General self-learning device and method for air-ground two-stage flight controller

Abstract

The application discloses a ground-air dual-stage flight controller general self-learning device and method. The device comprises an array camera visual attitude solution system, a surrounding wind field generator, a high-speed camera recording system, a ground four-degree-of-freedom aircraft support, an aircraft, a smoke flow guide pipe, a top truss, a flexible line and a multi-working-condition simulation nozzle. The surrounding wind field generator comprises three array wind field generating devices perpendicular to the ground, and the wind speed and direction can be defined. The aircraft is connected to a vertically telescopic support rod standing on a base through a three-degree-of-freedom spherical hinge, and is used for learning on the ground. An overhead umbilical assembly is arranged above the aircraft and connected with the flexible line of the truss, and is used for learning in the air. A flight controller is arranged at a mass point of the aircraft. The application can be used for controller self-learning design and verification of any configuration fixed-wing / variable fixed-wing aircraft in a real aerodynamic environment, and realizes intelligent flight controller attitude and position control under simulated multi-working-condition wind field conditions.

Description

Technical Field

[0001] The present invention relates to a universal self-learning device and method for a ground-to-air dual-stage flight controller, belonging to the fields of aircraft intelligent control, flight testing and experimental platforms. Background Art

[0002] The control and attitude optimization of aircraft are extremely challenging in complex flight environments, especially when faced with uncertain airflow conditions and complex tasks. The design and testing of flight controllers are particularly critical. Most existing flight tests rely on actual flights or wind tunnel experiments, which are costly and cannot run for a long time due to environmental conditions. At the same time, the verification and improvement of flight control simulation algorithms are difficult to fully implement in actual flights, especially the research on adaptive learning and reconstruction of controllers, which requires a more flexible and economical experimental platform. Based on this, the present invention proposes a universal self-learning device and experimental method for a ground-to-air dual-stage flight controller, which can combine the ground and air stages, and use a ground simulated wind field generating device, a movable bracket, a small fixed-wing UAV / scaled fixed-wing aircraft to achieve multi-degree-of-freedom attitude control and adaptive learning, providing a more feasible and flexible experimental scheme for the development of flight controllers. Summary of the Invention

[0003] In order to address the deficiencies of the prior art, the present invention aims to provide a universal self-learning device and method for a ground-to-air dual-stage flight controller.

[0004] The present invention is achieved through the following technical solutions:

[0005] A universal self-learning device for a ground-to-air dual-stage flight controller, comprising an array camera visual attitude calculation system, a surround wind field generator, a high-speed camera recording system, a ground four-degree-of-freedom aircraft bracket, an aircraft, a smoke flow guide tube, a top truss, a flexible cable, and a multi-condition simulation nozzle;

[0006] The surround wind field generator includes three array wind field generating devices perpendicular to the ground, which can customize the wind speed and direction;

[0007] A top truss is located above the array wind farm generator, equipped with multiple hanging points for installing high-speed cameras and multi-condition simulation nozzles; the ground four-degree-of-freedom aircraft bracket includes a three-degree-of-freedom spherical joint, a base, and a vertically retractable support rod; the bottom of the base is equipped with rollers;

[0008] The aircraft is connected to a vertically retractable support pole standing on a base through a three-degree-of-freedom spherical joint for learning in the ground phase;

[0009] A hanging umbilical cord assembly is installed above the aircraft, which is used to connect to the truss's soft line. The soft line provides external power supply and safe tethering to the aircraft during flight training. It has certain elasticity and is used for learning in the air.

[0010] A flight controller is installed at the mass point of the aircraft for autonomous flight control in wind field environments;

[0011] The array camera visual attitude solution system includes multiple infrared cameras, which use attitude algorithms to achieve attitude solution within the field of view and calculate the actual attitude angle and position of the aircraft;

[0012] The high-speed camera recording system includes three altitude cameras with front, top and left viewing angles, which are placed on the front, top and left side of the aircraft respectively, and are used to record relevant data during the aircraft learning process.

[0013] The array wind field generating device includes a smoke flow guide tube and multiple rotor propeller modules with independently adjustable wind speeds, which are used to generate steady / customized unsteady time-varying repeatable low-speed wind fields to simulate low-altitude flying air flow.

[0014] The array camera visual attitude solution system is composed of an array of infrared cameras surrounding the top of the surround wind farm generator. Each camera actively receives the incident light angle and time to collaboratively achieve attitude angle solution of the rigid body envelope within the field of view, and determines the relative position of the observed rigid body within the field of view based on the starting world coordinate calibration. It supports free assembly and addition and deletion of cameras to adjust the effective attitude solution range; it is fixed to the truss, which is fixed to the ground and has no direct connection with the surround wind farm generator.

[0015] The surround wind field generator includes an array of servo motors, each of which is independently driven by an electronic controller. The drive mode of all servo motors is PWM, the pulse modulation range is 1000-2000 ms, and the wind speed range is 0-52 m / s. The integrated control mode is web control, and network communication is via WiFi. It supports any device in the local area network to connect to the access control terminal, supports any custom wind field form and time and space sequence setting; supports the saving and import of custom wind field scripts; and any module unit supports modular splicing and supports free assembly into vertical, cylindrical, and trapezoidal surfaces; the smallest unit of a single array can be equipped with blades of different diameters as needed, with a diameter range of 20mm~180mm. The smallest units are connected in a mortise and tenon manner. A smoke flow guide tube is provided inside the unit to support controllable smoke flow output for flow field visualization.

[0016] The flight controller learns to fly by using an embedded algorithm based on a wind field environment that simulates a real air flow field when the precise aerodynamic model and dynamic model of the aircraft are unknown.

[0017] The top truss includes multiple hanging points, which can realize the installation of cameras, smoke nozzles, rain nozzles, flame nozzles, and mud and sand nozzles, and realize the simulation of multiple working conditions during flight.

[0018] The aircraft includes a set of flight controller hardware and an iterative adaptive learning algorithm. It learns and optimizes the optimal control law in real time by acquiring real flow field control response data during ground and air phases.

[0019] A method using a universal self-learning device for a ground-to-air dual-stage flight controller, wherein self-learning comprises two stages: ground and air. In the ground stage, basic attitude flight control laws are learned, and flight control law parameters under basic flight conditions are continuously learned in real data responses interacting with wind fields, and a stable control law is learned through online self-adaptation. Physical limitation of attitude angles is achieved through the action of spherical hinges, and telescopic support rods are remotely raised and lowered by a worm motor, thereby achieving attitude flight control learning of the aircraft in four degrees of freedom. In the air stage, the neighborhood performance optimal control law is learned to achieve optimal attitude control. Simultaneously, in combination with a high-speed camera recording system, the design and optimization of a position controller are achieved.

[0020] Beneficial effects of the present invention:

[0021] First, the actual aircraft flow field is directly simulated using a ground wind field generator, which greatly improves the experimental efficiency compared to traditional outdoor flight methods. Second, through an open annular wind field, it can support in-loop control test flights of real aircraft with a wingspan range of 0.3m-8m, greatly reducing the experimental cost of traditional low-altitude wind tunnels. Through the ground-air two-stage self-learning device and method, the full-process whole-machine controller design and optimization can be realized in an indoor environment, greatly improving the efficiency of controller design. In addition, the design also supports rain, smoke, mud, sand, and wind shear flow fields, reducing the dependence on experimental weather. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a system structure diagram of the universal self-learning device of the ground phase flight controller.

[0023] Figure 2 It is a system structure diagram of a universal self-learning device for an aerial phase flight controller.

[0024] Figure 3 It is a schematic diagram of the four-degree-of-freedom bracket structure during the ground learning phase.

[0025] Figure 4 It is a flow chart of the self-learning control method of the ground-to-air two-stage aircraft.

[0026] In the figure, the array camera attitude solution system 1, the surround wind field generator 2, the rotor propeller module 3, the high-speed camera recording system 4, the aircraft 5, the self-learning flight controller 6, the smoke duct 7, the top truss 8, the flexible wire 9, the three-degree-of-freedom spherical joint 10, the base 11, the vertically retractable support rod 12, and the multi-working condition simulation nozzle 13. DETAILED DESCRIPTION

[0027] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0028] like Figure 1 As shown, a universal self-learning device for a ground-to-air dual-stage flight controller includes a top array camera attitude solution system 1, a surround wind field generator 2, a rotor propeller module 3, a three-view high-speed camera recording system 4, an aircraft 5, a self-learning flight controller 6, a smoke duct 7, a top truss 8, a flexible wire 9, a three-degree-of-freedom spherical joint 10, a base 11, a vertically retractable support rod 12, and a multi-working condition simulation nozzle 13.

[0029] The surround wind generator 2 includes three array wind field generating devices perpendicular to the ground. Each array wind field consists of multiple rotor propeller modules 3 with independently adjustable wind speed and smoke ducts 7. The cluster control instructions of all modules are controlled by a web interface deployed in the cloud, which is used to generate steady / customized unsteady time-varying repeatable low-speed wind fields to simulate low-altitude flying air flow.

[0030] The three-view high-speed camera recording system 4 includes three height cameras with front, top and left viewing angles, which are respectively placed on the front, top and left sides of the aircraft to record relevant data during the aircraft learning process.

[0031] The aircraft 5 is fixedly connected to the end of the soft wire 9 hanging down from the top truss 8 (in the air stage), or the aircraft 5 is fixedly connected to the vertically retractable support rod 12 on the base 11 through a three-degree-of-freedom spherical joint 10 (on the ground stage), which is used to realize the virtual flight tethered test during the aircraft control law attitude / unknown learning process. During the learning process, the power unit and all servo actuators are controlled in real time in the loop.

[0032] The self-learning flight controller 6 is installed at the center of mass of the aircraft and can be used for autonomous flight control in a wind field environment.

[0033] The top truss 8 is placed above the wind field generator, has multiple hanging points, and can install the top-view high-speed camera 4 and the multi-working-condition simulation nozzle 13 to simulate the flight working condition of a complex scene.

[0034] The bottom of the base 11 is provided with a folding roller to realize movement in a horizontal plane and adjust the distance between the aircraft and the annular wind field generator.

[0035] As shown in Figure 1 The ground learning stage of the ground-air two-stage flight controller universal self-learning device system adopts the layout of the surrounding wind field generator, and the arbitrary real or scaled model aircraft 5 is fixed to the telescopic connecting rod 12 through the spherical hinge 10, the connecting rod is placed on the movable base 11, and the self-learning flight controller is installed at the center of mass of the aircraft 5. The aircraft power supply and the wind field generator are started, the aircraft enters the iterative adaptive learning process, the adaptive neural network parameters are learned through continuous learning iteration in multiple types of wind fields, the control law is reconstructed in real time, the performance is evaluated through the pose positioning system, the stable attitude flight control in the ground stage is realized.

[0036] As shown in Figure 2 The air learning stage of the ground-air two-stage flight controller universal self-learning device system also adopts the layout of the surrounding wind field generator, and the arbitrary real or scaled model aircraft 5 is fixed to the telescopic soft rope 9 through the truss 8, and the self-learning flight controller is installed at the center of mass of the aircraft 5. The aircraft power supply and the wind field generator are started, the aircraft enters the iterative adaptive air learning process, the adaptive neural network parameters are learned through continuous learning iteration in multiple types of wind fields, the control law is reconstructed in real time, the performance is evaluated through the pose positioning system, the attitude flight control optimization and the position controller design optimization in the air stage are realized.

[0037] As shown in Figure 3 The movable support includes the spherical hinge 10, the base 11, and the telescopic support rod 12, wherein the upper end groove of the spherical hinge is matched with the abdomen, the bottom spherical surface is engaged with the groove surface of the telescopic rod 12, the built-in ball layer reduces the sliding friction, the flexible rotation of pitch ±50°, roll ±50°, and heading 360° is realized. The rigid telescopic rod is connected with the ground base.

[0038] As shown in Figure 4 The application process of the ground-air two-stage self-learning control device and method in the controller whole-process design is shown, wherein the second-stage ground learning and the third-stage air learning correspond to the attitude learning device and method in the ground support wind field environment and the attitude control optimization and position controller design device and method in the air hanging soft rope wind field environment.

[0039] The embodiments described above may be further combined or replaced, and the embodiments are merely descriptions of preferred embodiments of the present invention and do not limit the concept and scope of the present invention. Various changes and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the design concept of the present invention are within the scope of protection of the present invention. The scope of protection of the present invention is given by the appended claims and any equivalents thereof.

Claims

1. A universal self-learning device for a ground-to-air dual-stage flight controller, characterized by: It includes an array camera visual attitude solving system (1), a surround wind field generator (2), a high-speed camera recording system (4), a ground four-degree-of-freedom aircraft bracket, an aircraft (5), a smoke flow guide tube (7), a top truss (8), a flexible wire (9), and a multi-working condition simulation nozzle (13); The surround wind field generator (2) includes three array wind field generating devices perpendicular to the ground, and the wind speed and wind direction can be customized; A top truss (8) is provided above the array wind field generating device, and has multiple hanging points for installing a high-speed camera and a multi-working condition simulation nozzle (13); The ground four-degree-of-freedom aircraft bracket comprises a three-degree-of-freedom spherical joint (10), a base (11), and a vertically retractable support rod (12); a roller is installed at the bottom of the base (11); The aircraft (5) is connected to a vertically retractable support rod (12) standing on a base (11) through a three-degree-of-freedom spherical joint (10) for learning in the ground phase; A hanging umbilical cord assembly is provided above the aircraft (5) for connecting to a flexible cord (9) of the truss (8). The flexible cord (9) provides external power supply and secure mooring to the aircraft being learned to fly, and has a retractable characteristic for learning in the air. A flight controller (6) is provided at a mass point of the aircraft (5) for autonomous flight control in a wind field environment; The array camera visual attitude solution system (1) includes multiple infrared cameras, which realizes attitude solution within the field of view through attitude algorithm, and is used to calculate the actual attitude angle and position of the aircraft; The high-speed camera recording system (4) includes three height cameras with front, upper and left viewing angles, which are respectively placed on the front, upper and left sides of the aircraft to record relevant data during the aircraft learning process; The array wind field generating device comprises a smoke flow guide tube (7) and a plurality of rotor propeller modules (3) capable of independently adjusting wind speed, and is used to generate a steady / customized unsteady time-varying repeatable low-speed wind field to simulate the incoming air flow of low-altitude flight; The array camera visual attitude solution system (1) is composed of an array of infrared cameras surrounding the top of the surround wind field generator (2). Each camera actively receives the incident light angle and time to achieve attitude angle solution of the rigid body envelope within the field of view, and determines the relative position of the observed rigid body within the field of view based on the initial world coordinate calibration. It supports free assembly and addition and deletion of cameras to adjust the effective attitude solution range; it is fixed to the truss (8), the truss is fixed to the ground, and has no direct connection with the surround wind field generator (2).

2. The universal self-learning device for a ground-to-air dual-stage flight controller according to claim 1, characterized in that: The surround wind field generator (2) includes array servo motors, each motor is independently driven by an electric controller, all servo motors are driven in PWM mode, the pulse modulation range is 1000-2000 ms, the wind speed range is 0-52 m / s, the integrated control mode is web control, network communication is via wifi, support any device in the local area network to connect to the access control terminal, support any custom wind field form and time and space sequence setting; support custom wind field script saving and import; and any module unit supports modular splicing, support free assembly into vertical, cylindrical, trapezoidal surfaces; a single array minimum unit can be equipped with blades of different diameters according to needs, with a diameter range of 20mm~180mm, the minimum units are connected in a mortise and tenon manner, and a smoke flow guide tube is provided inside the unit to support controllable smoke flow output for flow field visualization.

3. The universal self-learning device for a ground-to-air dual-stage flight controller according to claim 1, characterized in that: The flight controller (6) learns to fly by using an embedded algorithm based on a wind field environment that simulates a real air flow field when the precise aerodynamic model and dynamic model of the aircraft are unknown.

4. The universal self-learning device for a ground-to-air dual-stage flight controller according to claim 1, characterized in that: The multi-working condition simulation nozzle (13) includes a smoke nozzle, a rain nozzle, a flame nozzle, and a mud and sand nozzle, which realizes the condition simulation of multiple working conditions during the flight process.

5. The universal self-learning device for a ground-to-air dual-stage flight controller according to claim 1, characterized in that: The aircraft (5) includes a set of flight controller hardware and an iterative adaptive learning algorithm; the optimal control law is learned in real time by acquiring real flow field control response data during the ground and air phases.

6. A method using the universal self-learning device for a ground-to-air dual-stage flight controller according to claim 1, characterized in that: Self-learning consists of two phases: ground and air. The ground phase involves learning the basic attitude flight control law. The flight control law parameters under basic flight conditions are continuously learned from real data responses interacting with the wind field. A stable control law is learned through online adaptive learning. The physical limitation of attitude angle is achieved through the action of spherical hinges, and the telescopic support rod is remotely raised and lowered by a worm motor, thereby realizing attitude flight control learning of the aircraft in four degrees of freedom; in the air phase, the optimal control law of neighborhood performance is learned to achieve optimal attitude control. At the same time, combined with the high-speed camera recording system, the design and optimization of the position controller is realized.

Images