A design method of sequential multi-sensor track fusion effect simulation simulator

Abstract

The application relates to a design method of a sequential multi-sensor track fusion effect simulation simulator, and belongs to the field of simulation modeling and information fusion processing. In order to overcome the problem that in system simulation, a fusion simulator cannot flexibly simulate different fusion track qualities in the simulation of a sensor networking center, the application is improved on the basis of following a basic framework of a multi-sensor track fusion algorithm, and by innovatively adopting an error association probability table and a filter model correction function as external source parameters, an estimation process with different association and filtering performances which can be self-defined and configured is provided for users, different quality fusion tracks output by a sensor networking system are simulated, and therefore the flexibility of the fusion track simulation is improved, and higher freedom and authenticity are provided for system simulation.

Description

Technical Field

[0001] This invention belongs to the field of simulation modeling and information fusion processing, and specifically relates to the design method of a sequential multi-sensor trajectory fusion effect simulation simulator. Background Technology

[0002] In system simulation, multi-sensor network systems provide unified target trajectory information to various users through sensor trajectory fusion, which is a crucial process requiring simulation. Because many simulation behaviors of other models rely on the fused trajectory information provided by the network system, simulating the multi-sensor trajectory fusion effect is extremely important. Currently, there are generally two approaches to simulating fusion effects. One approach uses a specific multi-sensor trajectory fusion algorithm as the trajectory fusion model. Its advantage is that it uses a completely realistic algorithm, resulting in high model reliability. However, such models are not generally representative, and the quality of the fused trajectory they provide only represents the effect of that specific algorithm. Therefore, for system simulation, the degree of freedom of changing the fused trajectory quality cannot be changed. The other approach is a highly simplified information fusion model that does not consider the actual fusion process, only adding appropriate errors to the true value of the simulated target trajectory. This method can only simulate the accuracy factor in trajectory quality and cannot simulate complex situations such as mixed batches and discontinuous batches that occur in real-world scenarios. Furthermore, because it deviates from the basic principles of the algorithm, the simulation accuracy is not high. Therefore, there is currently a lack of a sequential multi-sensor trajectory fusion effect simulation simulator that can flexibly simulate fused trajectories of different qualities and realistically simulate the characteristics of fused trajectories from a mechanistic perspective.

[0003] This invention follows the basic framework of multi-sensor track fusion algorithms, including three processes: track initiation, association simulation, and filtering simulation. The association simulation process modifies the nearest neighbor association algorithm framework, innovatively introducing an error association probability table as an external configuration parameter. Combined with randomized simulation, it can simulate batch mixing or batch breaking phenomena caused by erroneous associations in fused tracks. Because the error association probability table can be flexibly configured manually, the degree of batch mixing or batch breaking can be controlled. The filtering simulation process modifies the Kalman rate wave framework, innovatively introducing a motion model correction function as an external configurable parameter. It can simulate excessive tracking errors, even filtering divergence and batch breaking phenomena, caused by target maneuvering. Because the motion model correction function can be flexibly defined manually, the quality of maneuvering target tracking can be controlled. Therefore, the multi-sensor track fusion effect simulator proposed in this invention can achieve high-fidelity simulation of fused tracks of different qualities. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] The technical problem to be solved by this invention is how to provide a design method for a sequential multi-sensor trajectory fusion effect simulation simulator, so as to solve the problem that in system simulation, there is a lack of flexible and configurable fusion simulators for simulating different fusion trajectory qualities for the sensor network center simulation.

[0006] (II) Technical Solution

[0007] To address the aforementioned technical problems, this invention proposes a design method for a sequential multi-sensor track fusion simulation simulator. This simulator includes a track association simulator, a track initiation module, and a track filtering simulator. The track simulator is started by sensor input track points. First, the track association simulator runs. If the input sensor track point is associated with an existing track, the track point and associated track are input into the track filtering simulator; otherwise, they are input into the track initiation module to generate a new fused track. The track filtering simulator updates the current target motion state and saves the update results to the fused track data list.

[0008] The trajectory association simulator performs the following steps:

[0009] Step S100: Input sensor j at t k The original track point at time r j _track(t k ), z is the original track point r j _track(t k The motion state vector of );

[0010] Step S101: Determine if the fused track list is empty. If it is empty, start the track module; otherwise, proceed to step S102.

[0011] Step S102: Select a track i from the track list, and estimate the track point position of track i at the previous moment. Forecast t k Predicted position of track point at time Where F is the motion state transition matrix, predicting t k The state covariance matrix at time t is Where P is the state covariance matrix and Q is the process noise covariance matrix. It is the estimated value of the state covariance matrix. It is the predicted value of the state covariance matrix;

[0012] The fused track list is f_track i (k);i=1,...,n;k=1,...,K i , where n is the number of tracks in the track list, and K if_track represents the number of points in track i; x represents the list of merged tracks. i The motion state vector of (k);

[0013] Step S103: Calculate the target association separation degree SDOA stands for Target Separation Degree, which defines the difficulty of calculating the association between two targets.

[0014] Step S104: Set up two error association probability tables, divided into two categories of errors. The first category is the probability table of points that should have been associated but were not, denoted as TAB1. The second category is the probability table of points that should not have been associated but were incorrectly associated, denoted as TAB2. The tables list the probabilities of the two types of error associations under different SDOA values. In the simulation, the ID of the target corresponding to the original trajectory point to be associated is z| target_ID The target ID corresponding to the last point of the merged track is f_track i (K i )| target_ID

[0015] Step S105: If f_track i (K i )| target_ID =z| target_ID If we query table TAB1, the probability of a join is 1-P. TAB1 (SDOA), otherwise check the TAB2 table; the association probability is P. TAB2 (SDOA)

[0016] Step S106: Based on the association probability in S105, generate random numbers according to the Bernoulli distribution. If the result is 1, correlate z with f_track. i (K i If the path is associated with the path, proceed to the path filtering simulator; otherwise, proceed to the path initiation module.

[0017] The trajectory filtering simulator performs the following steps:

[0018] Step S200: Input the estimated value of the fused track point from the previous time step for fused track i. Associated current moment sensor raw track point z k =r j _track(t k Simulation target t k Time-track truth value x k ;

[0019] Step S201: Calculate the predicted trajectory. Where G(·) is a function of the current state truth value and the predicted value;

[0020] Step S202: Calculate the state covariance prediction. The estimated value of the filter state covariance matrix at the previous time step is The process noise covariance matrix is ​​Q k-1 ;

[0021] Step S203: Calculate the gain matrix H is the observation matrix; the original track error covariance of the sensor is R. k ;

[0022] Step S204: Calculate the updated value of the waypoint position estimate Output the fused and filtered track points and store them in f_track. i (k);

[0023] Step S205: Update t k The estimated value of the state covariance matrix at time step is

[0024] The track initiation module performs the following steps:

[0025] Step S300: Determine whether the sensor track point is associated with an existing track. If yes, proceed to step S301; otherwise, proceed to step S302.

[0026] Step S301: No new track starts;

[0027] Step S302: The new sensor track point is used as the first point of the new track, and the track number is the historical track number plus 1.

[0028] (III) Beneficial Effects

[0029] This invention proposes a design method for a sequential multi-sensor trajectory fusion effect simulation simulator, belonging to the field of simulation modeling and information fusion processing. This invention aims to overcome the problem in system simulation where there is a lack of flexible and configurable fusion simulators for simulating different fused trajectory qualities at the sensor network center. Traditional sensor fusion trajectory simulations either employ specific sensor data fusion algorithms, representing only the fused trajectory quality provided by that specific algorithm, or they fail to consider the fundamental principles of the fusion algorithm, resulting in oversimplification and an inability to simulate phenomena such as batch mixing, batch breaking, and increased filtering errors due to target maneuvering that degrade fused trajectory quality. The method of this invention finds a compromise between these two approaches. While adhering to the basic framework of multi-sensor trajectory fusion algorithms, it improves upon them by innovatively using an error association probability table and a filtering model correction function as external parameters. This provides users with a customizable estimation process for different association and filtering performance, thereby simulating the sensor network system providing fused trajectories of varying quality to information users. This improves the flexibility of fusion trajectory simulation and provides greater freedom and realism for system simulation. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the components of the trajectory fusion effect simulation simulator of the present invention;

[0031] Figure 2 The overall process of running the simulation simulator for track fusion effect;

[0032] Figure 3 A comparison of the basic process of the track association algorithm and the basic process of the track association simulator;

[0033] Figure 4 A comparison of the general workflow of Karman filter algorithms and the workflow of a fusion filter simulator;

[0034] Figure 5 A schematic diagram simulating the optimal quality of the fused trajectory;

[0035] Figure 6 This is a schematic diagram to simulate the effect of a fused track that is perfectly correlated but has slightly poor maneuver tracking capability.

[0036] Figure 7 A schematic diagram to simulate the effect of a fused track that is perfectly correlated but has poor maneuver tracking capability;

[0037] Figure 8 This is a schematic diagram simulating the effect of a fused trajectory with correlation errors but optimal maneuver tracking capability;

[0038] Figure 9 This is a schematic diagram simulating a fused trajectory with poor maneuver tracking capabilities due to correlation errors;

[0039] Figure 10 This is a schematic diagram simulating a fused track with poor maneuver tracking capabilities due to correlation errors. Detailed Implementation

[0040] To make the objectives, contents, and advantages of the present invention clearer, the specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples.

[0041] The technical problem to be solved by this invention is how to design a sequential multi-sensor trajectory fusion effect simulation simulator to overcome the problem that in system simulation, there is a lack of flexible and configurable fusion simulators for simulating different fusion trajectory qualities for the sensor network center simulation.

[0042] This invention relates to a design method for a simulation simulator of multi-sensor trajectory fusion effects, belonging to the field of simulation modeling and information fusion processing. This invention aims to overcome the problem in system simulation where there is a lack of flexible and configurable fusion simulators for simulating different fusion trajectory qualities at the sensor network center. Traditional sensor fusion trajectory simulations either employ specific sensor data fusion algorithms, representing only the fusion trajectory quality provided by that specific algorithm, or they fail to consider the fundamental principles of the fusion algorithm, resulting in oversimplification and an inability to simulate phenomena such as batch mixing, batch breaking, and increased filtering errors due to target maneuvering that degrade fusion trajectory quality. The method of this invention strikes a balance between these two approaches. While adhering to the basic framework of multi-sensor trajectory fusion algorithms, it improves upon them by innovatively using an error association probability table and a filtering model correction function as external parameters. This allows users to customize the association and filtering estimation processes with different performance levels, thereby simulating the sensor network system providing fusion trajectories of varying quality to information users. This improves the flexibility of fusion trajectory simulation and provides greater freedom and realism for system simulation.

[0043] The basic architecture of the sequential multi-sensor trajectory fusion simulation simulator of this invention is as follows: Figure 1 As shown, it includes three sub-modules: track association simulator, track initiation module, and track filtering simulator.

[0044] The operation flow of the trajectory fusion effect simulation simulator is as follows: Figure 2 As shown. The trajectory simulator is started by inputting trajectory points from the sensors. First, the trajectory association simulator runs. If the input sensor trajectory point is associated with an existing trajectory, the trajectory point and associated trajectory are input into the trajectory filtering simulator; otherwise, they are input into the trajectory initiation module to generate a new fused trajectory. The trajectory filtering simulator updates the current target motion state and saves the update results to the fused trajectory data list.

[0045] 1. Track Association Simulator

[0046] Step S100: Input sensor j at t k The original track point at time r j _track(t k ), z is the original track point r j _track(t k The motion state vector of );

[0047] Step S101: Determine if the fused track list is empty. If it is empty, start the track module; otherwise, proceed to step S102.

[0048] Step S102: Select a track i from the track list, and estimate the track point position of track i at the previous moment. Forecast t k Predicted position of track point at time Where F is the motion state transition matrix, P is the state covariance matrix, and Q is the process noise covariance matrix. It is the estimated value of the state covariance matrix. It is the predicted value of the state covariance matrix; the above formula is the basic formula in Kalman filtering;

[0049] The fused track list is f_track i (k);i=1,...,n;k=1,...,K i , where n is the number of tracks in the track list, and K i f_track represents the number of points in track i; x represents the list of merged tracks. i The motion state vector of (k);

[0050] Step S103: Calculate the target association separation degree SDOA (Target Association Separation Degree) is used to define the difficulty of calculating the association between two targets; see Context ComplexityMetric of Association Problem for Target Tracking[C] / / 2019IEEE International Conference on Signal, Information and Data Processing(ICSIDP).IEEE,2019;

[0051] S103.1: Input sensor raw track error covariance σ lat σ is the standard deviation of latitude of the original track. long This is the standard deviation of longitude of the original track; it can be set based on prior information about the accuracy of the original track.

[0052] S103.2: The probability density function of the original waypoint location to be associated is:

[0053]

[0054] Where z represents the observation value of the original track point to be associated;

[0055] The probability density function of the extrapolated predicted position of the last estimated point in the track is:

[0056]

[0057] S103.3: Let M = (N1 + N2) / 2 be the average of the probability density function, and let N1, N2, and M be represented by N1(x), N2(x), and M(x), respectively;

[0058] Calculate the KL divergence between N1 and M, and N2 and M respectively.

[0059] Where, i = 1, 2

[0060] S103.4: Calculate the JS divergence of N1 and N2, i.e.

[0061] Step S104: Set up two error association probability tables, divided into two categories: the first category is the probability table of points that should have been associated but were not (denoted as TAB1), and the second category is the probability table of points that should not have been associated but were incorrectly associated (denoted as TAB2). The tables list the probabilities of performing the two types of error associations under different SDOA values. These two tables can be configured in advance as configuration tables before simulator execution. In the simulation, the ID of the target corresponding to the original trajectory point to be associated is z| target_ID The target ID corresponding to the last point of the merged track is f_track i (K i )| target_ID

[0062] Step S105: If f_track i (K i )| target_ID =z| target_ID If we query table TAB1, the probability of a join is 1-P. TAB1 (SDOA), otherwise check the TAB2 table; the association probability is P. TAB2 (SDOA)

[0063] Step S106: Based on the association probability in S105, generate random numbers according to the Bernoulli distribution. If the result is 1, correlate z with f_track. i (K iIf the path is associated with the path, proceed to the path filtering simulator; otherwise, proceed to the path initiation module.

[0064] 2. Track Filtering Simulator

[0065] Step S200: Input the estimated value of the fused track point from the previous time step for fused track i. Associated current moment sensor raw track point z k =r j _track(t k Simulation target t k Time-track truth value x k .

[0066] Step S201: Calculate the predicted trajectory. Wherein, G(·) is a function of the current state true value and the predicted value, which can be flexibly defined in the trajectory filtering simulator to simulate the filtered motion model from the linear motion model to the case of no model error;

[0067] Step S202: Calculate the state covariance prediction. The estimated value of the filter state covariance matrix at the previous time step is The process noise covariance matrix is ​​Q k-1 .

[0068] Step S203: Calculate the gain matrix H is the observation matrix; the original track error covariance of the sensor is R. k ,

[0069] Step S204: Calculate the updated value of the waypoint position estimate Output the fused and filtered track points and store them in f_track. i (k)

[0070] Step S205: Calculate the estimated state covariance matrix and update the prediction.

[0071] 3. Track Initiation Module

[0072] Step S300: Determine whether the sensor track point is associated with an existing track. If yes, proceed to step S301; otherwise, proceed to step S302.

[0073] Step S301: No new track starts;

[0074] Step S302: The new sensor track point is used as the first point of the new track, and the track number is the historical track number plus 1.

[0075] Example 1:

[0076] (I) Symbol Conventions

[0077] (1) Merge track list f_track i (k), i = 1, ..., n; k = 1, ..., K i The list contains n merged tracks, and the track numbered i has K tracks. i One waypoint.

[0078] (2) The k-th track point in the track with batch number i, f_track i (k)=(time track_ID radar_IDlong lat alt v_long v_lat target_ID)

[0079] (3)x i,k for f_track i The motion state vector in (k), i.e., x i,k =(long lat v_long v_lat) T

[0080] (4) Sensor j at t k The track points reported regularly, r j _track(t k )=(time track_IDradar_ID long lat alt v_long v_lat target_ID)

[0081] (5)z k For r j _track(t k The motion state vector in ) is z k =(long lat v_long v_lat) T

[0082] (6)f_track i (k) and r j _track(t k )middle

[0083] time — the time of the waypoint;

[0084] track_ID — Batch number of fused track / single sensor track;

[0085] radar_ID — The associated sensor number / sensor number of the fused track;

[0086] long — longitude of fused track / single sensor track;

[0087] lat — Latitude of fused track / single sensor track;

[0088] alt—High Longitude of Fusion Track / Single Sensor Track;

[0089] v_long — horizontal longitude velocity of fused track / single sensor track;

[0090] v_lat — Horizontal latitude velocity of fused track / single sensor track;

[0091] target_ID — The actual ID of the target

[0092] (II) Explanation of Basic Principles

[0093] 1. Simulator Components and Simulation Process

[0094] The basic architecture of the trajectory fusion effect simulation simulator is as follows: Figure 1 As shown, the abstract includes three sub-modules: track association simulator, track initiation module, and track filtering simulator.

[0095] The operation flow of the trajectory fusion effect simulation simulator is as follows: Figure 2 As shown. The trajectory simulator is started by inputting trajectory points from the sensors. First, the trajectory association simulator runs. If the input sensor trajectory point is associated with an existing trajectory, the trajectory point and associated trajectory are input into the trajectory filtering simulator; otherwise, they are input into the trajectory initiation module to generate a new fused trajectory. The trajectory filtering simulator updates the current target motion state and saves the update results to the fused trajectory data list.

[0096] 2. Basic principles of associated simulators

[0097] The association algorithm is the core and most challenging process in the fusion algorithm. In track fusion, incorrect association can lead to target batch breaks, new batch creation, and mixed batches. Track association simulation is not a real association algorithm, but rather, within the framework of basic track association principles, it calculates the SDOA between targets and uses an error association rate configuration table as an external parameter to generate incorrect associations or failures to associate targets with a certain probability. This causes batch breaks, new batch creation, and mixed batches to occur during the simulator's association algorithm execution. The advantage of the method in this invention is that the association simulator has high flexibility. It can simulate association algorithms with different association effects by flexibly setting the error association probability table, i.e., simulating the effects of association algorithms with different performance indicators. Furthermore, the error association probability table data can be provided using empirical data obtained from experiments with real association algorithms, thereby simulating the basic effect of a specific association algorithm. At the same time, this simulation method is designed based on the basic logical architecture of the association algorithm, thus possessing high simulation reliability. Figure 3 This paper compares the basic process of the track association algorithm and the basic process of the track association simulator.

[0098] 3. Basic Principles of a Filter Simulator

[0099] Kalman filtering algorithms are generally used to estimate the motion state of a target by introducing a motion model to suppress the influence of observation errors on the state estimation. Figure 4 The left figure shows the general flow of Kalman filter algorithms, which can describe the general principles of algorithms under various Kalman filter frameworks such as KF, EKF, UKF, and IMM. The effectiveness of Kalman filtering depends on the accuracy of the motion equations and measurement equations in representing the target's motion state and observations. For this track fusion system, the track observation input is latitude and longitude, and the estimation is also latitude and longitude. Therefore, the measurement equations are linear equations without truncation errors. The filtering error mainly comes from the error in representing nonlinear motions such as maneuvers by the motion equations. The larger the representation error, the more delayed the state update's reaction time to maneuvers, the larger the estimation error, and even the more likely it is to cause filter divergence.

[0100] Figure 4 The right figure shows the flowchart of the fusion filter simulator. The calculation process of this simulator is basically the same as that of KF, but the difference lies in the track prediction stage, where KF uses...

[0101] If it is a six-dimensional filter, that is, the latitude, longitude, and altitude are each adjusted for their respective velocities, then...

[0102] If it is a four-dimensional filter, that is, latitude and longitude plus their respective velocities, then Δt is the time interval.

[0103] That is, retaining the first-order term of the equation of motion, while in the simulator... Additional correction items Where x k Let G(·) be the true state of the trajectory at time k, and let G(·) be a function of the current true state and the predicted value. G(·) can be flexibly defined in the simulator to simulate a filtered motion model ranging from a linear motion model to one with no model error. Therefore, the timeliness and accuracy of the filtering algorithm's response to different maneuvers can be simulated by designing and adjusting G(·).

[0104] Several ways to define the model correction function G(·) are given below, but it is not limited to these. Users can reasonably customize G(·) according to their needs and subsequent research.

[0105] (1) G(x1,x2)=α(x1-x2), where the parameter setting is 0≤α≤1. When α=0, it can simulate the effect of KF filter. When α=1, it can simulate an ideal motion model with no truncation error and completely consistent with the real motion. When 0<α<1, it is a filter between a linear filter and an ideal filter, which can linearly correct the truncation error.

[0106] (2) in This represents the motion equation in a user-defined filter. In this case, the simulator simulates the filtering effect under the user-defined motion equation.

[0107] This invention proposes a design method for a sequential multi-sensor trajectory fusion effect simulation simulator, which can be used in the simulation field to simulate multi-sensor fusion systems. By innovatively employing an error association probability table and a filter model correction function as external parameters, it allows for flexible user-defined configuration of association and filter estimation processes with different performance levels, thereby simulating fusion systems at different levels and enhancing the openness of multi-sensor fusion effect simulation. Traditional sensor fusion trajectory simulations either employ specific sensor data fusion algorithms, representing only the capabilities of a particular algorithm, or fail to consider the fundamental principles of fusion algorithms, resulting in insufficient simulation realism and an inability to simulate mixed-batch associations or situations where target maneuvers lead to increased filtering errors. The method of this invention strikes a balance between these two approaches, reflecting the data fusion process while also giving the simulation strong generalization capabilities.

[0108] Example 2

[0109] This embodiment specifically describes a method proposed in this invention for simulating fused tracks of different qualities by setting different simulation parameters.

[0110] The simulation scenario includes four aircraft targets, with the flight start time being simulation time 0. The flight path parameters are shown in Table 1-4.

[0111] Table 1 Flight parameters of target 1

[0112] Waypoints latitude longitude high speed 1 39.5 115 10000 200 2 39.55 115.05 10000 200 3 39.5 116.05 10000 200 4 39.55 117 10000 200

[0113] Table 2 Flight parameters of target 2

[0114] Waypoints latitude longitude high speed 1 39.52 115 10000 200 2 39.57 115.05 10000 200 3 39.501 116.05 10000 200 4 39.551 117 10000 200

[0115] Table 3 Flight parameters of target 3

[0116] Waypoints latitude longitude high speed 1 39.54 115 10000 200 2 39.59 115.05 10000 200 3 39.502 116.05 10000 200 4 39.552 117 10000 200

[0117] Table 4 Flight parameters of target 4

[0118] Waypoints latitude longitude high speed 1 39.5 115 10000 200 2 39.59 117 10000 200

[0119] The sensor performs a circular scan with a scan cycle of 10 seconds. Its elevation angle range is 0-90 degrees, and its detection radius is 40 kilometers (50% detection probability). The azimuth measurement error is 0.2 degrees (1 sigma), the elevation angle measurement error is 0.2 degrees (1 sigma), and the distance measurement error is 50 meters. The sensor deployment parameters are shown in Table 5.

[0120] Table 5 Sensor Deployment Location Parameters

[0121] serial number latitude longitude serial number latitude longitude 1 39.5 115 7 39.56 115 2 39.5 116 8 39.56 116 3 39.5 117 9 39.56 117 4 39.53 115 10 39.59 115 5 39.53 116 11 39.59 116 6 39.53 117 12 39.59 117

[0122] The above parameter configuration generates raw tracks from 12 sensors for 4 targets, which serve as input to a sequential multi-sensor track fusion simulation simulator. By configuring the simulator parameters, fused tracks of varying quality can be simulated. With the correlation gate parameter set to 5000 meters:

[0123] (1) Simulate the optimal quality fusion trajectory

[0124] Set the filter configuration parameter α = 1, and the error correlation rate to 0%. The simulated fused trajectory is as follows: Figure 5 As shown in the figure. The points in the figure are the original track points reported by the sensors, and the four lines represent the fused track. The fused track is continuous, without any gaps or mixing, and has excellent tracking capability during turning maneuvers.

[0125] (2) Simulated fused tracks with perfect correlation but poor maneuver tracking capability.

[0126] With the filter parameter α = 0.1 and the error correlation rate set to 0%, the simulated fused trajectory is as follows: Figure 6 As shown in the figure, the points are the original track points reported by the sensors, and the four lines represent the fused track. The fused track is continuous without any gaps or batches, but the tracking error increases during the turning maneuver, simulating the "bulge" phenomenon in maneuvering and turning tracking.

[0127] (3) Simulates a fused track that is perfectly correlated but has poor maneuver tracking capability.

[0128] With the filter parameter α = 0 and the error correlation rate set to 0%, the simulated fused trajectory is as follows: Figure 7 As shown in the figure. The points in the figure are the original track points reported by the sensors, and the colored curves represent the fused track. During the target turning maneuver, the track tracking "disconnected," and after the maneuver, the fusion system simulated the phenomenon of "re-starting" the track.

[0129] (4) The simulation has correlation errors, but the fused track has the best maneuver tracking capability.

[0130] With the filter parameter α = 1 and the error correlation rate set to 10%, the simulated fused trajectory is as follows: Figure 8As shown, the phenomenon of "mixed batches" of flight paths was simulated when paths intersected and approached each other.

[0131] (5) The simulation of the fused track has correlation errors and poor maneuver tracking capability.

[0132] With the filter parameter α = 0.1 and the error correlation rate set to 10%, the simulated fused trajectory is as follows: Figure 9 As shown in the figure, the simulation also demonstrated phenomena such as "mixed batches," "discontinuous batches," and excessively large machine errors that occurred in the fusion system.

[0133] (6) Poor simulation correlation and maneuver tracking of the fused track

[0134] With the filter parameter α = 0 and the error correlation rate set to 10%, the simulated fused trajectory is as follows: Figure 10 As shown, a scenario is simulated where the target tracking capability of the fusion system is very poor. During the fusion tracking process, the tracking error of maneuvering targets increases, or even the targets are lost, and phenomena such as "mixed batches," "broken batches," and "incorrect batches" of tracks occur.

[0135] This invention proposes a design method for a sequential multi-sensor trajectory fusion effect simulation simulator, which can be used in the field of military simulation to simulate multi-sensor fusion systems. By innovatively employing an error association probability table and a filter model correction function as external parameters, it allows for flexible user-defined configuration of association and filter estimation processes with different performance levels, thereby simulating fusion systems at different levels and enhancing the openness of multi-sensor fusion effect simulation. Traditional sensor fusion trajectory simulations either employ specific sensor data fusion algorithms, representing only the capabilities of a particular algorithm, or fail to consider the fundamental principles of fusion algorithms, resulting in insufficient simulation realism and an inability to simulate mixed-batch associations or situations where target maneuvers lead to increased filtering errors. The method of this invention strikes a balance between these two approaches, reflecting the data fusion process while also giving the simulation strong generalization capabilities.

[0136] 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 technical principles 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 design method for a sequential multi-sensor trajectory fusion simulation simulator, characterized in that, The simulation simulator includes a track association simulator, a track initiation module, and a track filtering simulator. The track simulator is started by inputting track points from sensors. First, the track association simulator runs. If the input sensor track point is associated with an existing track, the track point and associated track are input into the track filtering simulator; otherwise, they are input into the track initiation module to generate a new fused track. The track filtering simulator updates the current target motion state and saves the update results to the fused track data list. The trajectory association simulator performs the following steps: Step S100: Input sensor j in The original track point at that time was z is the original waypoint The motion state vector; Step S101: Determine if the fused track list is empty. If it is empty, start the track module; otherwise, proceed to step S102. Step S102: Select a track i from the track list, and estimate the track point position of track i at the previous moment. ,forecast Predicted position of track point at time Where F is the motion state transition matrix, and the prediction... The state covariance matrix at time t is Where P is the state covariance matrix and Q is the process noise covariance matrix. It is the estimated value of the state covariance matrix. It is the predicted value of the state covariance matrix; Among them, the fused track list is , where n is the number of tracks in the track list, K i x is the number of points in track i; x is the list of merged tracks. The motion state vector; Step S103: Calculate the target association separation degree ;SDOA stands for Target Associativity Separation Degree, which defines the difficulty of calculating the association between two targets; Step S104: Set up two error association probability tables, divided into two categories of errors. The first category is the probability table of points that should have been associated but were not, denoted as TAB1. The second category is the probability table of points that should not have been associated but were incorrectly associated, denoted as TAB2. The tables list the probabilities of the two types of error associations under different SDOA values. In the simulation, the ID of the target corresponding to the original trajectory point to be associated is... The target ID corresponding to the last point of the merged track is Step S105: If If we query table TAB1, the probability of a join is... Otherwise, query the TAB2 table; the probability of a join is... ; Step S106: Based on the correlation probability in S105, generate random numbers according to the Bernoulli distribution. If the result is 1, then... and If the connection is established, proceed to the trajectory filtering simulator; otherwise, proceed to the trajectory initiation module. The trajectory filtering simulator performs the following steps: Step S200: Input the estimated value of the fused track point from the previous time step for fused track i. The associated current moment sensor original track point Simulation target Time Track True Value ; Step S201: Calculate the predicted trajectory. ;in, It is a function of the current state's true value and the predicted value; Step S202: Calculate the state covariance prediction. The estimated value of the filter state covariance matrix at the previous time step is The process noise covariance matrix is ; Step S203: Calculate the gain matrix , Here is the observation matrix; the original track error covariance of the sensor is... ; Step S204: Calculate the updated value of the waypoint position estimate Output fused and filtered waypoints and store them. ; Step S205: Update The estimated value of the state covariance matrix at time step is ; The track initiation module performs the following steps: Step S300: Determine whether the sensor track point is associated with an existing track. If yes, proceed to step S301; otherwise, proceed to step S302. Step S301: No new track starts; Step S302: The new sensor track point is used as the first point of the new track, and the track number is the historical track number plus 1; in, Step S103 specifically includes the following steps: S103.1: Input sensor raw track error covariance , The standard deviation of latitude of the original flight path. The longitude standard deviation of the original track is set based on prior information about the accuracy of the original track. S103.2: The probability density function of the original waypoint location to be associated is: The probability density function of the extrapolated predicted position of the last estimated point in the track is: ; S103.3: Order Let N1, N2, and M be the average of the probability density functions, and let N1(x), N2(x), and M(x) be denoted by N1(x), N2(x), and M(x), respectively. Calculate separately and , and KL dispersion between, i.e. ,in ; S103.4: Calculation and The JS divergence, i.e. .

2. The design method of the sequential multi-sensor trajectory fusion simulation simulator as described in claim 1, characterized in that, The error association probability table is configured in advance as a configuration table before the simulator is executed.

3. The design method of the sequential multi-sensor trajectory fusion simulation simulator as described in claim 1, characterized in that, In the aforementioned trajectory filtering simulator, for a six-dimensional filter, i.e., latitude, longitude, altitude, and their respective velocities, then... , , For time intervals.

4. The design method of the sequential multi-sensor trajectory fusion simulation simulator as described in claim 1, characterized in that, In the aforementioned trajectory filtering simulator, for a four-dimensional filter, i.e., latitude, longitude, and latitude plus their respective velocities, then... , , For time intervals.

5. The design method of the sequential multi-sensor trajectory fusion simulation simulator as described in any one of claims 1-4, characterized in that, function for ,in The settings parameters.

6. The design method of the sequential multi-sensor trajectory fusion simulation simulator as described in claim 5, characterized in that, At that time, the trajectory filtering simulator simulates the effect of the KF filter.

7. The design method of the sequential multi-sensor trajectory fusion simulation simulator as described in claim 5, characterized in that, At that time, the trajectory filtering simulator simulates an ideal motion model with no truncation error and is completely consistent with the real motion.

8. The design method of the sequential multi-sensor trajectory fusion simulation simulator as described in claim 5, characterized in that, At that time, the trajectory filtering simulator simulates a filter that is between a linear filter and an ideal filter, with linear correction for truncation error.

9. The design method of the sequential multi-sensor trajectory fusion simulation simulator as described in any one of claims 1-4, characterized in that, function for ,in This represents the motion equation in a user-defined filter. In this case, the trajectory filtering simulates the filtering effect under the user-defined motion equation.

Images