A method for generating test cases for reliability testing of a UAV ground dynamic target identification and tracking scene

Abstract

This invention addresses unmanned aerial vehicles (UAVs) and proposes a method for generating reliability test cases for UAV ground dynamic target recognition and tracking scenarios. It comprises five steps: (1) analyzing and determining the static reliability influencing factors in UAV ground dynamic target recognition and tracking scenarios; (2) analyzing the dynamic reliability influencing factors in UAV ground dynamic target recognition and tracking scenarios; (3) analyzing and determining the occlusion impact caused by building influence factors in UAV ground dynamic target recognition and tracking scenarios; (4) collecting data, performing a goodness-of-fit test on the distribution function, and selecting the optimal distribution of parameters affecting UAV ground dynamic target recognition and tracking; and (5) performing importance sampling to generate a UAV target recognition and tracking test environment and determine test cases.

Description

Technical Field

[0001] This invention addresses the needs of unmanned aerial vehicles (UAVs) and proposes a method for generating reliability test cases for UAV ground dynamic target recognition and tracking scenarios. This method targets the typical task of UAV target vehicle recognition and tracking. Within a road network of a certain width, it analyzes static and dynamic reliability influencing factors in the task environment, identifies the impact of buildings on UAV recognition and tracking, determines key parameters and their distribution types, and finally performs importance sampling to generate important scenario test cases for UAV ground dynamic target recognition and tracking reliability. This provides support for reliability assessment of UAV ground dynamic target recognition and tracking. This invention belongs to the field of reliability engineering. Background Technology

[0002] Because of their advantages such as low cost, high maneuverability, ease of use, and high security, drones are used in various important missions and activities, including target identification and tracking, aerial photography, and security monitoring. At the same time, the reliability of drones is becoming increasingly prominent, requiring designers to accurately assess their reliability levels under typical mission scenarios during the drone development process, thereby providing a basis for optimizing drone reliability design. During the development phase, drone reliability assessment data primarily comes from testing; ensuring comprehensive test coverage directly affects the accuracy of the assessment results.

[0003] When unmanned aerial vehicles (UAVs) perform target vehicle identification and tracking tasks in urban roads, they are often affected by static factors such as weather, roads, and buildings, as well as dynamic factors such as the movement of other vehicles and target vehicles, and the initial altitude and speed of the UAV. However, using simulation methods to test the reliability of UAV target vehicle identification and tracking is costly, inefficient, and has poor repeatability. Therefore, this invention proposes an importance-based test case generation method. By analyzing the static and dynamic influencing factors in the UAV target vehicle identification and tracking scenario, and constructing a suitable distribution model for these factors, importance sampling is performed to generate reliability test cases, providing input for accurate reliability assessment of UAV ground dynamic target identification and tracking. Summary of the Invention

[0004] This invention provides a method for generating reliability test cases in a UAV ground dynamic target recognition and tracking scenario. The main steps of this invention are as follows:

[0005] Step 1: Analyze and determine the static reliability influencing factors in UAV ground dynamic target recognition and tracking scenarios.

[0006] Static reliability factors affecting UAV target identification and tracking include weather, roads, and buildings.

[0007] Step 1: Analyze and determine weather influencing factors

[0008] Weather influencing factors include weather type, weather grade, wind force, and light intensity. The selection of weather type and weather grade affects the selection of light intensity.

[0009] (1) There are w weather types such as sunny, rainy, foggy, snowy, etc., denoted as {α1, α2, α3, …, α r2 , p , ln , lp , l1 , p , lp , rn , r1 , l2 , n , p , p , lp , Figure 2};

[0010] (2) Each weather type has 4 weather grades: general, relatively severe, severe, and extremely severe, denoted as {β1, β2, β3, β4};

[0011] (3) There are 12 grades of wind force, denoted as {c1, c2, c3, …, c 12};

[0012] (4) The light intensity varies every day. Denote the light intensity as d, d > 0, unit: lux (lx).

[0013] Step 2: Analyze and determine road influencing factors

[0014] Road influencing factors include the number of roads, length, and width. Collect the road environments of different cities, number each road in the urban road environment, denoted as {R1, R2, …, R n}, where n represents the total number of roads in the city. Taking any intersection as the origin, the positive x-axis direction as the due east direction, the positive y-axis direction as the due north direction, and the direction perpendicular to the urban ground upward as the positive z-axis direction, establish a three-dimensional coordinate system in the urban road environment to record the positions of each road. Consider the two sides of the road as two parallel lines, represent them with expressions in the three-dimensional coordinate system. Through the expressions, the length and width of each road can be obtained. Number the expressions on the left and right sides of each road respectively, denoted as {R l1 , R l2 , …, R ln}, {R r1 , R r2 , …, R rn}. The expression number on the left side of the p-th road is R lp , and the expression is: R lp : A p x + B p y + C lp = 0, z = 0 (-l < x < l'), as shown in Figure 2 , where A p and B p are coefficient constants used to jointly determine the road orientation, obtained by collecting multiple sample points on the same side of the road and performing linear fitting calculations; Clp The term representing the offset constant of the left-hand road relative to the origin is in A. p and B p Given the information, collect the points on the left side of the road and substitute them into the above expression to obtain the solution; x represents the horizontal coordinate of the road in the coordinate system, and y represents the vertical coordinate of the road in the coordinate system.

[0015] Step 3: Analyze and determine the set of building influencing factors. Represent the height and length of buildings on both sides of the road in a three-dimensional coordinate system, and analyze the impact of building shading. In the three-dimensional coordinate system, number the buildings on the left and right sides of each road, denoted as {b1, b2, ..., b m Record the coordinates of each building on the z=0 plane at the two vertices closest to the road. For example, the building numbered 'a' has coordinates 'b'. a Given a variable {(x,y,0),(x',y',0)} (a∈(1,2,…,m)), record the height of each building as {h1,h2,…,h...}. m}

[0016] Step 2: Analyze the factors affecting dynamic reliability in UAV ground dynamic target recognition and tracking scenarios.

[0017] This study analyzes the dynamic interference factors that may occur during the identification and tracking of ground targets by unmanned aerial vehicles (UAVs). The main dynamic influencing factors for UAV target identification and tracking include: the target object, other moving objects, and the preset state of the UAV.

[0018] Step 1: Analyze and determine the influencing factors of dynamic ground targets and other interfering targets.

[0019] Factors influencing ground dynamic targets and other interfering targets include target type, speed, width, length, and color.

[0020] (1) Construct a set of types of ground dynamic targets and other interference targets, denoted as {T0,T1,T2,…,T…} n}, where T0 represents the type of dynamic ground target, T i (i>0) indicates the type of other interfering targets numbered i;

[0021] (2) After determining the types of ground dynamic targets and other interfering targets, their width and length are also determined, denoted as {W0, W1, W2, ..., W...} n} and {L0,L1,L2,…,L n}, where W0 represents the width of the dynamic ground target, W i (i>0) represents the width of the other interfering target numbered i; L0 represents the length of the dynamic ground target, Li (i>0) indicates the length of other interfering targets numbered i;

[0022] (3) Construct a velocity set for ground dynamic targets and other interfering targets, denoted as {v0(t), v1(t), v2(t), ..., v n v(t)}, where v0(t) represents the velocity of the dynamic ground target at time t, and v i (t)(i>0) represents the velocity of the other interfering target numbered i at time t. Simultaneously, the center position of the ground dynamic target in the three-dimensional coordinate system at time t is recorded and denoted as (x... o (t),y o (t),0);

[0023] (4) Construct a color set for ground dynamic targets and other interfering targets, denoted as {c1, c2, c3, ..., c n}, where c0 represents the color of the dynamic ground target, c i (i>0) represents the color of other interfering targets numbered i.

[0024] Step 2: Analyze the influencing factors of the drone itself.

[0025] Factors influencing the performance of a drone include its initial velocity, altitude, and flight direction. Based on the characteristics of drones, their velocity is represented in a three-dimensional coordinate system using v. x (t) represents the velocity of the UAV along the x-axis at time t, v y (t) represents the velocity of the UAV in the y-axis direction at time t, v z (t) represents the velocity of the drone in the z-axis direction at time t, and records the coordinate position (x) of the drone at the start of its flight. u ,y u ,h u By presetting different initial speeds and positions of the drone, the reliability of drone target recognition and tracking can be further analyzed.

[0026] Step 3: Analyze and determine the occlusion effects caused by building factors in the scenario of UAV identification and tracking of dynamic ground targets.

[0027] The analysis determines the occlusion effect of buildings on ground dynamic targets in scenarios where UAVs identify and track ground dynamic targets, and identifies the occlusion range. The specific steps are as follows:

[0028] Step 1: Determine whether the building can provide shade.

[0029] At time t, the position coordinates of the UAV are: The coordinates of the ground dynamic target's position are (xo (t),y o (t), 0), located on road p. The expressions for the left and right sides of this road are R and R, respectively. lp :A p x+B p y+C lp =0, z=0 (-l<x<l'), R rp :A p x+B p y+C rp =0, z=0 (-l<x<l'), where, C rp The offset constant term of the right-hand road relative to the origin is obtained by collecting points on the right side of the road and substituting them into the above expression.

[0030] Calculate the distance from the drone's location to both sides of the road, when When the building is in use, it can obstruct dynamic targets on the ground; otherwise, there is no obstruction effect.

[0031] Step 2: Calculate the obstruction range of the building.

[0032] After determining that the building could cause obstruction, determine which side of the road the drone should be on. At this time, the building on the left will obstruct the view of the dynamic target on the ground. The building closest to the dynamic target on the left road is b. a The height is h a At this time, the obstruction range w provided by the building l for:

[0033]

[0034] Similarly, when At this time, the building on the right will obstruct the view of dynamic targets on the ground. The building closest to the dynamic target on the right road is b. s The height is h s At this time, the obstruction range w provided by the building r for:

[0035]

[0036] Step 3: Calculate the occlusion range of buildings on dynamic targets on the ground.

[0037] (1) When the UAV is on the left side of the road, the buildings on the left side of the road obstruct the view of the ground-based dynamic target vehicle. The distance from the ground-based dynamic target vehicle to the left side of the road is:

[0038]

[0039] Therefore, the occlusion range w0 caused by the buildings on the left side of the road to dynamic target vehicles on the ground can be obtained as:

[0040]

[0041] Among them, w l W0 represents the width of the obstruction by the building on the left, and W0 represents the width of the dynamic ground target.

[0042] (2) When the UAV is on the right side of the road, the buildings on the right side of the road obstruct the view of the ground-based dynamic target vehicle. The distance from the ground-based dynamic target vehicle to the right side of the road is:

[0043]

[0044] Therefore, the occlusion range w0 caused by the buildings on the right side of the road to dynamic target vehicles on the ground can be obtained as:

[0045]

[0046] Among them, w r W0 represents the width of the obstruction by the building on the right, and W0 represents the width of the dynamic ground target. Figure 4 As shown.

[0047] Step 4: Collect data, perform a goodness-of-fit test on the distribution function, and select the optimal distribution of parameters affecting the identification and tracking of ground dynamic targets by UAVs.

[0048] Collect data on the following factors in a UAV-based dynamic target recognition and tracking scenario, process the data, and select the optimal distribution of each influencing parameter of the UAV. This step includes the following three steps:

[0049] Step 1: Collect data from the UAV during target identification and tracking tasks, process and analyze the data to obtain the possible distribution types of each parameter.

[0050] (1) Based on the scenario of unmanned vehicle identification and tracking of dynamic ground targets, collect data on the reliability influencing factors of unmanned vehicles in flight scenarios {X1,X2,X3…,X…} n}

[0051] (2) The collected data from the UAVs in real-world scenarios is filtered and noisy data is removed. The minimum and maximum values ​​of the parameters affecting UAV reliability are taken [X]. min ,X max This serves as a reference range for the parameter's value, and the data is sorted from smallest to largest. Using the empirical formula t = 1 + 3.3lgn (where n is the number of data points, and t is rounded up), the data is divided into t groups. Then, according to the formula Vx = (X... max -Xmin The class interval is determined by t. The interval in each class (x1, x2, ..., xt) is calculated. t The frequency of )

[0052]

[0053] (3) Based on the obtained data, draw the distribution diagram of each reliability parameter of the UAV, fit the distribution to obtain the distribution function, compare it with the theoretical distribution function, and obtain the possible distribution of the parameters.

[0054] Step 2: Use the chi-square distribution goodness-of-fit test to examine the discrete distribution parameters among the parameters affecting UAV reliability, and select the optimal distribution type. In Step 1, the usable data obtained for the UAV's identification and tracking of dynamic ground targets is {x1, x2, x3, ..., x...}. t In possible distribution models, its measure is:

[0055]

[0056] Where, x i 'For possible distribution models with x i For the corresponding position, the critical value can be obtained for a given significance level α = 0.05. when The observed value is less than or equal to the critical value Then it can be assumed that the influencing parameter follows this distribution.

[0057] Step 3: Use the grey relational analysis method based on grey theory to test the continuous distribution parameters in the parameters affecting UAV reliability and select the optimal distribution type.

[0058] Let the available data in the scenario of UAV identification and tracking of dynamic ground targets be denoted as reference sequence x. (0) : {x1,x2,x3,…,x t Simultaneously, the frequency of the position corresponding to the reference sequence in distribution type 1 can be calculated to obtain the comparison sequence x. (1) Similarly, we can obtain the comparison sequence x. (2) ,x (3) ,…,x (j) Based on grey theory analysis, the correlation coefficients for each distribution type of the unmanned aerial vehicle (UAV) reliability impact parameters can be calculated:

[0059]

[0060] Where, ξ (i)k Let be the correlation coefficient of point k in the i-th reference sequence, and ρ be the resolution coefficient, which takes a value between 0 and 1, typically 0.5. Therefore, the correlation degree can be calculated as:

[0061]

[0062] Finally, the comparison sequence with the largest correlation coefficient is selected to obtain the optimal distribution of parameters affecting UAV reliability.

[0063] Step 5: Perform importance sampling to generate a test environment for UAV target recognition and tracking, and determine test cases.

[0064] Based on the cross-entropy optimization method, the important sampling distribution of parameters affecting UAV reliability is determined, and test scenarios for UAV identification and tracking of dynamic ground targets are generated by sampling. The specific steps for determining the test scenarios for UAVs are as follows:

[0065] (1) Obtain the probability density functions f1(x), f2(x), ..., f of the various reliability-affecting parameters of the UAV through step four. n (x), using the cross-entropy optimization method, determine the important sampling probability density functions of each influencing parameter of the UAV. and weight

[0066] (2) The important sampling probability density function of each independent reliability impact parameter of the UAV Sample from each parameter set {α,β,c,d,…} and generate n test scenarios for UAV recognition and tracking of dynamic ground targets.

[0067] (3) Place the UAV in the test scenario, set the test time, and record the time when the UAV successfully identifies and tracks the ground dynamic target during the test.

[0068] By following the steps above, key test scenarios for UAV target recognition and tracking tasks can be obtained, and the reliability of the UAV in performing target recognition and tracking tasks can be evaluated through the test results. (See attached diagram.)

[0069] Figure 1 Method and Flow

[0070] Figure 2 UAV-based identification and tracking of dynamic ground targets: road plan diagram

[0071] Figure 3 Schematic diagram of building obstruction in drone testing scenario

[0072] Figure 4 Schematic diagram of building occlusion of dynamic ground targets

[0073] Figure 5 Comparison of drone impact parameter distribution fitting chart Specific implementation methods

[0074] Step 1: Analyze and determine the static reliability influencing factors in UAV ground dynamic target recognition and tracking scenarios.

[0075] Static reliability factors affecting UAV target identification and tracking include weather, roads, and buildings.

[0076] Step 1: Analyze and determine the factors influencing weather

[0077] Weather factors include weather type, weather level, wind force, and sunlight intensity, and the choice of weather type and weather level will affect the choice of sunlight intensity.

[0078] (1) Weather has w types, such as sunny, rainy, foggy, and snowy, which are denoted as {α1, α2, α3, ..., α...} w};

[0079] (2) Each weather type has 4 weather levels: general, moderate, severe, and extremely severe, denoted as {β1,β2,β3,β4};

[0080] (3) There are 12 levels of wind force, denoted as {c1, c2, c3, ..., c 12};

[0081] (4) The light intensity varies every day. Let the light intensity be d, where d>0, and the unit is lux (lx).

[0082] Step 2: Analyze and determine the factors affecting the road

[0083] Road influencing factors include the number of roads, their length, and their width. Road environment data from different cities is collected, and each road in the urban road environment is numbered and denoted as {R1, R2, ..., R...}. n} where n represents the total number of roads in the city. A three-dimensional coordinate system is established within the urban road environment, with a certain intersection as the origin, east as the positive x-axis, north as the positive y-axis, and the direction perpendicular to the city ground as the positive z-axis, recording the position of each road. The two sides of each road are considered as two parallel lines, represented by expressions in the three-dimensional coordinate system. The length and width of each road can be obtained from these expressions. The expressions for the left and right sides of each road are numbered and denoted as {R}. l1 ,R l2 ,…,R ln}, {R r1 ,R r2 ,…,R rn}. For example, the expression on the left side of the p-th road is numbered R. lp The expression is: R lp :A p x+B pyz + C lp = 0, z = 0 (-l < x < l'), as Figure 2 shown.

[0084] Step 3: Analyze and determine the set of building influence factors, represent the heights and lengths of the buildings on both sides of the road in a three-dimensional coordinate system, and analyze the influence brought by building occlusion. In the three-dimensional coordinate system, number the buildings on the left and right sides of each road respectively, denoted as {b1, b2, …, b m}, record the coordinate positions of the two vertices of each building on the side close to the road in the z = 0 plane, denoted as b a : {(x a , y a , 0), (x' a , y' a , 0)} (a ∈ (1, 2, …, m)), and record the height of each building, denoted as {h1, h2, …, h m}.

[0085]

Example

[0086]

[0087] Step 2: Analyze the dynamic reliability influence factors in the scenario of UAV ground dynamic target recognition and tracking.

[0088] Analyze the possible dynamic interference factors during the process of UAV identifying and tracking ground dynamic targets. The dynamic influence factors of UAV target recognition and tracking mainly include: target objects, other dynamic objects, and the preset state of the UAV.

[0089] Step 1: Analyze and determine the influence factors of ground dynamic targets and other interference targets

[0090] The influence factors of ground dynamic targets and other interference targets include the type, speed, width, length, and color of the targets.

[0091] (1) Construct a set of types of ground dynamic targets and other interference targets, denoted as {T0, T1, T2, …, T n}, where T0 represents the type of ground dynamic targets, and T i (i > 0) represents the type of the i-th other interference target;

[0092] (2) After determining the types of ground dynamic targets and other interference targets, their widths and lengths are also determined accordingly, denoted as {W0, W1, W2, …, W n} and {L0,L1,L2,…,L n}, where W0 represents the width of the dynamic ground target, W i (i>0) represents the width of the other interfering target numbered i; L0 represents the length of the dynamic ground target, L i (i>0) indicates the length of other interfering targets numbered i;

[0093] (3) Construct a velocity set for ground dynamic targets and other interfering targets, denoted as {v0(t), v1(t), v2(t), ..., v n v(t)}, where v0(t) represents the velocity of the dynamic ground target at time t, and v i (t)(i>0) represents the velocity of the other interfering target numbered i at time t. Simultaneously, the center position of the ground dynamic target in the three-dimensional coordinate system at time t is recorded and denoted as (x... o (t),y o (t),0);

[0094] (4) Construct a color set for ground dynamic targets and other interfering targets, denoted as {c1, c2, c3, ..., c n}, where c0 represents the color of the dynamic ground target, c i (i>0) represents the color of the other interfering target numbered i;

[0095] Step 2: Analyze the influencing factors of the drone itself.

[0096] Factors influencing the performance of a drone include its initial velocity, altitude, and flight direction. Based on the characteristics of drones, their velocity is represented in a three-dimensional coordinate system using v. x (t) represents the velocity of the UAV along the x-axis at time t, v y (t) represents the velocity of the UAV in the y-axis direction at time t, v z (t) represents the velocity of the drone in the z-axis direction at time t, and records the coordinate position (x) of the drone at the start of its flight. u ,y u ,h u By presetting different initial speeds and positions of the drone, the reliability of drone target recognition and tracking can be further analyzed.

[0097]

Example

[0098] Table 1 Characteristics of different types of vehicles

[0099]

[0100] Step 3: Analyze and determine the occlusion effects caused by building factors in the scenario of UAV identification and tracking of dynamic ground targets.

[0101] The analysis determines the occlusion effect of buildings on ground dynamic targets in scenarios where UAVs identify and track ground dynamic targets, and identifies the occlusion range. The specific steps are as follows:

[0102] Step 1: Determine whether the building can provide shade.

[0103] At time t, the position coordinates of the UAV are: The coordinates of the ground dynamic target's position are (x o (t),y o (t), 0), located on road p. The expressions for the left and right sides of this road are R and R, respectively. lp :A p x+B p y+C lp =0, z=0 (-l<x<l'), R rp :A p x+B p y+C rp =0, z=0 (-l<x<l').

[0104] Calculate the distance from the drone's location to both sides of the road, when When the building is in use, it can obstruct dynamic targets on the ground; otherwise, there is no obstruction effect.

[0105] Step 2: Calculate the building's obstruction range

[0106] After determining that the building could cause obstruction, determine which side of the road the drone should be on. At this time, the building on the left will obstruct the view of the dynamic target on the ground. The building closest to the dynamic target on the left road is b. a The height is h a At this time, the obstruction range w provided by the building l for:

[0107]

[0108] Similarly, when At this time, the building on the right will obstruct the view of dynamic targets on the ground. The building closest to the dynamic target on the right road is b. s The height is h s At this time, the obstruction range w provided by the building r for:

[0109]

[0110] Step 3: Calculate the occlusion range of buildings on dynamic targets on the ground.

[0111] (1) When the UAV is on the left side of the road, the buildings on the left side of the road obstruct the view of the ground-based dynamic target vehicle. The distance from the ground-based dynamic target vehicle to the left side of the road is:

[0112]

[0113] Therefore, the occlusion range w0 caused by the buildings on the left side of the road to dynamic target vehicles on the ground can be obtained as:

[0114]

[0115] Among them, w l W0 represents the width of the obstruction by the building on the left, and W0 represents the width of the dynamic ground target.

[0116] (2) When the UAV is on the right side of the road, the buildings on the right side of the road obstruct the view of the ground-based dynamic target vehicle. The distance from the ground-based dynamic target vehicle to the right side of the road is:

[0117]

[0118] Therefore, the occlusion range w0 caused by the buildings on the right side of the road to dynamic target vehicles on the ground can be obtained as:

[0119]

[0120] Among them, w r W0 represents the width of the obstruction by the building on the right, and W0 represents the width of the dynamic ground target.

[0121]

Example

[0122] (1) Determine whether the building can provide shade

[0123] Therefore, buildings can have an obstructive effect.

[0124] (2) Calculate the shading range of the building.

[0125] Therefore, the obstruction is caused by the building on the right, and the area of ​​obstruction caused by the building is as follows:

[0126]

[0127] (3) Calculate the obstruction range of buildings on dynamic targets on the ground.

[0128]

[0129] Therefore, the obstruction range of the target vehicle caused by the buildings on the right side of the road can be calculated as follows:

[0130]

[0131] At this point, the building completely obscures the target vehicle.

[0132] Step 4: Collect data, perform a goodness-of-fit test on the distribution function, and select the optimal distribution of parameters affecting the identification and tracking of ground dynamic targets by UAVs.

[0133] Collect data on the following factors in a UAV-based dynamic target recognition and tracking scenario, process the data, and select the optimal distribution of each influencing parameter of the UAV. This step includes the following three steps:

[0134] Step 1: Collect data from the UAV during target identification and tracking tasks, process and analyze the data to obtain the possible distribution types of each parameter.

[0135] (1) Based on the scenario of unmanned vehicle identification and tracking of dynamic ground targets, collect data on the reliability influencing factors of unmanned vehicles in flight scenarios {X1,X2,X3…,X…} n}

[0136] (2) The collected data from the UAVs in real-world scenarios is filtered and noisy data is removed. The minimum and maximum values ​​of the parameters affecting UAV reliability are taken [X]. min ,X max This serves as a reference range for the parameter's value, and the data is sorted from smallest to largest. Using the empirical formula t = 1 + 3.3lgn (where n is the number of data points, and t is rounded up), the data is divided into t groups. Then, according to the formula Vx = (X... max -X min The class interval is determined by t. The interval in each class (x1, x2, ..., xt) is calculated. t The frequency of )

[0137]

[0138] (3) Based on the obtained data, draw the distribution diagram of each reliability parameter of the UAV, fit the distribution to obtain the distribution function, compare it with the theoretical distribution function, and obtain the possible distribution of the parameters.

[0139] Step 2: Use the chi-square distribution goodness-of-fit test to examine the discrete distribution parameters among the parameters affecting UAV reliability, and select the optimal distribution type. In Step 1, the usable data obtained for the UAV's identification and tracking of dynamic ground targets is {x1, x2, x3, ..., x...}. t In possible distribution models, its measure is:

[0140]

[0141] Where, x i 'For possible distribution models with x i For the corresponding position, the critical value can be obtained for a given significance level α = 0.05. when The observed value is less than or equal to the critical value Then it can be assumed that the influencing parameter follows this distribution.

[0142] Step 3: Use the grey relational analysis method based on grey theory to test the continuous distribution parameters in the parameters affecting UAV reliability and select the optimal distribution type.

[0143] Let the available data in the scenario of UAV identification and tracking of dynamic ground targets be denoted as reference sequence x. (0) : {x1,x2,x3,…,x t Simultaneously, the frequency of the position corresponding to the reference sequence in distribution type 1 can be calculated to obtain the comparison sequence x. (1) Similarly, we can obtain the comparison sequence x. (2) ,x (3) ,…,x (j) Based on grey theory analysis, the correlation coefficients for each distribution type of the unmanned aerial vehicle (UAV) reliability impact parameters can be calculated:

[0144]

[0145] Where, ξ (i)k Let be the correlation coefficient of point k in the i-th reference sequence, and ρ be the resolution coefficient, typically taken as 0.5. Therefore, the correlation degree can be calculated as:

[0146]

[0147] Finally, the comparison sequence with the largest correlation coefficient is selected to obtain the optimal distribution of parameters affecting UAV reliability.

[0148]

Example

[0149] Table 2 Vehicle Length

[0150]

[0151] After filtering and removing noisy data from the length data of the aforementioned 1000 vehicles, 821 vehicle length data points were obtained. The minimum and maximum values ​​of these data points are: [X] min ,X max = [2.00, 5.00], and sort them. From the empirical formula, t = 1 + 3.3lgn = 10.62, rounding up gives t = 11, and the class interval for each group is Vx = (X max -X min ) / (t-1)=0.3. The frequencies of each group of data are shown in Table 3.

[0152] Table 3 Frequency of each group of factors affecting vehicle length

[0153]

[0154] Plotting revealed that the parameters influencing vehicle length may follow normal, Weibull, and log-normal distributions. Since the factors affecting vehicle length are continuously distributed, the grey relational analysis method based on grey theory was used to examine the road influence parameters. Therefore, the frequency of road data in each group can be used as the reference sequence, and the normal distribution as the comparison sequence x. (1) The Weibull distribution and the log-normal distribution are used as comparison sequences x. (2) ,x (3) The correlation coefficients for each point are shown in Table 4.

[0155] Table 4 Correlation Coefficient Table

[0156] <![CDATA[ξ (1) ]]> 0.50 0.38 0.57 0.46 0.44 0.82 0.64 0.35 0.96 1.00 0.45 <![CDATA[ξ (2) ]]> 0.74 0.77 1.00 0.54 0.41 0.39 0.37 0.56 0.84 0.62 0.44 <![CDATA[ξ (3) ]]> 0.65 0.48 0.49 0.80 0.96 1.00 0.61 0.73 0.69 0.42 0.36

[0157] The correlation degree r of each reference sequence is obtained. (1) =0.6<r (2) =0.61<r (3) =0.65, therefore the log-normal distribution is chosen as the distribution type for the factors affecting vehicle length.

[0158] Step 5: Perform importance sampling to generate a test environment for UAV target recognition and tracking, and determine test cases.

[0159] Based on the cross-entropy optimization method, the important sampling distribution of parameters affecting UAV reliability is determined, and test scenarios for UAV identification and tracking of dynamic ground targets are generated by sampling. The specific steps for determining the test scenarios for UAVs are as follows:

[0160] (1) Obtain the probability density functions f1(x), f2(x), ..., f of the various reliability-affecting parameters of the UAV through step four. n (x), using the cross-entropy optimization method, determine the important sampling probability density functions of each influencing parameter of the UAV. and weight

[0161] (2) The important sampling probability density function of each independent reliability impact parameter of the UAV Sample from each sample to obtain n sets of parameters {α,β,c,d,…}, and generate n test environments for UAVs to identify and track dynamic ground targets;

[0162] (3) Place the UAV in the test scenario, set the test time, and record the time when the UAV successfully identifies and tracks the ground dynamic target during the test.

[0163] By following the steps above, we can obtain the key test scenarios for UAV target recognition and tracking tasks, and evaluate the reliability of UAVs in performing target recognition and tracking tasks through the test results.

[0164] [Example] The test scenarios and results obtained by sampling UAV ground dynamic target recognition and tracking scenarios using the importance-based sampling method are shown in the table below:

[0165] Table 5 Test scenarios and results

[0166]

[0167]

[0168] By setting a test time, the time it takes for the UAV to successfully identify and track a target during the process of identifying and tracking dynamic targets on the ground is recorded, and the reliability of the UAV is finally obtained.

Claims

1. A method for generating reliability test cases for UAV ground dynamic target recognition and tracking scenarios, characterized in that... It includes the following steps: Step 1: Analyze and determine the static reliability influencing factors in UAV ground dynamic target recognition and tracking scenarios; Step 1: Analyze and determine the factors influencing weather; (1) Weather has w types, such as sunny, rainy, foggy, and snowy, which are denoted as {α1, α2, α3, ..., α...} w }; (2) Each weather type has 4 weather levels: general, moderate, severe, and extremely severe, denoted as {β1,β2,β3,β4}; (3) There are 12 levels of wind force, denoted as {c1, c2, c3, ..., c 12 }; (4) The light intensity varies from day to day. Let the light intensity be d, where d>0. The unit is lux (lx). Step 2: Analyze and determine the factors affecting the road; Collect road environment data from different cities, and assign a number to each road in the urban road environment, denoted as {R1, R2, ..., R...}. n }, where n represents the total number of roads in the city, and a three-dimensional coordinate system is established in the urban road environment with a certain intersection as the origin, due east as the positive x-axis, due north as the positive y-axis, and the direction perpendicular to the city ground as the positive z-axis, recording the position of each road. The expressions on the left and right sides of each road are numbered and denoted as {R}. l1 ,R l2 ,…,R ln }, {R r1 ,R r2 ,…,R rn }; Step 3: In the three-dimensional coordinate system, number the buildings on the left and right sides of each road, denoted as {b1, b2, ..., b m Record the coordinates of each building on the two vertices closest to the road on the z=0 plane, denoted as b. a :{(x a ,y a ,0),(x' a ,y' a Given a variable {h1, h2, ..., m} (a ∈ (1, 2, ..., m)), record the height of each building, denoted as {h1, h2, ..., hm}. m }; Step 2: Analyze the factors affecting dynamic reliability in UAV ground dynamic target recognition and tracking scenarios; Step 1: Analyze and determine the influencing factors of dynamic ground targets and other interfering targets; Factors influencing ground dynamic targets and other interfering targets include target type, speed, width, length, and color: (1) Construct a set of types of ground dynamic targets and other interference targets, denoted as {T0,T1,T2,…,T…} n }, where T0 represents the type of dynamic ground target, T i (i>0) indicates other types of interference targets with the number i; (2) After determining the types of ground dynamic targets and other interfering targets, their width and length are also determined, denoted as {W0, W1, W2, ..., W...} n } and {L0,L1,L2,…,L n }, where W0 represents the width of the dynamic ground target, W i (i>0) represents the width of the other interfering target numbered i; L0 represents the length of the dynamic ground target, L i (i>0) indicates the length of other interfering targets numbered i; (3) Construct a velocity set for ground dynamic targets and other interfering targets, denoted as {v0(t), v1(t), v2(t), ..., v n Let v0(t)} represent the velocity of the ground dynamic target at time t, and v i (t)(i>0) represents the velocity of the other interfering target numbered i at time t. Simultaneously, the center position of the ground dynamic target in the three-dimensional coordinate system at time t is recorded and denoted as (x... o (t),y o (t),0); (4) Construct a color set for ground dynamic targets and other interfering targets, denoted as {c1, c2, c3, ..., c n }, where c0 represents the color of the dynamic ground target, c i (i>0) represents the color of the other interfering target numbered i; Step 2: Analyze the influencing factors of the drone itself; Represent the drone's velocity in a three-dimensional coordinate system: using v x (t) represents the velocity of the UAV in the x-axis direction at time t, v y (t) represents the velocity of the UAV in the y-axis direction at time t, v z (t) represents the velocity of the drone in the z-axis direction at time t, and records the coordinate position (x) of the drone at the start of its flight. u ,y u ,h u By presetting different initial speeds and positions of the drone, the reliability of drone target recognition and tracking can be further analyzed; Step 3: Analyze and determine the occlusion effects caused by building factors in the scenario of UAV identification and tracking of dynamic ground targets; Step 1: Determine whether the building will provide shade; At time t, the position coordinates of the UAV are: The coordinates of the ground dynamic target's position are (x o (t),y o (t), 0), located on road p, where the expressions for the left and right sides of the road are R and R, respectively. lp :A p x+B p y+C lp =0, z=0 (-l<x<l'), R rp :A p x+B p y+C rp =0, z=0 (-l<x<l'), where, A p and B p The constant coefficient used to jointly determine the road orientation is obtained by collecting multiple sample points on the same side of the road and performing straight line fitting calculations; C lp C rp These are the offset constants of the left and right roads relative to the origin, respectively, in A. p and B p Given the information, collect points on both sides of the road and substitute them into the above expression to obtain the solution; x represents the x-coordinate of the road in the coordinate system, and y represents the y-coordinate of the road in the coordinate system. Calculate the distance from the drone's location to both sides of the road, when The presence of a building indicates that it can obstruct dynamic targets on the ground; otherwise, there is no obstruction effect. Step 2: Calculate the obstruction range of the building; After determining that the building could cause obstruction, determine which side of the road the drone should be on. At this time, the building on the left will obstruct the view of the dynamic target on the ground. The building closest to the dynamic target on the ground on the left road is b. a The height is h a At this time, the shading range w caused by the building is: Similarly, when At this time, the building on the right will obstruct the view of dynamic targets on the ground. The building closest to the dynamic target on the right road is b. s The height is h s At this time, the shading range w caused by the building is: Step 3: Calculate the occlusion range of buildings on dynamic targets on the ground; (1) When the UAV is on the left side of the road, the buildings on the left side of the road obstruct the view of the ground dynamic target vehicle. At this time, the distance from the ground dynamic target vehicle to the left side of the road is: Therefore, the occlusion range w0 caused by the buildings on the left side of the road to dynamic target vehicles on the ground can be obtained as: (2) When the UAV is on the right side of the road, the buildings on the right side of the road obstruct the view of the ground dynamic target vehicle. At this time, the distance from the ground dynamic target vehicle to the right side of the road is: Therefore, the occlusion range w0 caused by the buildings on the right side of the road to dynamic target vehicles on the ground can be obtained as: Step 4: Collect data, perform a goodness-of-fit test on the distribution function, and select the optimal distribution of parameters affecting the identification and tracking of ground dynamic targets by UAVs; Step 1: Collect data from the UAV during target identification and tracking tasks, process and analyze the data to obtain the possible distribution types of each parameter; (1) Based on the scenario of unmanned vehicle identification and tracking of dynamic ground targets, collect data on the reliability influencing factors of unmanned vehicles in flight scenarios {X1,X2,X3…,X…} n }; (2) The collected data of the UAV in real-world scenarios is filtered and processed to remove noisy data, and the minimum and maximum values ​​of the parameters affecting UAV reliability are taken [X]. min ,X max This serves as a reference range for the parameter's value. The data is sorted from smallest to largest, and divided into t groups using the empirical formula t = 1 + 3.3lgn (where n is the number of data points, and t is rounded up). Then, according to the formula Vx = (X... max -X min Determine the class interval by ) / t, and calculate the interval in each class (x1, x2, ..., xt). t The frequency of ) (3) Based on the obtained data, draw the distribution diagram of each reliability parameter of the UAV, fit the distribution to obtain the distribution function, compare it with the theoretical distribution function, and obtain the possible distribution of the parameters; Step 2: Use the chi-square distribution goodness-of-fit test to examine the discrete distribution parameters among the parameters affecting UAV reliability, and select the optimal distribution type; In Step 1, the usable data obtained for UAV identification and tracking of ground dynamic targets in the scenario are {x1, x2, x3, ..., x t In possible distribution models, its measure is: Where, x i 'For possible distribution models with x i For the corresponding position, the critical value can be obtained for a given significance level α = 0.

05. when The observed value is less than or equal to the critical value Then it can be assumed that the influencing parameter follows this distribution; Step 3: Use the grey relational analysis method based on grey theory to test the continuous distribution parameters in the parameters affecting UAV reliability and select the optimal distribution type; Let the available data in the scenario of UAV identification and tracking of dynamic ground targets be denoted as reference sequence x. (0) :{x1,x2,x3,…,x t Simultaneously, the frequency of the corresponding position in distribution type 1 corresponding to the reference sequence can be calculated to obtain the comparison sequence x. (1) Similarly, the comparison sequence x can be obtained. (2) ,x (3) ,…,x (j) Based on the grey theory analysis method, the correlation coefficients of various distribution types of UAV reliability impact parameters can be calculated: Where, ξ (i)k ξ represents the correlation coefficient of the i-th comparison sequence relative to the reference sequence at point k, used to characterize the closeness of the two sequences at that time; ρ is the resolution coefficient, taking values ​​between 0 and 1, used to adjust the correlation coefficient ξ. (i)k This improves the resolution and computational stability; therefore, the correlation degree can be calculated as: Finally, the comparison sequence with the largest correlation coefficient is selected to obtain the optimal distribution of the parameters affecting UAV reliability; Step 5: Perform importance sampling to generate a test environment for UAV target recognition and tracking, and determine test cases; (1) Obtain the probability density functions f1(x), f2(x), ..., f of the various reliability-affecting parameters of the UAV through step four. n (x), using the cross-entropy optimization method, determine the important sampling probability density functions of each influencing parameter of the UAV. and weight (2) The important sampling probability density function of each independent reliability impact parameter of the UAV Sample from each sample to obtain n sets of parameters {α,β,c,d,…}, and generate n test environments for UAVs to identify and track dynamic ground targets; (3) Place the UAV in the test scenario, set the test time, and record the time when the UAV successfully identifies and tracks the ground dynamic target during the test.

Images