering synchronization threshold, and their contribution directions are the same, it is determined that there is a mutual overlap relationship between the two types of driving sources, expressed as:

[0102] ;

[0103] in, For the first Class and the Class-driven source in the first The result of mutual overlap determination on flight state variables. For the first The trigger synchronization threshold corresponding to the flight state quantity, , The first Class and the Contribution directions from class-driven sources;

[0104] ;

[0105] in, , These represent the mean and standard deviation of the effective trigger time difference for different drive sources within the same flight state quantity in historical samples. This refers to the synchronization tolerance coefficient.

[0106] ;

[0107] in, For the first Class-driven source for the first The direction of contribution of flight-like state quantities. This is a sign function, and its value can be positive, negative, or zero.

[0108] When the two types of driving sources contribute in opposite directions to the same flight state variable, and the magnitude of the changes in both types of driving sources is greater than the corresponding cancellation threshold, it is determined that there is a mutual cancellation relationship between the two types of driving sources, which is expressed as:

[0109] ;

[0110] in, For the first Class and the Class-driven source in the first The mutual cancellation determination results of flight state quantities , The first Class and the The threshold for offsetting class-driven sources;

[0111] ;

[0112] in, , These are the first in the historical normal samples. The mean and standard deviation of the absolute values ​​of the contributions of class-driven source changes. To offset the sensitivity coefficient, For the first Class-driven source in the first The threshold for offset determination on flight-like state quantities;

[0113] When the first significant change of the latter driving source occurs later than the first significant change of the former driving source, and the direction of the state change of the latter driving source before its own triggering is consistent with the contribution direction of the former driving source, and the change in control command, environmental perception, or navigation drift corresponding to the latter driving source does not reach its significant change threshold, a pseudo-continuous transmission relationship is determined to exist between the two types of driving sources, expressed as:

[0114] ;

[0115] in, For the first Class and the Class-driven source in the first The result of pseudo-continuous transmission determination on flight-like state quantities. For the first Class-driven sources at the moment of their first significant change Previously, and located in the first The moment when the class-driven source first significantly changes The direction of the subsequent state change, For the first The change in the trigger input corresponding to the class-driven source;

[0116] ;

[0117] in, This is the pre-observation interval used to determine pseudo-continuous propagation;

[0118] ;

[0119] in, For the first The class-driven source is the trigger input value at the moment of the first significant change. For its length of pre-observation time The previous trigger input value, To preset the pre-observation duration;

[0120] S133, based on the determination results of mutual coverage, mutual cancellation, and pseudo-continuous transmission relationships, takes active manipulation, environmental disturbance, and navigation drift as nodes, and the driving source pairs that satisfy any kind of occlusion relationship as edges. The relationship type, corresponding flight state quantity, triggering order, change amplitude, and occluded segment are written into the edge attributes to generate a behavior-coupled occlusion graph, represented as:

[0121] ;

[0122] ;

[0123] ;

[0124] ;

[0125] ;

[0126] in, For behavior coupling masking map. To drive the set of source nodes, For the set of edges with occlusion relation, For the set of edge attributes, For the first Class-driven origins and the first Class-driven source in the first The occlusion relationship edges formed on the flight-like state variables, For the edge The corresponding attributes The types of occlusion relationships include mutual coverage, mutual cancellation, or pseudo-continuous conduction. For the corresponding flight state quantities, To determine the order of triggering, The magnitudes of the changes from the two types of driving sources, The obscured behavior segment.

[0127] S2 includes:

[0128] S21, based on the behavior-coupled occlusion graph, read the edge attributes corresponding to each occlusion relationship edge, and combine them with the first... Within each time window, the contribution weights of each driving source to attitude adjustment, path deviation, altitude change, and heading correction are calculated. The behavioral occupancy weight of each driving source on the corresponding flight state variable is then determined. The driving source with the largest behavioral occupancy weight is identified as the dominant driving source for that flight state variable within the current time window. Time windows where the same dominant driving source is continuously maintained are merged into a dominant contribution interval, represented as:

[0129] ;

[0130] ;

[0131] in, For the first Class-driven source in the first Within the time window, for the first The right to occupy the behavior of flight-like state quantities. For the first Class-driven source in the first Contribution weight within each time window For the first Class-driven source for the first The magnitude of the contribution of changes in flight-like state quantities. To prevent extremely small constants with a denominator of zero, For the first Flight-like state quantities in the first The dominant driving force within a time window;

[0132] S22, based on the pseudo-continuous transmission relationship in the behavior occupancy right and behavior coupling occupancy map, determine whether there is a behavior continuity relationship formed by the pseudo-inheritance of environmental compensation actions between adjacent time windows, and generate the cut-off flight behavior fragments.

[0133] S23, for the severed flight behavior segment, reverse the search along the masking relationship edge in the behavior coupling masking map to find the preceding driving source that has a mutual covering, mutual cancellation or pseudo-continuous transmission relationship with the segment, and calculate the responsibility backtracking score according to the triggering time, behavior occupancy right and masking relationship type of the preceding driving source. The behavior segment corresponding to the preceding driving source with the highest responsibility backtracking score is determined as the original triggering behavior hidden after the compensation action, and a behavior responsibility decoupling chain is generated in the order of original triggering behavior segment, disturbance covering behavior segment, compensation takeover behavior segment and drift correction behavior segment.

[0134] S22 includes:

[0135] S221, based on the first The first time window and the first The behavior occupancy rights of the first time window are determined, and the first action within the adjacent time window is determined respectively. The dominant driving force of flight-like state variables is identified, and the direction of state change, changes in the dominant driving force, and changes in the behavior occupancy weight are compared between adjacent time windows to form the behavior inheritance result between adjacent windows, represented as follows:

[0136] ;

[0137] ;

[0138] in, For the first Within the first time window The dominant driving force of flight-like state quantities For the first Class-driven source in the first Within the time window, for the first The right to occupy the behavior of flight-like state quantities. This is the result of determining the consistency of the direction of state changes between adjacent time windows. , The first The first time window and the first Within the first time window The direction of state change of flight-like state quantities;

[0139] S222, the first A flight state change within a time window that is dominated by environmental disturbances and whose behavior occupies the threshold for compensation takeover is defined as a compensation takeover behavior. The compensation takeover action is formed within a time window, and the direction of its state change is the same as that of the first... The first time window is consistent, and at the same time the second When the change in the environmental trigger input corresponding to a time window does not reach a significant change threshold, and there are pseudo-continuous transmission relationship edges pointing to the source of environmental disturbance in the behavior coupling occlusion graph, it is determined that there is row continuity between adjacent time windows formed by pseudo-inheritance of environmental compensation actions, which is represented as:

[0140] ;

[0141] ;

[0142] ;

[0143] ;

[0144] in, For the first The determination of whether a compensatory takeover action has been taken within a specific time window. For the environmental disturbance driving source in the first Within the time window, for the first The right to occupy the behavior of flight-like state quantities. To compensate for the threshold for taking over the case, For the first The first time window and the first Does a pseudo-inheritance relationship exist between environmental compensation actions across different time windows? The change in the trigger input corresponding to the source of the environmental disturbance. The threshold for significant changes corresponding to the driving sources of environmental disturbances. For the first The result of determining the pseudo-continuous transmission relationship between class-driven sources and environmental disturbance-driven sources. , The environmental disturbance driving sources in historical normal flight samples are respectively related to the first Mean and standard deviation of the behavioral occupancy weights for flight-like state quantities. To compensate for the sensitivity coefficient of the receiver, For the current flight phase Environmental disturbance enhancement factor corresponding to flight-like state quantities This is the environmental disturbance correction factor. The first in the current flight phase The real-time environmental disturbance intensity corresponding to flight-like state quantities. For the first time under normal flight conditions in history The intensity of the baseline environmental disturbance corresponding to the flight-like state quantity;

[0145] S223, when it is determined that there is a pseudo-inheritance relationship of environmental compensation actions between adjacent time windows, the first... Using the start time of each time window as the cutting boundary, the original continuous flight behavior is split into pre-cut behavior segments and post-cut behavior segments. Time windows that continuously satisfy the pseudo-inheritance condition are merged to generate the cut flight behavior segment, represented as:

[0146] ;

[0147] ;

[0148] in, For the first Flight-like state quantities in the first The first time window and the first The behavior cuts off the boundary between time windows. For the first The start time of each time window To avoid performing a cutoff, For the generated fragments of truncated flight behavior, For the first A time window, The number of time windows during which the conditions for determining false inheritance are met consecutively.

[0149] S23 includes:

[0150] S231, for the severed flight behavior segment, determine the corresponding severed driving source. Cut-off start time and corresponding flight state quantities and with the source of the cut-off drive With the endpoint as the starting point, a reverse search is performed along the occlusion relationship edges in the coupling occlusion graph. Precursor sources that were triggered before the cutoff start time and have mutual overlap, mutual cancellation, or pseudo-continuous transmission relationships with the cutoff driving source are selected to form a candidate precursor source set, denoted as:

[0151] ;

[0152] in, To target the cut-off driving source In the The set of candidate preceding drivers for flight-like state variables. As a candidate preceding driving source, The driving source was cut off. For the first Class-driven source points to the first Class-driven occlusion relationship edges, Let be the set of occlusion relation edges in the behavior coupling occlusion graph. For the first Class-driven origins and the first The types of occlusion relationships between class-driven sources. For mutual coverage relationship, Indicates a mutual offsetting relationship. Indicates a pseudo-continuous conduction relationship. For the first Class-driven source in the first The moment of the first significant change in flight-like state quantities The moment when the severed flight behavior segment was cut off;

[0153] S232, for each candidate preceding driving source in the candidate preceding driving source set, read its first significant change time, the behavior occupancy weight within the corresponding time window, and the occlusion relationship type with the cut-off driving source, and calculate the responsibility backtracking score according to the temporal proximity, behavior occupancy weight, and occlusion relationship correction coefficient, as shown below:

[0154] ;

[0155] ;

[0156] in, For the first Candidate preceding driver source to cut-off driver source In the Responsibility retrospective score in flight-like state quantities For the first The candidate class pre-processor source is in its triggering time window The right to occupy within the premises, For the first The moment when the candidate preceding driving source first significantly changes The time window to which it belongs The start time of the cut-off of the segment of flight behavior. For the first The time scale for responsibility backtracking corresponding to flight-like state quantities. This is a correction factor for the occlusion relationship type. This is a correction factor for mutual coverage relationships. This is a correction factor for mutual cancellation. This is a correction factor for pseudo-continuous conduction relations;

[0157] S233, the candidate preceding driving source with the highest responsibility backtracking score is identified as the original triggering driving source hidden after the compensation action. The behavior segment corresponding to this original triggering driving source is extracted as the original triggering behavior segment. Based on the occlusion relationship edge from the original triggering driving source to the cut-off driving source in the behavior coupling occlusion map, the disturbance coverage behavior segment, the compensation takeover behavior segment, and the drift correction behavior segment are extracted in sequence. A behavior responsibility decoupling chain is generated according to the triggering sequence and responsibility transmission sequence, as shown below:

[0158] ;

[0159] ;

[0160] in, The original trigger driver with the highest responsibility tracing score. For the first The behavioral responsibility decoupling chain corresponding to the flight state quantity. Original trigger driver source The corresponding original trigger behavior segment, For the original trigger-driven source With the cut-off driving source The perturbation coverage behavior segment formed by the mutual coverage or mutual cancellation relationship between them. For the cut-off drive source The corresponding compensation and takeover action section, This refers to drift correction behavior segments formed before and after the same cut segment, driven by navigation drift sources.

[0161] S3 includes:

[0162] S31, based on the behavior responsibility decoupling chain, aggregates the original triggering behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment, and drift correction behavior segment according to the flight phase, and connects behavior segments with the same original triggering source, the same masking relationship type, or the same state change direction in adjacent flight phases across phases to generate a cross-phase behavior recombination sequence, specifically including:

[0163] (1) Based on the behavioral responsibility decoupling chain, according to at least one of the flight phases of takeoff, climb, cruise, obstacle avoidance, hovering, and landing, the original triggering behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment, and drift correction behavior segment are determined to belong to the same flight phase. Behavior segments belonging to the same flight phase are grouped into a phase behavior set, which is represented as:

[0164] ;

[0165] ;

[0166] ;

[0167] in, For behavior segment The corresponding flight phase, For the flight phase assembly, For behavior segment With the Matching degree of each flight phase For the first The set of stage behaviors corresponding to each flight stage. For the first The behavior corresponding to the flight state variable. For highly variable matching factors, For velocity variation matching factor, For attitude change matching factor, For path change matching factors, , , , Each stage is assigned a corresponding weight.

[0168] (2) For the set of stage behaviors corresponding to adjacent flight stages, compare the original triggering source, occlusion relationship type, and state change direction of the behavior segments. When the behavior segments in adjacent flight stages satisfy at least one connection condition, establish a cross-stage connection relationship between the corresponding behavior segments and generate a cross-stage behavior connection edge, as shown below:

[0169] ;

[0170] ;

[0171] in, For behavior segment With behavior segment Whether a cross-stage connection relationship is established between them , Behavior segment With behavior segment The corresponding original trigger driver source, , Behavior segments With behavior segment The corresponding occlusion relationship type, , Behavior segments With behavior segment The corresponding direction of state change, For the first The set of cross-stage behavior connection edges corresponding to flight state variables. For behavior segment With behavior segment Cross-stage behavior connection edges between;

[0172] (3) Based on the set of cross-stage behavior connection edges, connect the corresponding behavior segments sequentially according to the flight stage order, and merge behavior segments with continuous cross-stage connection relationships to generate a cross-stage behavior recombination sequence; when there is a cross-stage behavior connection edge between the next behavior segment and the previous behavior segment, append the next behavior segment to the current cross-stage behavior recombination sequence; otherwise, generate a new cross-stage behavior recombination sequence, represented as:

[0173] ;

[0174] ;

[0175] in, For the first A cross-stage behavioral reorganization sequence, For the first The first in the cross-stage behavioral reorganization sequence Each behavior segment, For the first The number of behavioral segments in a cross-stage behavioral reorganization sequence. For the first The first behavior segment and the first Cross-stage behavior connection edges between behavior segments Represents the set of edges connecting cross-stage behaviors;

[0176] S32, based on the cross-stage behavior recombination sequence, calculate the behavior migration matching degree of the low-altitude aircraft in obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering and narrow area passage scenarios respectively. When the behavior migration matching degree of the scenario reaches the migration judgment threshold of the scenario, the corresponding cross-stage behavior recombination sequence is identified as the implicit behavior migration path in the scenario.

[0177] S33. Based on the implicit behavior migration path, the original triggering behavior segment, the disturbance coverage behavior segment, the compensation takeover behavior segment, and the drift correction behavior segment are mapped as latent cause nodes, and the cross-stage connection relationship, the occlusion relationship type, and the behavior migration matching result are mapped as latent cause edges to construct a latent cause model of low-altitude aircraft flight behavior. The flight behavior latent cause model is used to represent the implicit evolution mechanism of flight behavior from original triggering, disturbance coverage, compensation takeover to drift correction under different low-altitude flight scenarios.

[0178] S32 includes:

[0179] S321, based on the cross-stage behavior reorganization sequence, extracts the behavior type, original triggering source, occlusion relationship type, state change direction, compensation takeover marker, drift correction marker, and flight stage sequence of each behavior segment, and maps them to a scene migration feature vector. Obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering, and narrow area navigation are each configured with corresponding standard migration feature templates, represented as follows:

[0180] ;

[0181] in, For the first Scene migration feature vectors corresponding to cross-stage behavior recombination sequences. For the first The first in the cross-stage behavior recombination sequence One transfer feature, This represents the number of transfer features;

[0182] S322, the scene transfer feature vector is matched with the standard transfer feature templates corresponding to obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering, and narrow area passage, respectively, and the behavior transfer matching degree under each scenario is calculated, expressed as:

[0183] ;

[0184] in, For the first The cross-stage behavioral reorganization sequence and the first Behavioral transfer matching degree between low-altitude flight scenarios These respectively represent obstacle avoidance and flight maneuvering, wind resistance correction, instability recovery, low-altitude hovering, and navigating narrow areas. For the first The first in the standardized transfer feature template corresponding to low-altitude flight scenarios One standard feature, For the first In low-altitude flight scenarios, the first The weights of each migration feature, For the first The cross-stage behavioral reorganization sequence and the first Consistency of phase sequence between low-altitude flight scenarios For the first The integrity of key behavioral segments in a cross-stage behavioral reorganization sequence. , , These are the weighting coefficients for feature consistency, stage sequence consistency, and key behavior segment completeness, respectively. To prevent extremely small constants with a denominator of zero;

[0185] ;

[0186] ;

[0187] in, For the first The number of standard flight phases corresponding to low-altitude flight scenarios. For the first The first cross-stage behavioral reorganization sequence The first behavior segment and the first The actual stage transition relationship between each behavior segment For the first The first type of low-altitude flight scenario The standard stage migration relationship, For phase transition consistency function, To prevent extremely small constants with a denominator of zero;

[0188] ;

[0189] ;

[0190] in, For the first The number of key behavior segments corresponding to low-altitude flight scenarios. For the first In low-altitude flight scenarios, the first The importance weight of each key behavior segment For the first Does the sequence of cross-stage behavioral reorganization include the first one? Key behavioral segments, For the first The first type of low-altitude flight scenario Key behavior segment template For the first A cross-stage behavior reorganization sequence;

[0191] S323, compare the behavior transfer matching degree in each scenario with the corresponding transfer judgment threshold, when the first... The cross-stage behavioral recombination sequence in the first When the behavior transfer matching degree in a low-altitude flight scenario reaches the corresponding transfer judgment threshold, the cross-stage behavior recombination sequence is marked as the first... The implicit behavior transfer path in low-altitude flight scenarios is recorded, including its path label, start and end flight stages, constituent behavior segments, and matching scenario, and is represented as follows:

[0192] ;

[0193] ;

[0194] in, For the first Whether the cross-stage behavioral recombination sequence is identified as the first Implicit behavior transfer paths in low-altitude flight-like scenarios For the first Migration determination thresholds for low-altitude flight scenarios For the first A set of implicit behavior transfer paths in low-altitude flight scenarios. For the first A cross-stage behavior reorganization sequence;

[0195] ;

[0196] ;

[0197] in, , These are the first in the historical normal flight samples. The mean and standard deviation of the behavioral transfer matching degree corresponding to low-altitude flight scenarios. For migration sensitivity coefficient, For the first Environmental complexity correction factor for low-altitude flight scenarios This is a correction factor for environmental complexity. For the first time in the current flight environment Real-time environmental complexity corresponding to low-altitude flight scenarios. For the first time under normal flight conditions in history The baseline environmental complexity corresponding to low-altitude flight scenarios.

[0198] S33 includes:

[0199] S331, based on the implicit behavior migration path, extracts the original triggering behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment, and drift correction behavior segment from the path, and maps them to original triggering latent cause nodes, disturbance coverage latent cause nodes, compensation takeover latent cause nodes, and drift correction latent cause nodes according to the behavior segment type. Simultaneously, the flight stage, corresponding driving source, state change direction, behavior occupancy right, and start and end times of each latent cause node are written into the node attributes, forming a set of latent cause nodes, represented as:

[0200] ;

[0201] ;

[0202] in, For the set of hidden cause nodes, For the action segment Latent nodes formed by mapping, For the first Implicit behavior transfer paths in low-altitude flight-like scenarios For behavior segment Behavior segment types, including , , , These represent the original trigger behavior segment, the disturbance coverage behavior segment, the compensation takeover behavior segment, and the drift correction behavior segment, respectively. For behavior segment The flight phase to which it belongs For behavior segment The corresponding driver source, For behavior segment The corresponding direction of state change, For behavior segment The corresponding right of possession, , Behavior segments The start and end times;

[0203] S332, in the implicit behavior migration path, according to the temporal adjacency relationship, cross-stage connection relationship, occlusion relationship type, and behavior migration matching result between behavior segments, a connection determination is made for any adjacent latent cause nodes. When there is a cross-stage behavior connection edge, occlusion relationship edge, or same implicit behavior migration path affiliation relationship between the behavior segments corresponding to two latent cause nodes, a latent cause edge is generated between the two latent cause nodes, and the connection source, occlusion relationship type, migration scene category, and behavior migration matching degree are written into the edge attribute, as follows:

[0204] ;

[0205] ;

[0206] in, Let be the set of hidden cause edges. For the hidden cause node Pointing to the hidden cause node The hidden cause edge, Hidden cause node Pointing to the hidden cause node The result of determining whether a hidden cause edge is established between them. For behavior segment With behavior segment Cross-stage behavior connection edges between them For cross-stage behavior connection edge set, For behavior segment With behavior segment The occlusion relationship between them Let be the set of occlusion relation edges in the behavior coupling occlusion graph. For the first Whether the cross-stage behavioral recombination sequence is identified as the first Implicit behavior transfer paths in low-altitude flight scenarios;

[0207] S333, based on the set of latent cause nodes and the set of latent cause edges, constructs a latent cause model of low-altitude aircraft flight behavior. The cross-stage connection relationship, occlusion relationship type, and behavior migration matching result corresponding to each latent cause edge are written into the latent cause model of low-altitude aircraft flight behavior as an edge attribute set. This enables the flight behavior latent cause model to represent the implicit behavioral evolution path of low-altitude aircraft in different low-altitude flight scenarios, from original triggering, disturbance coverage, compensation takeover to drift correction, as shown below:

[0208] ;

[0209] ;

[0210] ;

[0211] in, For the latent cause model of low-altitude aircraft flight behavior, For the set of hidden cause nodes, Let be the set of hidden cause edges. For the set of hidden edge attributes, For hidden cause edge The corresponding edge attributes, , Behavior segments , The flight phase to which it belongs For behavior segment With behavior segment Types of occlusion relationships between them For the first The cross-stage behavioral reorganization sequence and the first Behavioral transfer matching degree between low-altitude flight scenarios This is a low-altitude flight scenario category.

[0212] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0213] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for modeling the flight behavior of low-altitude aircraft based on temporal decoupling of avionics data, characterized in that, Includes the following steps: S1. Acquire multi-source avionics data generated by the low-altitude aircraft during continuous flight. Separate the behavioral driving sources of the flight control data, attitude change data, navigation and positioning data, power response data and environmental perception data in the multi-source avionics data. Identify the composite action state formed by the combined effects of active aircraft control, environmental disturbance compensation and navigation drift correction in the same time period. Construct the corresponding behavioral coupling masking map based on the mutual coverage, mutual cancellation and pseudo-continuous transmission relationships between different driving sources in the composite action state. S2, based on the behavior coupling masking map, the occupancy rights of the continuous flight behavior generated by the low-altitude aircraft during flight are decomposed, the dominant contribution range of each driving source in attitude adjustment, path deviation, altitude change and heading correction is identified, and the behavior continuity relationship formed by pseudo-inheritance of environmental compensation actions is cut off. At the same time, reverse responsibility backtracking is performed on the cut-off flight behavior segments to identify the original triggering behavior hidden after the compensation action and generate the corresponding behavior responsibility decoupling chain. The behavior responsibility decoupling chain includes the original triggering behavior segment, the disturbance coverage behavior segment, the compensation takeover behavior segment and the drift correction behavior segment. S3. Based on the aforementioned behavior responsibility decoupling chain, the original triggering behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment, and drift correction behavior segment in different flight phases are recombined across phases to identify implicit behavior migration paths formed by low-altitude aircraft during obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering, and narrow area traversal. Based on these implicit behavior migration paths, a latent cause model of low-altitude aircraft flight behavior is constructed.

2. The method for modeling the flight behavior of low-altitude aircraft based on time-series decoupling of avionics data according to claim 1, characterized in that, S1 includes: S11: Acquire multi-source avionics data output from the flight control system, inertial navigation system, satellite navigation system, power system, and environmental perception system of the low-altitude aircraft, and perform timestamp correction, sampling frequency unification, and abnormal missing measurement supplementation according to a unified flight time axis to form a multi-source avionics synchronization sequence. The multi-source avionics synchronization sequence includes flight control sequence, attitude change sequence, navigation and positioning sequence, power response sequence, and environmental perception sequence. S12, based on the multi-source avionics synchronization sequence, extract active control features, environmental disturbance features and navigation drift features respectively, and calculate the contribution weight of each type of driving source to the change of flight state within the same time window. When at least two types of contribution weights exceed the corresponding judgment threshold at the same time, mark the time window as a composite action state. S13, for the composite action state, compare the sequential triggering relationship and contribution direction of different driving sources in attitude adjustment, heading deviation, altitude change and dynamic response, identify mutual coverage, mutual cancellation and pseudo-continuous transmission relationship, and construct a behavior coupling occlusion map with driving source as node and occlusion relationship as edge. The behavior coupling occlusion map is used to output the overlapping occupancy relationship, the actual triggering order and the occluded behavior segment between each driving source in the composite action state.

3. The method for modeling the flight behavior of low-altitude aircraft based on time-series decoupling of avionics data according to claim 2, characterized in that, S12 includes: S121, Based on the multi-source avionics synchronization sequence, a continuous time window is constructed according to a preset time length and sliding step size, and local changes in the flight control sequence, attitude change sequence, navigation and positioning sequence, power response sequence and environmental perception sequence are extracted in each time window to form a window state change set; S122, within each time window, active control features are extracted based on flight control sequence and dynamic response sequence, environmental disturbance features are extracted based on environmental perception sequence and attitude change sequence, and navigation drift features are extracted based on the deviation between navigation positioning sequence and inertial navigation attitude change sequence, generating active control contribution, environmental disturbance contribution and navigation drift contribution respectively. S123, normalize the contribution of active manipulation, environmental disturbance and navigation drift to obtain the contribution weight of each type of driving source in the same time window. When at least two types of contribution weights in the same time window exceed their corresponding judgment thresholds, the time window is marked as a composite action state; otherwise, it is marked as a single driving state.

4. The method for modeling the flight behavior of low-altitude aircraft based on time-series decoupling of avionics data according to claim 3, characterized in that, S13 includes: S131, for the time window marked as a composite action state, extract the first significant change moment in attitude adjustment, heading deviation, altitude change and dynamic response of the three types of driving sources: active control, environmental disturbance and navigation drift, and determine the sequential triggering relationship between different driving sources according to the first significant change moment; S132, based on the first significant change time, contribution direction, and change amplitude, determine the occlusion relationship between any two types of driving sources, specifically including: When the time difference between the first significant changes of two types of driving sources in the same flight state quantity is less than or equal to the trigger synchronization threshold, and the contribution directions are the same, it is determined that there is a mutual overlap relationship between the two types of driving sources. When the two types of driving sources contribute in opposite directions to the same flight state quantity, and the change amplitude between the two types of driving sources is greater than the corresponding cancellation determination threshold, it is determined that there is a mutual cancellation relationship between the two types of driving sources. When the first significant change of the latter driving source is later than the first significant change of the former driving source, and the direction of the state change of the latter driving source before its own triggering is consistent with the direction of contribution of the former driving source, and at the same time the change in control command, change in environmental perception, or change in navigation drift corresponding to the latter driving source does not reach its significant change threshold, it is determined that there is a pseudo-continuous transmission relationship between the two types of driving sources. S133, based on the determination results of mutual coverage relationship, mutual cancellation relationship and pseudo-continuous transmission relationship, takes active manipulation, environmental disturbance and navigation drift as nodes, and the driving source pair that satisfies any kind of occlusion relationship as edge, and writes the relationship type, corresponding flight state quantity, triggering order, change amplitude and occluded segment into the edge attribute to generate a behavior coupling occlusion map.

5. The method for modeling the flight behavior of low-altitude aircraft based on time-series decoupling of avionics data according to claim 4, characterized in that, S2 includes: S21, based on the behavior coupling occlusion graph, read the edge attributes corresponding to each occlusion relationship edge, and combine them with the first... The contribution weights of each driving source to attitude adjustment, path deviation, altitude change and heading correction within each time window are calculated. The behavioral occupancy weight of each driving source on the corresponding flight state quantity is calculated. The driving source with the largest behavioral occupancy weight is determined as the dominant driving source of the flight state quantity in the current time window. The time windows in which the same dominant driving source is continuously maintained are merged into the dominant contribution interval. S22, based on the behavior occupancy right and the pseudo-continuous transmission relationship in the behavior coupling occupancy map, determine whether there is a behavior continuity relationship formed by the pseudo-inheritance of environmental compensation actions between adjacent time windows, and generate the cut-off flight behavior segment. S23, for the severed flight behavior segment, the preceding driving source with mutual coverage, mutual cancellation or pseudo-continuous transmission relationship with the segment is retrieved in reverse along the occlusion relationship edge in the behavior coupling occlusion map, and the responsibility backtracking score is calculated according to the triggering time, behavior occupancy right and occlusion relationship type of the preceding driving source. The behavior segment corresponding to the preceding driving source with the highest responsibility backtracking score is determined as the original triggering behavior hidden after the compensation action, and the behavior responsibility decoupling chain is generated in the order of original triggering behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment and drift correction behavior segment.

6. The method for modeling the flight behavior of low-altitude aircraft based on time-series decoupling of avionics data according to claim 5, characterized in that, S22 includes: S221, based on the first The first time window and the first The behavior occupancy rights of the first time window are determined, and the first action within the adjacent time window is determined respectively. The dominant driving source of the flight-like state quantity is identified, and the direction of state change, changes in the dominant driving source, and changes in the behavior occupancy right are compared between adjacent time windows to form the behavior inheritance result between adjacent windows. S222, the first A flight state change within a time window that is dominated by environmental disturbances and whose behavior occupies the threshold for compensation takeover is defined as a compensation takeover behavior. The compensation takeover action is formed within a time window, and the direction of its state change is the same as that of the first... The first time window is consistent, and at the same time the second When the change in the environmental trigger input corresponding to a time window does not reach the significant change threshold, and there is a pseudo-continuous transmission relationship edge pointing to the source of environmental disturbance in the behavior coupling occlusion graph, it is determined that there is row continuity between adjacent time windows formed by pseudo-inheritance of environmental compensation actions. S223, when it is determined that there is a pseudo-inheritance relationship of environmental compensation actions between adjacent time windows, the first... The starting time of each time window is used as the cutting boundary. The original continuous flight behavior is split into pre-cut behavior segments and post-cut behavior segments. The time windows that continuously satisfy the pseudo-inheritance judgment condition are merged to generate the cut flight behavior segments.

7. A method for modeling the flight behavior of low-altitude aircraft based on time-series decoupling of avionics data as described in claim 6, characterized in that, S23 includes: S231, for the severed flight behavior segment, determine its corresponding severed driving source. Cut-off start time and corresponding flight state quantities and with the cut-off drive source With the endpoint as the target, the reverse search is performed along the occlusion relationship edges in the coupling occlusion graph to filter out the preceding driving sources that were triggered before the cutoff start time and have mutual coverage, mutual cancellation or pseudo-continuous transmission relationship with the cutoff driving source, forming a candidate preceding driving source set. S232, for each candidate preceding driving source in the candidate preceding driving source set, read its first significant change time, the behavior occupancy right in the corresponding time window, and the occlusion relationship type with the cut-off driving source, and calculate the responsibility backtracking score according to the time proximity, behavior occupancy right size and occlusion relationship correction coefficient. S233, the candidate preceding driving source with the highest responsibility backtracking score is determined as the original trigger driving source hidden after the compensation action, and the behavior segment corresponding to the original trigger driving source is extracted as the original trigger behavior segment. According to the occlusion relationship edge between the original trigger driving source and the cut-off driving source in the behavior coupling occlusion map, the disturbance coverage behavior segment, the compensation takeover behavior segment, and the drift correction behavior segment are extracted in sequence, and a behavior responsibility decoupling chain is generated according to the triggering time sequence and responsibility transmission sequence.

8. A method for modeling the flight behavior of low-altitude aircraft based on time-series decoupling of avionics data according to claim 7, characterized in that, S3 includes: S31, based on the behavior responsibility decoupling chain, the original triggering behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment and drift correction behavior segment are grouped according to the flight stage, and the behavior segments with the same original triggering drive source, the same masking relationship type or the same state change direction in adjacent flight stages are connected across stages to generate a cross-stage behavior recombination sequence. S32, Based on the cross-stage behavior recombination sequence, calculate the behavior migration matching degree of the low-altitude aircraft in obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering and narrow area passage scenarios respectively. When the behavior migration matching degree of the scenario reaches the migration judgment threshold of the scenario, the corresponding cross-stage behavior recombination sequence is identified as the implicit behavior migration path in the scenario. S33. Based on the implicit behavior migration path, the original trigger behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment and drift correction behavior segment are mapped as latent cause nodes respectively, and the cross-stage connection relationship, occlusion relationship type and behavior migration matching result are mapped as latent cause edges to construct a latent cause model of low-altitude aircraft flight behavior.

9. A method for modeling the flight behavior of low-altitude aircraft based on temporal decoupling of avionics data according to claim 8, characterized in that, S32 includes: S321. Based on the cross-stage behavior reorganization sequence, extract the behavior type, original triggering source, occlusion relationship type, state change direction, compensation takeover mark, drift correction mark and flight stage sequence of each behavior segment, and map them into scene migration feature vectors. Among them, obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering and narrow area passage are respectively configured with corresponding standard migration feature templates. S322, the scene migration feature vector is matched with the standard migration feature templates corresponding to obstacle avoidance, wind resistance correction, instability recovery, low-altitude hovering and narrow area passage, respectively, and the behavior migration matching degree under each scenario is calculated. S323, compare the behavior transfer matching degree in each scenario with the corresponding transfer judgment threshold, when the first... The cross-stage behavioral recombination sequence in the first When the behavior transfer matching degree in a low-altitude flight scenario reaches the corresponding transfer judgment threshold, the cross-stage behavior recombination sequence is marked as the first... The implicit behavior transfer path in low-altitude flight scenarios is recorded, along with its path label, start and end flight stages, constituent behavior segments, and matching scenarios.

10. A method for modeling the flight behavior of low-altitude aircraft based on temporal decoupling of avionics data according to claim 9, characterized in that, S33 includes: S331, based on the implicit behavior migration path, extract the original trigger behavior segment, disturbance coverage behavior segment, compensation takeover behavior segment and drift correction behavior segment in the path, and map them into original trigger hidden cause nodes, disturbance coverage hidden cause nodes, compensation takeover hidden cause nodes and drift correction hidden cause nodes according to the behavior segment type. At the same time, write the flight stage, corresponding driving source, state change direction, behavior occupancy right and start and end time of each hidden cause node into the node attributes to form a set of hidden cause nodes. S332, in the implicit behavior migration path, according to the temporal adjacency relationship, cross-stage connection relationship, occlusion relationship type and behavior migration matching result between behavior segments, any adjacent latent cause nodes are connected. When there is a cross-stage behavior connection edge, occlusion relationship edge or the same implicit behavior migration path belonging relationship between the behavior segments corresponding to two latent cause nodes, a latent cause edge is generated between the two latent cause nodes, and the connection source, occlusion relationship type, migration scene category and behavior migration matching degree are written into the edge attribute. S333, Based on the set of latent cause nodes and the set of latent cause edges, a latent cause model of low-altitude aircraft flight behavior is constructed, and the cross-stage connection relationship, occlusion relationship type and behavior migration matching result corresponding to each latent cause edge are written into the latent cause model of low-altitude aircraft flight behavior as a set of edge attributes.

Images