A method and system for high-fidelity modeling and visual simulation integration of aircraft motion

Abstract

The application discloses a kind of high fidelity modeling and visual simulation integration method and system of aircraft motion, it is related to flight control simulation technical field.The technical points of the present application include: constructing unmanned aerial vehicle three-dimensional geometric model according to the three-dimensional geometric parameters of unmanned aerial vehicle component;Aerodynamic data is obtained by using computational fluid dynamics software based on unmanned aerial vehicle three-dimensional geometric model;Aerodynamic data is extended to five-dimensional aerodynamic query matrix by linear extrapolation method;Five-dimensional aerodynamic query matrix is embedded in six-degree-of-freedom dynamic inversion unmanned aerial vehicle control model, to realize the closed loop digital simulation of unmanned aerial vehicle;Wherein, the six-degree-of-freedom dynamic inversion unmanned aerial vehicle control model includes dynamic inversion controller, aircraft mass control unit, six-degree-of-freedom dynamics unit, aerodynamic interpolation unit, path tracking unit.The present application can realize high fidelity simulation based on high fidelity six-degree-of-freedom model, and the simulation result is displayed by the rendering platform of Unreal Engine framework with high quality.

Description

Technical Field

[0001] This invention relates to the field of flight control simulation technology, specifically to a high-fidelity modeling and visualization simulation integration method and system for aircraft motion. Background Technology

[0002] As an unattended execution platform, unmanned aerial vehicles (UAVs) inherently possess numerous advantages such as high flexibility and maneuverability. With the rapid development of control technology, sensing technology, and device miniaturization, UAVs have gradually become an indispensable piece of equipment in various fields of society. UAVs are divided into two types: fixed-wing and multi-rotor. Fixed-wing UAVs maintain their motion by generating lift from the airflow speed difference above and below the wings, thus possessing characteristics such as long endurance, high speed, long range, high payload, and high service ceiling. Multi-rotor UAVs, on the other hand, are propelled by the airflow reaction force generated by the rotation of their rotors, thus possessing advantages such as small size, light weight, strong stealth, low cost, and high flexibility. In application areas, fixed-wing UAVs are mainly used in mission scenarios requiring long range and high altitude, such as highway monitoring, grassland or forest surveillance, and remote sensing; while multi-rotor UAVs are used in mission scenarios requiring flexibility, such as power line inspection, street view photography, police patrol, and short-distance transportation.

[0003] While drones are becoming increasingly sophisticated and widely used, the challenges and environments they face are also becoming more complex. During the development of fixed-wing drones, the high demands of their missions and the extreme nature of their experimental scenarios make it difficult to conduct flight tests in real-world applications. Unreal Engine (UE or UE5), developed by Epic Games, is one of the most widely used and powerful game engines due to its powerful real-time graphics rendering capabilities, highly realistic physical simulation graphics, and rich and powerful modular architecture. Unreal Engine is not only widely used in game development but also in high-precision simulation, industrial design, virtual reality, the film industry, and many other sectors.

[0004] Existing UAV simulation platforms often suffer from low realism, poor visual effects, and an inability to achieve collaborative simulation of multiple UAVs. In 2018, to better apply UAV simulation and training in experimental teaching, Chen Jinyin et al. [1] Based on the Robot Operating System (ROS), a developmental teaching drone simulation training platform was developed. This platform can be combined with various algorithms and is highly flexible. However, to achieve realistic simulations, it is necessary to use environmental information acquisition equipment to collect various information and construct the required ROS environment for the simulation, resulting in poor portability. Xian et al. [2]An inertial guidance sensor was used to replace the part of the platform that updates the UAV's position in the digital model, and a hardware-in-the-loop flight simulation platform was built. However, the Flight Gear and Google Earth visualization methods used had poor interactivity and presentation. Summary of the Invention

[0005] Therefore, this invention proposes an integrated method and system for high-fidelity modeling and visualization simulation of aircraft motion.

[0006] According to one aspect of the present invention, a high-fidelity modeling and visualization simulation integration method for aircraft motion is proposed, the method comprising the following steps:

[0007] Construct a three-dimensional geometric model of the UAV based on the three-dimensional geometric parameters of its components;

[0008] Based on the three-dimensional geometric model of the UAV, aerodynamic data are obtained by computational fluid dynamics software.

[0009] The aerodynamic data is expanded into a five-dimensional aerodynamic query matrix using linear extrapolation.

[0010] The five-dimensional aerodynamic query matrix is ​​embedded into a six-degree-of-freedom dynamic inverse UAV control model to achieve closed-loop digital simulation of the UAV.

[0011] Furthermore, the aerodynamic data includes triaxial aerodynamic resultant force and resultant torque; the process of obtaining aerodynamic data by computational fluid dynamics software includes: meshing the three-dimensional geometric model of the UAV to generate an unstructured mesh; and using the Reynolds time-averaged method to solve the problem based on the unstructured mesh to obtain the triaxial aerodynamic resultant force matrix and aerodynamic resultant torque matrix acting on the center of mass of the UAV.

[0012] Furthermore, the five-dimensional aerodynamic query matrix includes angle of attack, sideslip angle, elevator deflection, rudder deflection, and aileron deflection.

[0013] Furthermore, the six-degree-of-freedom dynamic inverse UAV control model includes a dynamic inverse controller, an aircraft mass control unit, a six-degree-of-freedom dynamics unit, and an aerodynamic interpolation unit. The dynamic inverse controller calculates control surface commands based on the UAV's position and attitude data, and then obtains simulated control surface angles through a first-order inertial element. The aerodynamic interpolation unit interpolates the control surface angles sent by the dynamic inverse controller based on a five-dimensional aerodynamic query matrix to obtain aerodynamic forces and moments at different state points, and outputs them to the six-degree-of-freedom dynamics unit. The aircraft mass control unit acquires and outputs the aircraft's current mass and moment of inertia to the six-degree-of-freedom dynamics unit. The six-degree-of-freedom dynamics unit converts the aerodynamic forces and moments into accelerations and angular accelerations based on the aircraft's current mass and moment of inertia, and then integrates these to obtain the UAV state at the next moment.

[0014] Furthermore, the dynamic inverse UAV control model also includes a path tracking unit, which is used to calculate the position vector between the current position of the UAV and the tracking path according to the pre-set tracking path, and perform a vector cross product with the velocity vector of the UAV to obtain the roll angle command of the UAV, thereby realizing continuous tracking of the tracking target.

[0015] Furthermore, the pre-set tracking path in the path tracking unit is obtained according to the following method:

[0016] Step 1: Calculate the optimal Dubins path between all M UAVs and all K targets, and store the cost of each path in the cost matrix C;

[0017] Step 2: Use the improved particle swarm optimization algorithm to search the cost matrix C and obtain the sequence I with the minimum maximum cost. u ;

[0018] Step 3: When the number of iterations reaches the maximum iteration limit, determine whether the following inequality holds true:

[0019]

[0020] In the formula, min(I u ) represents sequence I u The shortest path in the equation, max(I u ) represents sequence I u The longest path in v min v represents the minimum flight speed of the drone. max This indicates the maximum flight speed of the drone;

[0021] When the inequality holds, such that in sequence I uThe drone assigned the shortest path will circle in place one more time, shortening sequence I. u The difference between the shortest and longest paths in the equation is used to determine the inequality again.

[0022] When the inequality does not hold, based on sequence I u Calculate the desired velocity and desired arrival time so that the UAV follows sequence I. u The corresponding path is tracked at the desired speed, specifically: in sequence I u The flight speed of the drone assigned to the longest path is set to the maximum flight speed v. max Then, the other drones adjust their speeds according to their own path lengths. The expected speed of the i-th drone is as follows:

[0023]

[0024] The expected arrival time is the same for each drone, and the expected arrival time is calculated as follows:

[0025]

[0026] Furthermore, the process of searching the cost matrix C using the improved particle swarm optimization algorithm in the path tracing unit includes:

[0027] Step 21: Randomly generate initial particles and calculate their fitness values; specifically: set a feature information matrix P, P is a K*M matrix, M is the number of drones, K is the number of targets, the feature information matrix P is P(i,j)={0&1}, where a value of 1 indicates that drone j attacks target i, and a value of 0 indicates that drone j does not attack target i;

[0028] Performing column-to-column swaps on the matrix is ​​equivalent to two drones exchanging their attack targets; randomly generating swap combinations, and using these combinations to initialize the feature information matrix P, generating initial particles; the fitness value calculation process is as follows:

[0029] The cost vector x for calculating the particle allocation result i :

[0030] x i =diag(C×P)

[0031] In the formula, diag() represents the operation of extracting the diagonal elements of the matrix, and C represents the cost matrix;

[0032] The cost vector x i The position of the i-th particle is obtained by removing all zero elements, with the cost vector x. i The maximum value is taken as the fitness value of the particle;

[0033] Step 22: Compare the fitness values ​​of all particles with their historical best solutions. If the current fitness value is better, replace the historical best solution. When it is the initial state, directly use the current particle as the historical best solution. Take the historical best solution of each individual particle as the population's historical best solution.

[0034] Step 23: Perform a crossover operation between each particle and the best historical solution of the population, and select the offspring with the better fitness value;

[0035] Step 24: Perform cross-operation between each particle and its individual historical best solution, and select the offspring with the better fitness value;

[0036] Step 25: Each particle undergoes random mutation behavior, randomly swapping the attack targets among multiple groups of drones, directly updating the particle position without performing fitness value evaluation.

[0037] Step 26: Repeat the above steps before reaching the maximum number of iterations limit.

[0038] Furthermore, in the path tracking unit, during the search of the cost matrix C using the improved particle swarm optimization algorithm, the crossover operation is as follows: randomly select a set of identical crossover regions from the two parent particles; exchange the attack targets of the corresponding UAVs within the crossover regions of the two parent particles, while updating the attack targets of other UAVs, eliminating duplicate elements, and obtaining two child individuals.

[0039] Furthermore, the communication vector is set to embed the dynamic inverse UAV control model into the simulation platform and visualized through a virtual engine; the communication vector contains all UAV state information, including attitude angle, center of mass position, three-axis ground speed components, UAV angle of attack, sideslip angle, track tilt angle, and track deflection angle.

[0040] According to another aspect of the present invention, a high-fidelity modeling and visualization simulation integrated system for aircraft motion is proposed, the system comprising:

[0041] The 3D model building module is configured to build a 3D geometric model of the UAV based on the 3D geometric parameters of the UAV components.

[0042] The aerodynamic data acquisition module is configured to calculate aerodynamic data based on the three-dimensional geometric model of the UAV using computational fluid dynamics software. The aerodynamic data includes three-axis aerodynamic resultant force and resultant torque. The aerodynamic data is then extended into a five-dimensional aerodynamic query matrix through linear extrapolation. The five-dimensional aerodynamic query matrix includes angle of attack, sideslip angle, elevator deflection, rudder deflection, and aileron deflection.

[0043] The simulation control module is configured to embed the five-dimensional aerodynamic query matrix into a six-degree-of-freedom (DOF) dynamic inverse UAV control model to achieve closed-loop digital simulation of the UAV. The six-DOF dynamic inverse UAV control model includes a dynamic inverse controller, an aircraft mass control unit, a six-DOF dynamics unit, and an aerodynamic interpolation unit. The dynamic inverse controller calculates control surface commands based on the UAV's position and attitude data, and then obtains the simulated control surface angles through a first-order inertial element. The aerodynamic interpolation unit interpolates the control surface angles sent by the dynamic inverse controller based on the five-dimensional aerodynamic query matrix to obtain aerodynamic forces and moments at different state points, and outputs them to the six-DOF dynamics unit. The aircraft mass control unit acquires and outputs the aircraft's current mass and moment of inertia to the six-DOF dynamics unit. The six-DOF dynamics unit converts the aerodynamic forces and moments into accelerations and angular accelerations based on the aircraft's current mass and moment of inertia, and then integrates them to obtain the UAV state at the next moment.

[0044] Furthermore, the process of obtaining aerodynamic data using computational fluid dynamics software in the aerodynamic data acquisition module includes: meshing the three-dimensional geometric model of the UAV to generate an unstructured mesh; and using the Reynolds time-averaged method to solve the problem based on the unstructured mesh to obtain the three-axis aerodynamic resultant force matrix and aerodynamic resultant moment matrix acting on the center of mass of the UAV.

[0045] Furthermore, the dynamic inverse UAV control model also includes a path tracking unit, which is used to calculate the position vector between the current position of the UAV and the tracking path according to the pre-set tracking path, and perform a vector cross product with the velocity vector of the UAV to obtain the roll angle command of the UAV, thereby realizing continuous tracking of the tracking target.

[0046] Furthermore, the system also includes a visualization module, which is configured to set communication vectors, embed the dynamic inverse UAV control model into the simulation platform, and visualize it through a virtual engine.

[0047] The beneficial technical effects of this invention are:

[0048] This invention can achieve high-fidelity simulation based on a high-fidelity six-degree-of-freedom model and display the simulation results in high quality through a rendering platform based on the Unreal Engine architecture. Attached Figure Description

[0049] The present invention can be better understood by referring to the description given below in conjunction with the accompanying drawings, which together with the following detailed description are included in and form part of this specification, and are used to further illustrate preferred embodiments of the invention and explain the principles and advantages of the invention.

[0050] Figure 1 This is a flowchart illustrating a high-fidelity modeling and visualization simulation integration method for aircraft motion as described in an embodiment of the present invention.

[0051] Figure 2 This is a flowchart illustrating the path planning method in an embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram of the calculation process of the improved PSO algorithm in an embodiment of the present invention;

[0053] Figure 4 This is a flowchart of the simulation platform scheduling and control program in an embodiment of the present invention;

[0054] Figure 5 This is an example diagram of the simulation results of the path tracing process in an embodiment of the present invention. Detailed Implementation

[0055] To enable those skilled in the art to better understand the present invention, exemplary embodiments or examples of the present invention will be described below in conjunction with the accompanying drawings. Obviously, the described embodiments or examples are merely some, not all, of the embodiments or examples of the present invention. All other embodiments or examples obtained by those skilled in the art based on the embodiments or examples of the present invention without inventive effort should fall within the scope of protection of the present invention.

[0056] To address the shortcomings of existing simulation methods, such as low realism, poor visual effects, and inability to achieve collaborative simulation of multiple UAVs, this invention proposes a high-fidelity modeling and visualization simulation integration method and system for aircraft motion. It embeds a UAV mathematical model based on real six degrees of freedom into a simulation platform built on Unreal Engine to achieve high-fidelity, high-quality visualization simulation with better simulation results.

[0057] This invention proposes a high-fidelity modeling and visualization simulation integration method for aircraft motion, which includes the following steps:

[0058] Construct a three-dimensional geometric model of the UAV based on the three-dimensional geometric parameters of its components;

[0059] Based on the three-dimensional geometric model of the UAV, aerodynamic data are obtained by computational fluid dynamics software.

[0060] The aerodynamic data is expanded into a five-dimensional aerodynamic query matrix using linear extrapolation.

[0061] The five-dimensional aerodynamic query matrix is ​​embedded into a six-degree-of-freedom dynamic inverse UAV control model to achieve closed-loop digital simulation of the UAV.

[0062] The embodiments of the present invention will be described in detail below.

[0063] like Figure 1 As shown, the method includes: using 3D modeling software to construct a 3D geometric model of the UAV based on the 3D geometric parameters of the UAV components; using CFD software to calculate the 3D geometric model and obtain its basic aerodynamic data; establishing a five-dimensional aerodynamic query matrix (angle of attack, sideslip angle, elevator deflection, rudder deflection, aileron deflection) of the UAV through interpolation aerodynamic modeling methods; establishing a six-degree-of-freedom dynamic inverse UAV control model based on Simulink, introducing the five-dimensional aerodynamic query matrix into the control model to achieve closed-loop digital simulation of the UAV; converting the constructed UAV control model into C code using Simulink's code generation tool, and embedding the UAV control model into the co-simulation platform in the form of specified communication vectors; importing the UAV 3D geometric model into the co-simulation platform and displaying it through the UE5 engine, where the control surface rotation, attitude, and position parameters of the UAV geometric model are directly driven by the UAV control model and move in the visual scene.

[0064] According to an embodiment of the present invention, the process of constructing a three-dimensional geometric model of a UAV using three-dimensional modeling software includes: obtaining the three-dimensional geometric parameters of each component by performing proportional mapping of publicly available UAV data; then, establishing models of each component of the UAV using three-dimensional modeling software; and finally, combining the models of each independent component together to form a three-dimensional model of the UAV, wherein each independent component (elevator, aileron, rudder, propeller, etc.) has independent degrees of freedom of motion. The three-dimensional modeling software can be, for example, CATIA, or other three-dimensional modeling software.

[0065] According to an embodiment of the present invention, the process of performing aerodynamic calculations on a three-dimensional geometric model using computational fluid dynamics software includes: importing the obtained three-dimensional UAV model into the domestic mesh processing software NNW-GridStar for mesh generation, generating an unstructured mesh. Since the UAV surface is complex, it is impossible to generate a structured mesh, so an unstructured mesh is selected; then, the domestic CFD calculation software NNW-FlowStar is used to calculate aerodynamic data. The aerodynamic data is a data matrix of three-axis aerodynamic resultant force and resultant moment. The solution method is the Reynolds time-averaged method (RANS), and further, the SST (three-dimensional model) can be used to solve for the three-axis aerodynamic resultant force matrix and aerodynamic resultant moment matrix acting on the center of mass of the UAV.

[0066] According to an embodiment of the present invention, the process of establishing a five-dimensional aerodynamic query matrix (angle of attack, sideslip angle, elevator deflection, rudder deflection, aileron deflection) of an unmanned aerial vehicle (UAV) using an interpolation aerodynamic modeling method includes: based on the aerodynamic force variation law, the aerodynamic force matrix of the UAV is extended into a five-dimensional aerodynamic query matrix by linear extrapolation, the matrix is ​​imported into Simulink of MATLAB, and the aerodynamic parameters are interpolated using an nD lookup table unit to obtain the aerodynamic force and aerodynamic torque at different state points.

[0067] Because computational fluid dynamics software is time-consuming, the number of state points to be calculated is minimized when determining the calculation state. Other state points are obtained through symmetry. Taking the longitudinal channel as an example, its coordinate axes are angle of attack, sideslip angle, and elevator deflection. By utilizing the symmetry of aerodynamic data, the aerodynamic data of the sideslip angle can be linearly extrapolated to complete the calculation, expanding the triaxial aerodynamic forces and triaxial aerodynamic moments in the range of [0,3] to the range of [-3,0,3], as shown below:

[0068] CA(α,-β,δ z )=CA(α,β,δ z )

[0069] CN(α,-β,δ z )=CN(α,β,δ z )

[0070] Cz(α,-β,δ z )=-Cz(α,β,δ z )

[0071] m x (α,-β,δ z )=-m x (α,β,δ z )

[0072] m y (α,-β,δ z )=-m y (α,β,δ z )

[0073] m z (α,-β,δ z )=m z (α,β,δ z )

[0074] In the formula, (α,β,δ) z ) represent angle of attack, sideslip angle, and pitch deflection, respectively; CA, CN, Cz, m x m y m zThese represent the axial force, lateral force, normal force, body axis rolling moment, body axis yaw moment, and body axis pitch moment of the aircraft, respectively.

[0075] After performing aerodynamic compensation on all three channels (yaw, pitch, and roll), all data are merged into a five-dimensional aerodynamic query matrix through linear superposition:

[0076] C A (α,β,δ) x ,δ y ,δ z ) = C A (α,β,0,0,0)+[C A (α,β,δ) x ,0,0)-C A (α,β,0,0,0)]+[C A (α,β,0,δ) y ,0)-C A (α,β,0,0,0)]+[C A (α,β,0,0,δ) z )-C A (α,β,0,0,0)]C N (α,β,δ) x ,δ y ,δ z ) = C N (α,β,0,0,0)+[C N (α,β,δ) x ,0,0)-C N (α,β,0,0,0)]+[C N (α,β,0,δ) y ,0)-C N (α,β,0,0,0)]+[C N (α,β,0,0,δ) z )-C N (α,β,0,0,0)]C Z (α,β,δ) x ,δ y ,δ z ) = C Z (α,β,0,0,0)+[C Z (α,β,δ) x ,0,0)-C Z (α,β,0,0,0)]+[C Z (α,β,0,δ) y ,0)-C Z (α,β,0,0,0)]+[C Z (α,β,0,0,δ) z )-CZ (a,b,0,0,0)]m x (a,b,d) x ,d y ,d z )=m x (a,b,0,0,0)+[m x (a,b,d) x ,0,0)-m x (a,b,0,0,0)]

[0077] +[m x (a,b,0,d) y ,0)-m x (a,b,0,0,0)]+[m x (a,b,0,0,d) z )-m x (a,b,0,0,0)]m y (a,b,d) x ,d y ,d z )=m y (a,b,0,0,0)+[m y (a,b,d) x ,0,0)-m y (a,b,0,0,0)]+[m y (a,b,0,d) y ,0)-m y (a,b,0,0,0)]+[m y (a,b,0,0,d) z )-m y (a,b,0,0,0)]m z (a,b,d) x ,d y ,d z )=m z (a,b,0,0,0)+[m z (a,b,d) x ,0,0)-m z (a,b,0,0,0)]+[m z (a,b,0,d) y ,0)-m z (a,b,0,0,0)]+[m z (a,b,0,0,d) z )-m z (a,b,0,0,0)]

[0078] According to an embodiment of the present invention, the process of establishing a six-degree-of-freedom dynamic inverse UAV control model based on Simulink includes: the UAV control model includes a linearized dynamic inverse controller based on singular perturbation theory, an aircraft mass control unit, a six-degree-of-freedom dynamic unit considering the Earth's rotation, and an aerodynamic interpolation unit. The operation flow is as follows: the dynamic inverse controller reads the current state of the UAV (center-of-mass coordinate data and aircraft attitude angles), calculates the control surface commands, and obtains the simulated control surface angles through a first-order inertial element; the aerodynamic interpolation unit receives the control surface angles and interpolates the aerodynamic forces and torques based on the current UAV attitude data (the aerodynamic interpolation module uses MATLAB's multidimensional interpolation module nD lookup table); the six-degree-of-freedom dynamic unit obtains the current mass state of the UAV from the mass control unit, converts the aerodynamic forces and torques into acceleration and angular acceleration, and then integrates to obtain the UAV state at the next moment. The aircraft mass control unit is used to output the current mass, moment of inertia, and other information of the aircraft, providing it to the six-degree-of-freedom unit to calculate acceleration and angular acceleration using Newton's second law.

[0079] As an example, the process of obtaining the simulated control surface angle in the dynamic inverse controller includes: acquiring the user-given altitude and speed commands, and subtracting them from the UAV's current actual altitude and speed to obtain altitude and speed errors. These errors are then converted into angle-of-attack and throttle commands via proportional and integral circuits. The angle-of-attack error is obtained by calculating the actual angle of attack and the angle-of-attack command, and converted into a pitch rate command via proportional and integral circuits. The pitch rate error is obtained by calculating the actual pitch rate and the pitch rate command, and converted into a pitch control surface command via proportional and integral circuits. The control surface command is output with limitations based on the servo status to prevent exceeding the servo's operating limits. The roll angle command control process during UAV operation is similar to the angle-of-attack control process, with the intermediate control variables being roll angle - roll rate - aileron deflection. A 0-degree sideslip angle hold strategy is adopted during UAV operation, requiring the sideslip angle to remain at 0 degrees throughout the entire flight. The intermediate control variables are sideslip angle - yaw rate - rudder deflection.

[0080] In addition to the aforementioned dynamic inverse UAV control model, the complete UAV model also includes a path tracking unit. This unit reads user-preset paths or paths planned by intelligent algorithms, stored as a two-dimensional dot matrix. The path tracking unit calculates the position vector between the UAV's current position and the tracked path, performs a cross product with the UAV's velocity vector to obtain the UAV's roll angle command, thus achieving continuous tracking of the predetermined path. Specifically, the path tracking unit employs a task planning strategy combining a Dubins curve-based path planning method with an improved particle swarm optimization (PSO) allocation scheme optimization method. This strategy calculates task planning schemes that satisfy the requirement of simultaneous arrival, providing planned paths for each UAV. Figure 2 As shown, the specific steps are as follows:

[0081] Step 1: Calculate the optimal Dubins path between all M UAVs and all K targets, and store the cost of each path in the cost matrix C, which has a dimension of M×K.

[0082] Specifically, the process involves: finding the left and right tangent circles of the starting and ending points based on the input aircraft position, speed, target position, optimal incident angle, minimum turning radius of the aircraft, and obstacle circle position; determining the intersection points of the centers of the starting and ending tangent circles with the obstacle circle to identify all tangent points; calculating the lengths of all Dubins curves under different combinations of tangent points, selecting the optimal path, and the cost is the path length; the optimal path includes the coordinates of all points, the type of connection between each point, and the curvature of the curve segment.

[0083] Step 2: Use the improved particle swarm optimization algorithm to search the cost matrix and obtain the sequence I with the minimum maximum cost. u ;

[0084] Step 3: When the number of iterations reaches the maximum iteration limit, determine whether the following inequality holds true:

[0085]

[0086] In the formula, min(I u ) represents sequence I u The shortest path in the equation, max(I u ) represents sequence I u The longest path in v min v represents the minimum flight speed of the drone. max This indicates the maximum flight speed of the drone;

[0087] When the inequality holds, such that in sequence I u The drone assigned the shortest path will circle in place one more time, shortening sequence I. uThe difference between the shortest and longest paths in the equation is used to determine the inequality again.

[0088] When the inequality does not hold, based on sequence I u Calculate the desired velocity and desired arrival time so that the UAV follows sequence I. u The corresponding path is tracked at the desired speed, specifically: in sequence I u The flight speed of the drone assigned to the longest path is set to the maximum flight speed v. max Then, the other drones adjust their speeds according to their own path lengths. The expected speed of the i-th drone is as follows:

[0089]

[0090] The expected arrival time is the same for each drone, and the expected arrival time is calculated as follows:

[0091]

[0092] like Figure 3 As shown, Step 2 involves searching the cost matrix using the improved particle swarm optimization algorithm as follows:

[0093] Step 21: Randomly generate initial particles and calculate their fitness values;

[0094] Specifically, without considering multiple drones attacking the same target during a single task allocation process, the allocation problem of multiple drones can be represented by matrix operations. Taking three drones attacking four targets as an example, the allocation sequence is shown in the table below.

[0095] Target I 0 0 1 0 Target II 1 0 0 0 Target III 0 1 0 0 Target IV 0 0 0 1

[0096] Set a feature information matrix P, which is a K*M matrix, where M is the number of drones and K is the number of targets. The feature information matrix P is P(i,j)={0&1}, where a value of 1 indicates that drone j attacks target i, and a value of 0 indicates that drone j does not attack target i.

[0097] As shown in the table above, the allocation result of the feature information matrix P is: UAV I attacks target II, UAV I attacks target III, UAV III attacks target I, and target IV is left empty. Performing column-to-column swapping operations on the matrix represents the two UAVs exchanging their attack targets. Randomly generated swap combinations are used to initialize the identity matrix E, generating initial particles.

[0098] Set a cost matrix C with dimension M*K, wherein the cost matrix C(j,i) = {l ji}, where l jiThe value of the attack target i by drone j is represented by the flight path length, which is calculated using the aforementioned Dubins method.

[0099] The cost matrix C and the feature information matrix P are further completed into a matrix of max(M,K)×max(M,K) dimensions. The completion method is to add a new row or column with all zeros after the last row or column.

[0100] Perform matrix operations to obtain the cost vector of the particle allocation results:

[0101] x i =diag(C×P)

[0102] In the formula, diag() represents the operation of extracting the diagonal elements of the matrix, C represents the cost matrix, and P represents the feature information matrix; x i The position of the i-th particle can be obtained by removing all zero elements, using max(x) i ) represents the fitness value of the particle.

[0103] Step 22: Compare the fitness values ​​of all particles with their historical best solutions. If the current fitness value is better, replace the historical best solution. When it is the initial state, directly use the current particle as the historical best solution. Take the historical best solution of each individual particle as the population's historical best solution.

[0104] Step 23: Perform a crossover operation between each particle and the best historical solution of the population, and select the offspring with the better fitness value.

[0105] Step 24: Perform cross-operation between each particle and its individual historical best solution, and select the offspring with the better fitness value.

[0106] Step 25: Each particle undergoes random mutation behavior, randomly swapping the attack targets among multiple groups of drones, directly updating the particle position without performing fitness value evaluation.

[0107] Step 26: When the maximum number of iterations N is reached... max Repeat the steps above.

[0108] The crossover operation is as follows: based on the characteristics of the UAV task allocation problem, the problem space dimension M is set to the number of UAVs, and a single particle represents the target allocation result of a group of UAVs. To address the difficulty in defining particle velocity, an operation is proposed to cross the current particle swarm with the individual historical best solution and the global historical best solution: 1) Randomly select a set of identical crossover regions from the two parent particles; 2) Exchange the attack targets of the corresponding UAVs within the crossover regions of the two parent particles, while updating the attack targets of other UAVs, eliminating duplicate elements, and obtaining two child individuals.

[0109] The advantages of the improved particle swarm optimization algorithm are as follows: Since the path planning optimization problem for multiple UAVs is a typical discrete NP-hard problem, exhaustive search is used to find the global optimum when the number of UAVs is small. However, as the dimensionality increases, the computational cost grows exponentially. When using the traditional PSO algorithm for searching, it is difficult to represent particle velocities in discrete space. Therefore, the improvement of this invention lies in proposing a crossover operation to replace particle velocities, and introducing particle mutation behavior to better help the swarm escape local optima.

[0110] like Figure 4 As shown, the process of converting the built UAV control model into C code using Simulink's code generation tool and embedding the UAV control model into the co-simulation platform through specified communication vectors includes: using the Simulink code generation tool Embedded Code-C++ to compile the built dynamic inverse six-degree-of-freedom control model into C++ code. The dynamic inverse six-degree-of-freedom control model has already been separated into user command input interfaces and UAV status output interfaces in Simulink. The input and output interfaces are represented as structures with variable names in the C++ code, and the structures are accessed through specific methods. The C++ code is divided into an initialization unit `initialize()`, an iteration unit `step()`, and a memory reclamation unit `terminate()`. The initialization unit initializes the memory locations and initial values ​​of time-related units (integral units, PID control units) in the six-degree-of-freedom control program. The iteration unit updates the next step's state data based on the previous step's state data and the input. The memory reclamation unit destroys all memory addresses and reclaims memory. The co-simulation platform has the ability to integrate a fully digital control model of the aircraft and drive the movement of the UAV model within the visual platform using the control model's output, and it has the ability for multiple UAVs to fly simultaneously within the visual platform.

[0111] The communication vector contains all UAV state information, including but not limited to: attitude angles in the launch frame, the position of the center of mass in the WCS coordinate system, the three-axis components of the UAV's ground speed, the UAV's angle of attack and sideslip angle, and the UAV's track tilt and track yaw angles. Attitude angles in the launch frame: used to store the UAV's attitude information, provided to the six-DOF unit in the next iteration step to update the UAV's attitude data. Also provided to the co-simulation platform to drive the attitude changes of the UAV model within it. Center of mass position: stores the UAV's position information, provided to the six-DOF unit in the next iteration step to update the UAV's position information. Also provided to the co-simulation platform to drive the position changes of the UAV model within it. Ground speed: same as above. Angle of attack, sideslip angle, track tilt angle, and track yaw angle: stored data, provided to the dynamic inverse controller in the next iteration step to calculate control surface command information. Also provided to the co-simulation platform as important UAV state data for display.

[0112] According to an embodiment of the present invention, the visualization simulation platform reads the UAV state information from the communication vector to drive the motion of the UAV's 3D digital model in the scene. Since Unreal Engine is a left-handed coordinate system, the UAV model, when outputting, converts the attitude angles under NED to those under Unreal Engine through coordinate transformation. Specifically, the transformation matrix from the NED geodetic coordinate system to the UAV body coordinate system is multiplied left by the coordinate transformation matrix from the NED coordinate system to the Unreal Engine's left-handed coordinate system, and then multiplied right by the coordinate transformation matrix from the Unreal Engine's left-handed coordinate system to the NED coordinate system. The resulting DCM coordinate transformation matrix is ​​then solved using Euler angle calculation to obtain the corresponding attitude angles in the Unreal Engine's left-handed coordinate system.

[0113] The high-fidelity modeling and visualization simulation integration method for aircraft motion described in this invention allows the UAV to automatically truncate itself by acquiring the user-defined initial position and velocity from the simulation platform at the start of the simulation. After the formal simulation, the UAV automatically maintains its own state information and sends its own data to the platform as called by the simulation platform. The truncation process is implemented by writing a MATLAB M-file, which is then converted to C++ code using a code generation tool. The initial latitude, longitude, and altitude of the UAV, along with its initial velocity in the WCS coordinate system, are input. A coordinate transformation matrix is ​​used to transform the position and velocity to the NED coordinate system. The M-file integrates an aerodynamic interpolation matrix, which can calculate the UAV's truncation angle of attack based on the current velocity information, thus determining the initial state.

[0114] The coordinate system used for internal data calculation in the six-degree-of-freedom control program of the UAV is the NED local coordinate system and its derived body coordinate system, track coordinate system, airflow coordinate system, etc. When interacting with the outside world, the six-degree-of-freedom program transforms the position and velocity information in the NED coordinate system to the WCS global coordinate system through a coordinate transformation matrix, thereby realizing parallel simulation of multiple UAVs in the simulation platform.

[0115] This invention can achieve high-fidelity simulation based on a high-fidelity six-degree-of-freedom model and display the simulation results in high quality through a rendering platform based on the Unreal Engine architecture. Figure 5 Simulation results of the path tracking process are presented, showing that the UAV flies according to the path given by the platform.

[0116] Another embodiment of the present invention proposes a high-fidelity modeling and visualization simulation integrated system for aircraft motion, the system comprising:

[0117] The 3D model building module is configured to build a 3D geometric model of the UAV based on the 3D geometric parameters of the UAV components.

[0118] The aerodynamic data acquisition module is configured to calculate aerodynamic data based on the three-dimensional geometric model of the UAV using computational fluid dynamics software. The aerodynamic data includes three-axis aerodynamic resultant force and resultant torque. The aerodynamic data is then extended into a five-dimensional aerodynamic query matrix through linear extrapolation. The five-dimensional aerodynamic query matrix includes angle of attack, sideslip angle, elevator deflection, rudder deflection, and aileron deflection.

[0119] The simulation control module is configured to embed the five-dimensional aerodynamic query matrix into a six-degree-of-freedom (DOF) dynamic inverse UAV control model to achieve closed-loop digital simulation of the UAV. The six-DOF dynamic inverse UAV control model includes a dynamic inverse controller, an aircraft mass control unit, a six-DOF dynamics unit, and an aerodynamic interpolation unit. The dynamic inverse controller calculates control surface commands based on the UAV's position and attitude data, and then obtains the simulated control surface angles through a first-order inertial element. The aerodynamic interpolation unit interpolates the control surface angles sent by the dynamic inverse controller based on the five-dimensional aerodynamic query matrix to obtain aerodynamic forces and moments at different state points, and outputs them to the six-DOF dynamics unit. The aircraft mass control unit acquires and outputs the aircraft's current mass and moment of inertia to the six-DOF dynamics unit. The six-DOF dynamics unit converts the aerodynamic forces and moments into accelerations and angular accelerations based on the aircraft's current mass and moment of inertia, and then integrates them to obtain the UAV state at the next moment.

[0120] In this embodiment, preferably, the process of obtaining aerodynamic data using computational fluid dynamics software in the aerodynamic data acquisition module includes: dividing the three-dimensional geometric model of the UAV into a mesh to generate an unstructured mesh; and using the Reynolds time-averaged method to solve the unstructured mesh to obtain the three-axis aerodynamic resultant force matrix and aerodynamic resultant moment matrix acting on the center of mass of the UAV.

[0121] In this embodiment, preferably, the dynamic inverse UAV control model further includes a path tracking unit. The path tracking unit is used to calculate the position vector between the current position of the UAV and the tracking path according to the preset tracking path, and perform a vector cross product between the position vector and the velocity vector of the UAV to obtain the roll angle command of the UAV, thereby realizing continuous tracking of the tracking target.

[0122] In this embodiment, preferably, the system further includes a visualization display module, which is configured to set communication vectors, embed the dynamic inverse UAV control model into the simulation platform, and perform visualization display through a virtual engine.

[0123] The functionality of the high-fidelity modeling and visualization simulation integrated system for aircraft motion in this embodiment of the invention can be described by the aforementioned high-fidelity modeling and visualization simulation integrated method for aircraft motion. Therefore, for the parts not described in detail in the system embodiment, please refer to the above method embodiment, and they will not be repeated here.

[0124] Although the invention has been described with respect to a limited number of embodiments, those skilled in the art will understand from the foregoing description that other embodiments are conceivable within the scope of the invention described herein. The disclosure of the invention is illustrative and not restrictive, and the scope of the invention is defined by the appended claims.

[0125] The references cited in this invention are as follows:

[0126] [1] Chen Jinyin, Yang Dongyong, Yu Shanqing, et al. Design and implementation of open experimental teaching for UAV simulation based on ROS [J]. Experimental Science and Technology, 2018, 16(01):40-43.

[0127] [2]Xian B, Zhao B, Zhang Y, et al.A low-cost hardware-in-the-loopsimulation testbed ofquadrotor UAVand implementation ofnonlinear controlschemes[J].Robotica,2017,35(3):588-612.

Claims

1. A method for integrating high-fidelity modeling and visualization simulation of aircraft motion, characterized in that, Includes the following steps: Construct a three-dimensional geometric model of the UAV based on the three-dimensional geometric parameters of its components; Based on the three-dimensional geometric model of the UAV, aerodynamic data are obtained by computational fluid dynamics software. The aerodynamic data is extended into a five-dimensional aerodynamic query matrix by linear extrapolation; the five-dimensional aerodynamic query matrix includes angle of attack, sideslip angle, elevator deflection, rudder deflection, and aileron deflection. The five-dimensional aerodynamic query matrix is ​​embedded into a six-degree-of-freedom dynamic inverse UAV control model to achieve closed-loop digital simulation of the UAV. The six-degree-of-freedom dynamic inverse UAV control model includes a dynamic inverse controller, an aircraft mass control unit, a six-degree-of-freedom dynamics unit, an aerodynamic interpolation unit, and a path tracking unit. The path tracking unit calculates the position vector between the UAV's current position and the tracking path based on a pre-set tracking path, and performs a cross product between this position vector and the UAV's velocity vector to obtain the UAV's roll angle command, thus achieving continuous tracking of the target. The pre-set tracking path is obtained using the following method: Step 1: Calculate the optimal Dubins path between all M UAVs and all K targets, and store the cost of each path in the cost matrix C; Step 2: Use the improved particle swarm optimization algorithm to search the cost matrix C and obtain the sequence with the minimum maximum cost. ; Step 3: When the number of iterations reaches the maximum iteration limit, determine whether the following inequality holds true: my( ) max( ); In the formula, min( ) represents a sequence The shortest path in, max( ) represents a sequence The longest path in This indicates the minimum flight speed of the drone. This indicates the maximum flight speed of the drone; When the inequality holds, such that in the sequence The drone assigned the shortest path will circle in place one more time to shorten the sequence. The difference between the shortest and longest paths in the equation is used to determine the inequality again. When the inequality does not hold, based on the sequence Calculate the desired speed and desired arrival time so that the drone follows the sequence. The corresponding path is tracked at the desired speed, specifically: in the sequence The flight speed of the drone assigned to the longest path is set to the maximum flight speed. Then, the other drones adjust their speeds according to their own path lengths. The expected speed of the i-th drone is as follows: ； The expected arrival time is the same for each drone, and the expected arrival time is calculated as follows: 。 2. The high-fidelity modeling and visualization simulation integration method for aircraft motion according to claim 1, characterized in that, The aerodynamic data includes the resultant force and resultant torque of the three-axis aerodynamics; The process of obtaining aerodynamic data using computational fluid dynamics software includes: meshing the three-dimensional geometric model of the UAV to generate an unstructured mesh; Based on an unstructured mesh, the Reynolds time-averaged method is used to solve for the three-axis aerodynamic resultant force matrix and aerodynamic resultant moment matrix acting on the UAV's center of mass.

3. The high-fidelity modeling and visualization simulation integration method for aircraft motion according to claim 1, characterized in that, The dynamic inverse controller is used to calculate control surface commands based on the UAV's position and attitude data, and then obtain simulated control surface angles through a first-order inertial element. The aerodynamic interpolation unit is used to interpolate the control surface angles sent by the dynamic inverse controller based on a five-dimensional aerodynamic query matrix to obtain aerodynamic forces and aerodynamic torques at different state points, and output them to the six-degree-of-freedom dynamics unit. The aircraft mass control unit is used to acquire and output the aircraft's current mass and moment of inertia to the six-degree-of-freedom dynamics unit. The six-degree-of-freedom dynamics unit is used to convert aerodynamic forces and aerodynamic torques into accelerations and angular accelerations based on the aircraft's current mass and moment of inertia, and then integrate them to obtain the UAV state at the next moment.

4. The high-fidelity modeling and visualization simulation integration method for aircraft motion according to claim 1, characterized in that, The process of searching the cost matrix C using the improved particle swarm optimization algorithm in the path tracing unit includes: Step 21: Randomly generate initial particles and calculate their fitness values; specifically: set the feature information matrix P, where P is K. An M-order matrix, where M represents the number of drones and K represents the number of targets; the feature information matrix P is... Where a value of 1 indicates that drone j attacks target i, and a value of 0 indicates that drone j does not attack target i; Performing column-to-column swaps on the matrix is ​​equivalent to two drones exchanging their attack targets; randomly generating swap combinations, and using these combinations to initialize the feature information matrix P, generating initial particles; the fitness value calculation process is as follows: The cost vector of the particle allocation result is calculated. : ； In the formula, diag represents the operation of extracting the diagonal elements of the matrix, and C represents the cost matrix; Cost vector The position of the i-th particle is obtained by removing all zero elements, using the cost vector. The maximum value is taken as the fitness value of the particle; Step 22: Compare the fitness values ​​of all particles with their historical best solutions. If the current fitness value is better, replace the historical best solution. When it is the initial state, directly use the current particle as the historical best solution. Take the historical best solution of each individual particle as the population's historical best solution. Step 23: Perform a crossover operation between each particle and the best historical solution of the population, and select the offspring with the better fitness value; Step 24: Perform cross-operation between each particle and its individual historical best solution, and select the offspring with the better fitness value; Step 25: Each particle undergoes random mutation behavior, randomly swapping the attack targets among multiple groups of drones, directly updating the particle position without performing fitness value evaluation. Step 26: Repeat the above steps before reaching the maximum number of iterations limit.

5. The high-fidelity modeling and visualization simulation integration method for aircraft motion according to claim 4, characterized in that, In the path tracing unit, the crossover operation during the search of the cost matrix C using the improved particle swarm optimization algorithm is as follows: randomly select a set of identical crossover regions from two parent particles; exchange the attack targets of the corresponding UAVs within the crossover regions of the two parent particles, and update the attack targets of other UAVs at the same time, eliminate duplicate elements, and obtain two child individuals.

6. The high-fidelity modeling and visualization simulation integration method for aircraft motion according to claim 1, characterized in that, The communication vector is set to embed the dynamic inverse UAV control model into the simulation platform and visualized through a virtual engine. The communication vector contains all UAV state information, including attitude angle, center of mass position, three-axis ground speed components, UAV angle of attack, sideslip angle, track tilt angle, and track deflection angle.

7. A high-fidelity modeling and visualization simulation integrated system for aircraft motion, characterized in that, The system is used to implement the high-fidelity modeling and visualization simulation integration method for aircraft motion as described in any one of claims 1-6; the system includes: The 3D model building module is configured to build a 3D geometric model of the UAV based on the 3D geometric parameters of the UAV components. The aerodynamic data acquisition module is configured to calculate aerodynamic data based on the three-dimensional geometric model of the UAV using computational fluid dynamics software. The aerodynamic data includes three-axis aerodynamic resultant force and resultant torque. The aerodynamic data is then extended into a five-dimensional aerodynamic query matrix through linear extrapolation. The five-dimensional aerodynamic query matrix includes angle of attack, sideslip angle, elevator deflection, rudder deflection, and aileron deflection. The simulation control module is configured to embed the five-dimensional aerodynamic query matrix into a six-degree-of-freedom (DOF) dynamic inverse UAV control model to achieve closed-loop digital simulation of the UAV. The six-DOF dynamic inverse UAV control model includes a dynamic inverse controller, an aircraft mass control unit, a six-DOF dynamics unit, and an aerodynamic interpolation unit. The dynamic inverse controller calculates control surface commands based on the UAV's position and attitude data, and then obtains the simulated control surface angles through a first-order inertial element. The aerodynamic interpolation unit interpolates the control surface angles sent by the dynamic inverse controller based on the five-dimensional aerodynamic query matrix to obtain aerodynamic forces and moments at different state points, and outputs them to the six-DOF dynamics unit. The aircraft mass control unit acquires and outputs the aircraft's current mass and moment of inertia to the six-DOF dynamics unit. The six-DOF dynamics unit converts the aerodynamic forces and moments into acceleration and angular acceleration based on the aircraft's current mass and moment of inertia, and then integrates them to obtain the UAV state at the next moment.

Images