Operation control method of aircraft-unmanned tractor

Abstract

The invention relates to an operation control method of an aircraft-unmanned tractor, belongs to the technical field of automatic control of unmanned tractors, and solves the problem that an unmanned tractor cannot run according to a central line of a planned path in the prior art. Acquiring an image of a to-be-dragged aircraft by adopting a camera, acquiring point cloud data of the to-be-dragged aircraft and the distance between the unmanned tractor and the to-be-dragged aircraft by adopting a laser radar, and acquiring a deflection angle of front wheels of the to-be-dragged aircraft based on the image and the point cloud data; the clamping operation of the unmanned tractor is executed based on the deflection angle and the distance, and after the clamping operation is completed, the operation path of the unmanned tractor is obtained; and selecting a vehicle control model corresponding to the model of the aircraft to be dragged from the vehicle control model library, and controlling the unmanned tractor to implement a traction task along the operation path based on the vehicle control model. According to the technical scheme provided by the invention, the operation accuracy, stability and adaptability of the unmanned tractor and the operation control efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology for unmanned tractors, and more particularly to an operation control method for an aircraft-unmanned tractor. Background Technology

[0002] Aircraft towing vehicles are an indispensable piece of ground support equipment at modern airports. Using aircraft towing vehicles to move aircraft saves fuel, reduces noise and exhaust pollution, and lowers engine wear. Traditional manned aircraft towing vehicles suffer from high labor costs and significant operational risks, severely hindering airport transfer efficiency and posing potential threats to flight safety. With the rapid development of unmanned aerial vehicle (UAV) technology based on sensor technology and automatic control, the safety and efficiency of unmanned equipment have gradually improved.

[0003] Currently, unmanned vehicles have the capability to complete pre-set tasks well under conditions of "manned monitoring and unmanned intervention," which provides a clear direction for the technological upgrade of aircraft towing vehicles. Through environmental fusion perception, autonomous task planning, and automatic control, unmanned aircraft towing vehicles can autonomously complete the entire process of aircraft towing operations according to the task requirements.

[0004] After acquiring the planned path, the unmanned towing vehicle will tow the aircraft to the designated location. Since most of the existing unmanned towing vehicle control methods are based on theoretical design or experience control, the existing unmanned towing vehicles cannot completely follow the planned path as the central path. Therefore, in actual operation, airport personnel need to constantly monitor the central control panel to prevent large lateral movement errors of each unmanned towing vehicle. Summary of the Invention

[0005] Based on the above analysis, the present invention aims to provide an operation control method for an aircraft-unmanned tractor, in order to solve the problem that existing unmanned tractors cannot travel along the planned path centerline due to large lateral travel errors.

[0006] This invention provides an operational control method for an aircraft-unmanned tractor, comprising:

[0007] S1: Use a camera to acquire images of the aircraft to be towed, use a lidar to acquire point cloud data of the aircraft to be towed and the distance between the unmanned towing vehicle and the front wheel of the aircraft to be towed, and obtain the deflection angle of the front wheel of the aircraft to be towed based on the images and the point cloud data.

[0008] S2: Perform the unmanned tractor clamping operation based on the deflection angle and the distance. After the clamping operation is completed, obtain the working path of the unmanned tractor.

[0009] S3: Select a vehicle control model corresponding to the aircraft model to be towed from the vehicle control model library, and control the unmanned towing vehicle to carry out the towing task along the operation path based on the vehicle control model;

[0010] The vehicle control model library stores vehicle control models corresponding to various types of aircraft. Each vehicle control model is built based on the PID-FF algorithm and is used for the lateral control of unmanned tractor vehicles. The control parameters of each vehicle control model are different, including the proportional gain coefficient, integral time parameter, derivative time parameter, and feedforward parameter for lateral control.

[0011] Based on a further improvement of the above method, obtaining the deflection angle of the front wheel of the aircraft to be towed based on the image information and the point cloud data includes:

[0012] S11: Identify the target front wheel in the image based on an image detection algorithm;

[0013] S12: Filter and downsample the point cloud data to obtain processed point cloud data;

[0014] S13: Based on the target front wheel, extract the target point cloud data corresponding to the front wheel from the processed point cloud data;

[0015] S14: Based on the comparison between the standard aircraft front wheel point cloud data and the target point cloud data, the deflection angle of the front wheel of the aircraft to be towed is obtained, and the deflection angle is used as the traction angle of the unmanned towing vehicle.

[0016] Based on a further improvement to the above method, the distance between the unmanned towing vehicle and the aircraft to be towed is calculated from the front wheel radius of the aircraft to be towed, the range of the lidar, and the relative distance between the lidar and the clamping device of the unmanned towing vehicle; specifically as follows:

[0017]

[0018] Where L is the distance between the unmanned tractor and the front wheel of the aircraft to be towed, l is the minimum distance measured by the lidar, r is the radius of the front wheel, α is the deflection angle of the front wheel of the aircraft to be towed, and s is the relative distance between the lidar and the clamping device of the unmanned tractor.

[0019] Based on a further improvement of the above method, the PID-FF algorithm is as follows:

[0020]

[0021] Where, δ n K is the steering angle of the front axle tires of the unmanned tractor. Py The proportional gain parameter for lateral control; T IyT is the integral time parameter for lateral control. Dy K is the derivative time parameter for lateral control. ffy For feedforward parameters of lateral control; e n e is the minimum lateral distance between the aircraft's nose landing gear and its flight path; θ denoted as , where is the angle between the heading angle of the unmanned tractor and the tangent at the pre-aiming distance s; is the path segment corresponding to the current driving route, n = 1, 2, 3, ..., T.

[0022] Based on further improvements to the above method, a vehicle control model for each aircraft model is established in the following manner, thereby forming the vehicle control model library:

[0023] A simulation model of an aircraft-unmanned tractor was built based on Gazeboll.

[0024] Based on historical vehicle control data, the value ranges of various control parameters in the PID-FF algorithm are set, including K. py T Iy T Dy K ffy And randomly generate multiple sets of simulation control parameters to be tested according to the value range;

[0025] Based on each set of simulated control parameters to be tested, the PID-FF algorithm is used to perform control simulation to obtain the first lateral distance error under the set of simulated control parameters to be tested. If the first lateral distance error is less than the first preset threshold, the corresponding simulated control parameters to be tested are used as the measured control parameters.

[0026] Select an actual driving route, and conduct on-site PID-FF algorithm control tests based on each set of measured control parameters to obtain the second lateral route error corresponding to that set of measured control parameters. If the second lateral route error is less than the second preset threshold, the corresponding measured control parameters are used as a vehicle control model corresponding to the aircraft model, thereby forming a vehicle control model library.

[0027] Based on a further improvement of the above method, the step of using the PID-FF algorithm to perform control simulation based on each set of simulated control parameters to obtain the first lateral distance error under that set of simulated control parameters includes:

[0028] A1: Obtain the simulation path and divide it into N path segments, N = 1, 2, 3, ..., N. Initially, the unmanned tractor is located at the starting point of the simulation path. The target position is the corresponding point on the simulation path with the distance from the unmanned tractor as the pre-aiming distance S. Initially, the lateral distance between the front landing gear of the aircraft to be towed and the path segment is 0, and the steering angle of the front axle tires of the unmanned tractor is 0.

[0029] A2: For each path segment, based on the heading angle θ NMinimum lateral distance e N The steering angle of the current path segment is calculated using the PID-FF algorithm;

[0030] Wherein, the heading angle θ N Based on the steering angle δ of the previous path segment N-1 The minimum lateral distance e is obtained. N It is the minimum lateral distance between the nose landing gear of the aircraft to be towed and the current path segment;

[0031] A3: After the driving is completed, the average value of the minimum lateral distance corresponding to each path segment is taken as the first lateral distance error corresponding to the current traction simulation control parameter.

[0032] Based on a further improvement of the above method, the step of selecting the actual driving route and performing on-site PID-FF algorithm control tests based on each set of measured control parameters to obtain the second lateral route error corresponding to that set of measured control parameters includes: performing the following operations for each set of measured control parameters:

[0033] B1: Divide the measured path into M path segments, M = 1, 2, 3, ..., M. Initially, the unmanned towing vehicle is located at the starting point of the measured path. Obtain the pre-aiming distance S′. Initially, the lateral distance between the front landing gear of the aircraft to be towed and the path segment is 0, and the steering angle of the front axle tires of the unmanned towing vehicle is 0.

[0034] B2: Based on the positioning information of the unmanned tractor, calculate the minimum lateral distance RN between the nose landing gear of the aircraft to be towed and the measured path segment corresponding to the current path segment M, and the angle SN between the heading angle of the unmanned tractor and the tangent at the pre-aiming distance S′. Based on RN, SN and the set of measured control parameters, obtain the steering angle δ of the unmanned tractor through the PID-FF algorithm, and calculate the difference between δ and the standard steering angle corresponding to the current path segment. Take the largest steering angle difference of the set of measured control parameters in each measured path segment as the first measured error.

[0035] B3: Based on the positioning information, obtain the actual driving route of the aircraft to be towed, and use the maximum difference in the lateral distance between the actual driving route and the measured path as the second measured error;

[0036] B4: The second lateral route error is obtained based on the first measured error and the second measured error.

[0037] Based on a further improvement to the above method, if no corresponding vehicle control model is obtained after on-site testing of all measured control parameters, the measured control parameters are fine-tuned to obtain the vehicle control model, including:

[0038] Sort all measured control parameters in ascending order of the second lateral route error, and select at least two sets of measured control parameters with the smallest error as target control parameters for fine-tuning.

[0039] Obtain the first and second measured errors corresponding to the target control parameters, and determine whether the quotient of the first and second measured errors is greater than 1. If it is greater than 1, then adjust the parameter K in the PID+FF algorithm. ffy Fine-tune the parameters; otherwise, adjust the parameter K in the PID+FF algorithm. Py Make minor adjustments.

[0040] Based on further improvements to the above method, for K ffy Adjustment is made using the first preset step size value A for K. Py The adjustment is made with the second preset step size value B, where A and B are both constants;

[0041] The K ffy The fine-tuning process includes:

[0042] D1: With K ffy +A reconstructs the target control parameters and conducts on-site testing. It is determined whether the second lateral route error corresponding to the reconstructed target control parameters is less than the second preset threshold. If it is less, the corresponding target control parameters are used as the vehicle control model. If it is greater, D2 is executed.

[0043] D2: With K ffy -A Reconstruct the target control parameters and conduct on-site testing. Determine whether the second lateral route error corresponding to the reconstructed target control parameters is less than the second preset threshold. If it is less, use the corresponding target control parameters as the vehicle control model. If it is greater, continue to execute D3.

[0044] D3: If K ffy The error of the second lateral route corresponding to +A is greater than K. ffy The second lateral route error corresponding to -A is then determined to be a reduction, and K is used as the adjustment direction. ffy -2A reconstructs the target control parameters and performs field testing; otherwise, the fine-tuning direction is determined to be increasing, and K is used as the benchmark. ffy +2A reconstructed the target control parameters and conducted field tests;

[0045] D4: If the second lateral route error corresponding to the reconstructed target control parameters is less than the second preset threshold, then the corresponding target control parameters are used as the vehicle control model. If it is greater, then continue to construct the target control parameters along the fine-tuning direction determined by D3 for on-site testing until the second lateral route error is less than the second preset threshold. Then, the target control parameters at this time are used as the vehicle control model.

[0046] The KPy The fine-tuning process and the K ffy The fine-tuning process is the same.

[0047] Based on a further improvement of the above method, the heading angle θ N Based on the steering angle δ of the previous path segment N-1 Specifically, the results are as follows:

[0048]

[0049] θ N =a t2 *T,

[0050] Among them, F xf and F yf These are the longitudinal force and the lateral force of the front wheel, respectively. tf l is the distance between the center of gravity and the front axle of the unmanned tractor. n F is the distance between the center of gravity of the unmanned tractor and the articulation point of the vehicle and the machine. yr For the lateral force of the rear wheel, l tr Let a be the distance between the center of gravity and the rear axle of the unmanned tractor. ty For the lateral acceleration of the unmanned tractor, I t Let be the moment of inertia of the unmanned tractor, m be the mass of the unmanned tractor, δ be the steering angle of the unmanned tractor, and a be the moment of inertia. t2 Let T be the angular acceleration of the unmanned tractor, and T be the current travel time.

[0051] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0052] 1. This invention provides an operation control method for an aircraft-unmanned tractor. Based on the PID-FF algorithm, experimental simulation method, and field measurement method, vehicle control models for various types of aircraft are finally obtained. Using the vehicle control model can ensure that the aircraft travels along the centerline of the planned route during the traction process, which greatly reduces lateral error, improves the accuracy of unmanned tractor operation, and ensures safe operation. By pre-storing the vehicle control model corresponding to each aircraft, the efficiency of unmanned tractor operation control is further improved.

[0053] 2. This invention provides an operation control method for an aircraft-unmanned tractor. First, the theoretically optimal control parameters are obtained by experimental simulation. However, due to factors such as transmission loss between the steering wheel and the unmanned tractor, road wear, and environment during actual traction, the unmanned tractor cannot perfectly follow the centerline of the planned route according to the theoretical parameters. Therefore, the vehicle control model obtained by the simulation method is tested on-site to further improve the accuracy, stability, and adaptability of the vehicle control model.

[0054] 3. This invention provides an operation control method for an aircraft-unmanned tractor. When a vehicle control model corresponding to the aircraft is not obtained, the possibility of PID-FF algorithm control parameters is further explored through a fine-tuning process to obtain the optimal vehicle control model that conforms to the actual operating environment.

[0055] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0056] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0057] Figure 1 This is a schematic diagram of an operation control method for an aircraft-unmanned tractor.

[0058] Figure 2 This is a schematic diagram showing the angle between the heading angle of the unmanned tractor and the tangent at the pre-aiming distance s. Detailed Implementation

[0059] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0060] A specific embodiment of the present invention discloses an operation control method for an aircraft-unmanned tractor, such as... Figure 1 As shown, it includes:

[0061] S1: Use a camera to acquire images of the aircraft to be towed, use a lidar to acquire point cloud data of the aircraft to be towed and the distance between the unmanned towing vehicle and the front wheels of the aircraft to be towed, and obtain the deflection angle of the front wheels of the aircraft to be towed based on the images and the point cloud data.

[0062] The distance between the unmanned towing vehicle's parking position and the front wheel of the aircraft to be towed is no more than 5m in the longitudinal direction, and the angular deviation between the unmanned towing vehicle and the front wheel of the aircraft is no more than 10°.

[0063] The camera can be selected based on actual needs, with the standard being the ability to capture clear images. A 48-line LiDAR can be selected.

[0064] Specifically, obtaining the deflection angle of the front wheel of the aircraft to be towed based on the image information and the point cloud data includes:

[0065] S11: Identify the target front wheel in the image based on an image detection algorithm.

[0066] The image detection algorithms include YOLO, R-CNN, Fast R-CNN, SSD, NanoDet, etc. For example, an aircraft nose wheel image dataset can be pre-built to train various image detection algorithms, and the detection algorithm with the best recognition performance can be used as the final image detection algorithm.

[0067] S12: Filter and downsample the point cloud data to obtain processed point cloud data.

[0068] Point cloud filtering can effectively remove noise points and redundant information, improving data quality. Common filtering techniques include voxel grid filtering, statistical filtering, conditional filtering, and radius filtering. Point cloud downsampling techniques aim to reduce data volume and improve processing efficiency. Common downsampling methods include random sampling, uniform sampling, and nearest neighbor sampling.

[0069] The process of point cloud processing using the above-mentioned techniques in this invention all adopts existing technologies and will not be described in detail.

[0070] S13: Extract the target point cloud data corresponding to the front wheel from the processed point cloud data based on the target front wheel.

[0071] S14: Based on the comparison between the standard aircraft front wheel point cloud data and the target point cloud data, the deflection angle of the front wheel of the aircraft to be towed is obtained, and the deflection angle is used as the traction angle of the unmanned towing vehicle.

[0072] Specifically, the standard aircraft nose wheel point cloud data is the point cloud data obtained by lidar when the nose wheel of this type of aircraft is not rotating.

[0073] The distance between the unmanned tractor and the aircraft to be towed is calculated from the radius of the aircraft's front wheels, the range of the lidar, and the relative distance between the lidar and the tractor's clamping device. For example, if the lidar is installed at the center of the tractor, and this center is on the same vertical line as the tractor's center of gravity, then the distance L between the tractor and the aircraft to be towed is...

[0074]

[0075] Where 1 is the minimum distance for lidar ranging, r is the front wheel radius, α is the deflection angle of the front wheel of the aircraft to be towed, and s is the relative distance between the lidar and the clamping device of the unmanned towing vehicle.

[0076] S2: Perform the unmanned tractor clamping operation based on the deflection angle and the distance. After the clamping operation is completed, obtain the working path of the unmanned tractor.

[0077] The operational path is calculated by the cloud control platform and transmitted wirelessly to the unmanned tractor. The cloud control platform, acting as the "brain" of the tractor system, mainly consists of remote control ground station hardware and the command and control software embedded within it. The cloud control platform is primarily used for global path planning (Dijkstra's algorithm), local path planning (TEB) algorithm, task planning and distribution, and individual vehicle remote control. It transmits this information to the individual tractor via a communication data link. In addition, the cloud control platform also receives real-time pose information, task information, and vehicle control data reported by the individual tractor.

[0078] The unmanned tractor's communication terminal connects to the VPN server and establishes a static route pointing to the remote control station. The addresses of the connected data acquisition and control devices (gateways) and control devices (host computers) are consistent with the internal network addresses of their respective communication terminals. The control devices (host computers) and one or more data acquisition and control devices (gateways) establish IP connections using VPN private network addresses, thereby realizing the communication data link between the tractor and the cloud control platform.

[0079] S3: Select a vehicle control model from the vehicle control model library that corresponds to the model of the aircraft to be towed, and control the unmanned towing vehicle to carry out the towing task along the operation path based on the vehicle control model.

[0080] The vehicle control model library stores vehicle control models corresponding to various types of aircraft. Each vehicle control model is built based on the PID-FF algorithm and is used for the lateral control of unmanned tractor vehicles. The control parameters of each vehicle control model are different, including the proportional gain coefficient, integral time parameter, derivative time parameter, and feedforward parameter for lateral control.

[0081] Because the front wheel steering angle of the unmanned tractor in aircraft towing projects is relatively small, this invention approximates the front wheel steering angle of the tractor to be equal to the vehicle's turning angle and the aircraft's front wheel steering angle. In this case, the tractor and aircraft assembly can be considered as a single unit with power and steering capabilities in the front wheels, thus yielding the PID-FF algorithm:

[0082]

[0083] Where, δ n K is the steering angle of the front axle tires of the unmanned tractor. Py The proportional gain parameter for lateral control; T Iy T is the integral time parameter for lateral control. Dy K is the derivative time parameter for lateral control. ffy For feedforward parameters of lateral control; e ne is the minimum lateral distance between the aircraft's nose landing gear and its flight path; θ Let be the angle between the heading angle of the unmanned towing vehicle and the tangent at the pre-aiming distance s, such as... Figure 2 As shown; n is the path segment corresponding to the current driving route, n = 1, 2, 3, ..., T.

[0084] The vehicle control model library is formed by establishing a vehicle control model for each aircraft model in the following manner:

[0085] Step 1: Build an aircraft-unmanned tractor simulation model based on Gazebo11.

[0086] This invention utilizes Gazebo11 to establish aircraft models, unmanned tractor models, and simulation path models. Gazebo11 is an open-source robot simulation software that provides a highly realistic 3D simulation environment for testing and validating robot algorithms. Key features of Gazebo11 include, but are not limited to: Physics Simulation: Gazebo11 uses an advanced physics engine to simulate the interaction between the robot and its environment, providing accurate collision detection and dynamic behavior. Sensor Simulation: Supports simulation of various sensor models, such as cameras, LiDAR, and inertial measurement units (IMUs), which can be used to test sensor data fusion and processing algorithms. Plugin System: Gazebo11 has a flexible plugin system, allowing developers to extend the simulator's functionality, such as adding new robot models, sensor models, or custom environments. Model Library: Provides a rich library of robot models and environments; users can choose from built-in models or import their own models for simulation. ROS Integration: Tightly integrated with ROS, supporting ROS message passing, service calls, and parameter servers, facilitating the development and testing of robot algorithms. Visual Interface: Provides an intuitive user interface, allowing users to control the simulation process and observe simulation results through graphical operations.

[0087] Of course, other simulation software such as Adams, Simpack, and Simulink can also be selected, and this invention does not impose any restrictions on them.

[0088] Step 2: Based on historical vehicle control data, set the value range of each control parameter in the PID-FF algorithm, including K. Py T Iy T Dy K ffy And, based on the value range, randomly generate multiple sets of simulation control parameters to be tested.

[0089] Specifically, a historical vehicle control dataset is obtained, and the range of each control parameter in the dataset is statistically analyzed. A clustering algorithm can be used to cluster any control parameter, and the cluster center is obtained. Based on experience, values ​​that are far from the cluster center are removed. The purpose is to improve the accuracy of the simulation experiment because values ​​that are far away have a lower probability of appearing, while points that are close to the cluster center are relatively more stable. Therefore, the range of the retained control parameters is statistically analyzed to form the value range of each control parameter. Table 1 lists the value range of each control parameter.

[0090] Table 1. Value ranges of each control parameter

[0091] Serial Number parameter Range of values Remark 1 <![CDATA[T Iy ]]> 0-∞ 2 <![CDATA[T Dy ]]> 0-0.2 3 <![CDATA[K Py ]]> 0-1 4 <![CDATA[K ffy ]]> 0-1

[0092] Clustering algorithms can include K-Means clustering, hierarchical clustering, DBSCAN clustering, Gaussian mixture model (GMM), etc., and this invention does not limit them.

[0093] Based on the value range of the above control parameters, at least 2000 sets of control parameters should be generated using a random algorithm. The random algorithm can be linear congruential method, Mersenne Twister, XORshift, random number tables, cryptographic hash algorithms, etc.

[0094] Understandably, this invention primarily studies the lateral control of unmanned tractor vehicles. For longitudinal control parameters, such as speed and acceleration, this invention assumes that the longitudinal control parameters remain consistent across all simulation experiments and field measurements, but all must ensure that the speed is less than or equal to 10 km / h and the acceleration is greater than or equal to 0.13 m / s². 2 Less than or equal to 1 m / s 2 .

[0095] Step 3: Based on each set of simulated control parameters to be tested, use the PID-FF algorithm to perform control simulation to obtain the first lateral distance error under that set of simulated control parameters to be tested. If the first lateral distance error is less than the first preset threshold, then the corresponding simulated control parameters to be tested are used as the measured control parameters.

[0096] The first preset threshold can be set according to the control accuracy and the road width. For example, the first preset threshold can be set to 0.3m.

[0097] Based on each set of simulated control parameters to be tested, the PID-FF algorithm is used for control simulation to obtain the first lateral distance error under that set of simulated control parameters, including:

[0098] A1: Obtain the simulation path and divide it into N path segments, N = 1, 2, 3, ..., N. Initially, the unmanned towing vehicle is located at the starting point of the simulation path. The target position is the point on the simulation path at a distance S from the unmanned towing vehicle. Initially, the lateral distance between the nose landing gear of the aircraft to be towed and the path segment is 0, and the steering angle of the front axle tires of the unmanned towing vehicle is 0. For example, the simulation path can be divided into N path segments to facilitate subsequent calculations.

[0099] A2: For each path segment, based on the heading angle θ N Minimum lateral distance e N The steering angle of the current path segment is calculated using the PID-FF algorithm.

[0100] Wherein, the heading angle θ N Based on the steering angle δ of the previous path segment N-1 Specifically, the results are as follows:

[0101]

[0102] θ N =a t2 *T,

[0103] Among them, F xf and F yf These are the longitudinal force and the lateral force of the front wheel, respectively. tf l is the distance between the center of gravity and the front axle of the unmanned tractor. n F is the distance between the center of gravity of the unmanned tractor and the articulation point of the vehicle and the machine. yr For the lateral force of the rear wheel, l tr Let a be the distance between the center of gravity and the rear axle of the unmanned tractor. ty For the lateral acceleration of the unmanned tractor, I t Let m be the moment of inertia of the unmanned tractor, and a be the mass of the unmanned tractor. t2 Let T be the heading angular acceleration of the unmanned tractor, and T be the time that the unmanned tractor has traveled so far, which can be obtained directly through simulation software.

[0104] See Figure 2 After obtaining the heading angle, e can be calculated based on the angle between the heading angle and the tangent at the pre-aiming distance s. θ The minimum lateral distance e N This is the minimum lateral distance between the nose landing gear of the aircraft to be towed and the current path segment. Specifically, once the simulation of a path segment is completed, the lateral distance between the nose landing gear of the aircraft to be towed and that path segment can be obtained. The minimum lateral distance appearing in that path segment is taken as the minimum lateral distance e. N .

[0105] For the PID+FF algorithm, after obtaining eθ and e N Then the turning angle of the current path segment can be calculated.

[0106] A3: After the driving is completed, the average value of the minimum lateral distance corresponding to each path segment is taken as the first lateral distance error corresponding to the current traction simulation control parameter.

[0107] After obtaining the first lateral distance error corresponding to each set of simulation control parameters, the measured control parameters that meet the requirements are selected based on the first preset threshold. Specifically, if the first lateral distance error is less than the first preset threshold, the corresponding simulation control parameter to be tested is used as the measured control parameter.

[0108] Step 4: Select the actual driving route, and conduct on-site PID-FF algorithm control test based on each set of measured control parameters to obtain the second lateral route error corresponding to the set of measured control parameters. If the second lateral route error is less than the second preset threshold, the corresponding measured control parameters are used as a vehicle control model corresponding to the aircraft model, thereby forming a vehicle control model library.

[0109] The second preset threshold can be set to 0.5.

[0110] The process of selecting an actual driving route and conducting on-site PID-FF algorithm control tests based on each set of measured control parameters to obtain the second lateral route error corresponding to that set of measured control parameters includes: performing the following operations for each set of measured control parameters:

[0111] B1: Divide the measured path into M path segments, M = 1, 2, 3, ..., M. Initially, the unmanned towing vehicle is located at the starting point of the measured path. Obtain the pre-aiming distance S′. Initially, the lateral distance between the front landing gear of the aircraft to be towed and the path segment is 0, and the steering angle of the front axle tires of the unmanned towing vehicle is 0. For example, the measured path segments can be divided into M path segments to facilitate subsequent calculations.

[0112] B2: Based on the positioning information of the unmanned tractor, calculate the minimum lateral distance RN between the nose landing gear of the aircraft to be towed and the measured path segment corresponding to the current path segment M, and the angle SN between the heading angle of the unmanned tractor and the tangent at the pre-aiming distance S′. Based on RN, SN, and the set of measured control parameters, obtain the steering angle δ of the unmanned tractor through the PID-FF algorithm, and calculate the difference between δ and the standard steering angle corresponding to the current path segment. The largest steering angle difference of the set of measured control parameters in each measured path segment is taken as the first measured error. The standard steering angle is the optimal steering angle corresponding to the path segment calculated by the central control system based on each path segment of the current measured driving route.

[0113] B3: Based on the positioning information, obtain the actual driving route of the aircraft to be towed, and use the maximum difference in the lateral distance between the actual driving route and the measured path as the second measured error;

[0114] B4: The second lateral route error is obtained based on the first measured error and the second measured error. For example, the numerical part of the first measured error is added to the second measured error to obtain the second lateral route error.

[0115] If a corresponding vehicle control model is not obtained after on-site testing of all measured control parameters, the measured control parameters are fine-tuned to obtain the vehicle control model, including:

[0116] Sort all measured control parameters in ascending order of the second lateral route error, and select at least two sets of measured control parameters with the smallest error as target control parameters for fine-tuning.

[0117] Obtain the first and second measured errors corresponding to the target control parameters, and determine whether the quotient of the first and second measured errors is greater than 1. If it is greater than 1, then adjust the parameter K in the PID+FF algorithm. ffy Fine-tune the parameters; otherwise, adjust the parameter K in the PID+FF algorithm. Py Make minor adjustments.

[0118] For K ffy Adjustment is made using the first preset step size value A for K. Py The adjustment is made with a second preset step size value B, where A and B are both constants. Preferably, both A and B can be set to 0.01.

[0119] The K ffy The fine-tuning process includes:

[0120] D1: With K ffy +A reconstructs the target control parameters and conducts on-site testing. It is determined whether the second lateral route error corresponding to the reconstructed target control parameters is less than the second preset threshold. If it is less, the corresponding target control parameters are used as the vehicle control model. If it is greater, D2 is executed.

[0121] D2: With K ffy -A Reconstruct the target control parameters and conduct on-site testing. Determine whether the second lateral route error corresponding to the reconstructed target control parameters is less than the second preset threshold. If it is less, use the corresponding target control parameters as the vehicle control model. If it is greater, continue to execute D3.

[0122] D3: If K ffy The error of the second lateral route corresponding to +A is greater than K. ffy The second lateral route error corresponding to -A is then determined to be a reduction, and K is used as the adjustment direction.ffy -2A reconstructs the target control parameters and performs field testing; otherwise, the fine-tuning direction is determined to be increasing, and K is used as the benchmark. ffy +2A reconstructed the target control parameters and conducted field tests;

[0123] D4: If the second lateral route error corresponding to the reconstructed target control parameters is less than the second preset threshold, then the corresponding target control parameters are used as the vehicle control model. If it is greater, then continue to construct the target control parameters along the fine-tuning direction determined by D3 for on-site testing until the second lateral route error is less than the second preset threshold. Then, the target control parameters at this time are used as the vehicle control model.

[0124] The K Py The fine-tuning process includes:

[0125] D1: With K ffy +B Reconstruct the target control parameters and conduct on-site testing. Determine whether the second lateral route error corresponding to the reconstructed target control parameters is less than the second preset threshold. If it is less, use the corresponding target control parameters as the vehicle control model. If it is greater, continue to execute D2.

[0126] D2: With K ffy -B Reconstruct the target control parameters and conduct on-site testing. Determine whether the second lateral route error corresponding to the reconstructed target control parameters is less than the second preset threshold. If it is less, use the corresponding target control parameters as the vehicle control model. If it is greater, continue to execute D3.

[0127] D3: If K ffy +B corresponds to a second lateral route error greater than K. ffy The second lateral route error corresponding to -B is then used to determine the fine-tuning direction as decreasing, and K is used as the reference point. ffy -2B Reconstruct the target control parameters and conduct field tests; otherwise, determine the fine-tuning direction as increasing, and use K... ffy +2B Reconstruct the target control parameters and conduct field tests;

[0128] D4: If the second lateral route error corresponding to the reconstructed target control parameters is less than the second preset threshold, then the corresponding target control parameters are used as the vehicle control model. If it is greater, then continue to construct the target control parameters along the fine-tuning direction determined by D3 for on-site testing until the obtained second lateral route error is less than the second preset threshold. Then, the target control parameters at this time are used as the vehicle control model.

[0129] This invention provides an operational control method for an aircraft-unmanned tractor. Based on the PID-FF algorithm, experimental simulation, and field testing, vehicle control models for various aircraft types are obtained. Using these models ensures that the aircraft travels along the centerline of the planned route during traction, significantly reducing lateral errors, improving the accuracy of the unmanned tractor operation, and ensuring safe operation. Furthermore, by pre-storing the vehicle control models corresponding to each aircraft, the efficiency of unmanned tractor operation control is further improved. Initially, experimental simulation is used to obtain theoretically optimal control parameters. However, due to factors such as transmission losses between the steering wheel and the unmanned tractor, road wear, and environmental conditions during actual traction, the unmanned tractor may not perfectly conform to the centerline of the planned route according to the theoretical parameters. Therefore, the vehicle control models obtained through simulation are subjected to field testing, further improving the accuracy, stability, and adaptability of the vehicle control models. When a vehicle control model corresponding to a specific aircraft is not available, a fine-tuning process is used to further explore the possibilities of the PID-FF algorithm control parameters to obtain the optimal vehicle control model that conforms to the actual operating environment.

[0130] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0131] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. An aircraft-unmanned tractor operation control method characterized by, The method comprises the following steps: S1: acquiring an image of the aircraft to be towed by using a camera, acquiring point cloud data of the aircraft to be towed by using a laser radar, and acquiring a distance between the unmanned towing vehicle and the front wheel of the aircraft to be towed, and acquiring a deflection angle of the front wheel of the aircraft to be towed based on the image and the point cloud data; S2: performing a clamping operation of the unmanned towing vehicle based on the deflection angle and the distance, and acquiring a working path of the unmanned towing vehicle after the clamping operation is completed; S3: selecting a vehicle control model corresponding to the model of the aircraft to be towed from the vehicle control model library, and controlling the unmanned towing vehicle to implement a towing task along the working path based on the vehicle control model; wherein the vehicle control model library stores vehicle control models corresponding to each type of aircraft, each vehicle control model is constructed based on a PID-FF algorithm and is used for lateral control of the unmanned towing vehicle, and control parameters of each vehicle control model are different, the control parameters including a proportional gain coefficient, an integral time parameter, a differential time parameter, and a feedforward parameter of lateral control.

2. The aircraft-unmanned tow tractor operation control method according to claim 1, characterized by, The method of acquiring the deflection angle of the front wheel of the aircraft to be towed based on the image information and the point cloud data comprises the following steps: S11: identifying a target front wheel in the image based on an image detection algorithm; S12: performing filtering and downsampling operations on the point cloud data to obtain processed point cloud data; S13: extracting target point cloud data corresponding to the front wheel from the processed point cloud data based on the target front wheel; S14: comparing standard aircraft front wheel point cloud data with the target point cloud data to obtain the deflection angle of the front wheel of the aircraft to be towed, and taking the deflection angle as a towing clamping angle of the unmanned towing vehicle.

3. The aircraft-unmanned tow tractor operation control method according to claim 2, characterized in that, The distance between the unmanned towing vehicle and the front wheel of the current aircraft to be towed is calculated based on a front wheel radius of the aircraft to be towed, a laser radar ranging distance, and a relative distance between the laser radar and the clamping device of the unmanned towing vehicle; specifically as follows: wherein L is the distance between the unmanned towing vehicle and the front wheel of the current aircraft to be towed, 1 is the minimum distance of laser radar ranging, r is the front wheel radius, a is the deflection angle of the front wheel of the aircraft to be towed, and s is the relative distance between the laser radar and the clamping device of the unmanned towing vehicle.

4. The aircraft-unmanned tow tractor operation control method according to claim 3, characterized by The PID-FF algorithm is as follows: wherein, δ n is the steering angle of the front axle tire of the unmanned towing vehicle; K Py is the proportional gain parameter of the lateral control; T Iy is the integral time parameter of the lateral control; T Dy is the differential time parameter of the lateral control; K ffy is the feedforward parameter of the lateral control; e n is the minimum lateral distance between the aircraft nose landing gear and the driving path; e θ is the included angle between the heading angle of the unmanned towing vehicle and the tangent at the preview distance s; n is the path segment corresponding to the current driving route, n = 1, 2, 3…, T.

5. The aircraft-unmanned tug operation control method according to claim 4, characterized by, The vehicle control model of each aircraft model is established in the following manner to form the vehicle control model library: an aircraft-unmanned towing vehicle simulation model is established based on Gazebo 11; Based on historical vehicle control data, set the value range of each control parameter in the PID-FF algorithm, including K Py , T Iy , T Dy , K ffy , and randomly generate multiple groups of to-be-tested simulation control parameters according to the value range; based on each group of to-be-tested simulation control parameters, a first lateral distance error under the to-be-tested simulation control parameters is obtained by performing control simulation based on the PID-FF algorithm, and if the first lateral distance error is less than a first preset threshold, the corresponding to-be-tested simulation control parameter is taken as a measured control parameter; an actual driving route is selected, and a second lateral route error corresponding to each group of measured control parameters is obtained by performing on-site PID-FF algorithm control test based on each group of measured control parameters, and if the second lateral route error is less than a second preset threshold, the corresponding measured control parameter is taken as a vehicle control model corresponding to the aircraft model, thereby forming the vehicle control model library.

6. The aircraft-unmanned tug operation control method according to claim 5, wherein The PID-FF algorithm is adopted to control simulation based on each set of to-be-tested simulation control parameters to obtain the first lateral distance error under the set of to-be-tested simulation control parameters, and the method comprises the following steps: A1: obtaining a simulation path and dividing the simulation path into N path segments, N=1, 2, 3…, N, and initially locating the unmanned towing vehicle at the starting point of the simulation path, driving to a corresponding point on the simulation path with a distance of a preview distance S from the unmanned towing vehicle as a target position, and initially setting the lateral distance between the front landing gear of the to-be-towed aircraft and the path segment to 0 and the steering angle of the front axle tire of the unmanned towing vehicle to 0; A2: for each path segment, calculate the steering angle based on the heading angle θ N , the minimum lateral distance e N and the PID-FF algorithm wherein the heading angle θ N the steering angle δ N-1 is obtained; the minimum lateral distance e N is the minimum lateral distance of the nose landing gear of the aircraft to be towed from the current path segment; A3: after driving, taking the average value of the minimum lateral distance corresponding to each path segment as the first lateral distance error corresponding to the current to-be-towed simulation control parameter.

7. The aircraft-unmanned tug operation control method according to claim 6, characterized by, The actual driving route is selected, and the PID-FF algorithm control test is performed based on each set of measured control parameters to obtain the second lateral route error corresponding to the set of measured control parameters, and the method comprises the following steps: B1: dividing the measured path into M path segments, M=1, 2, 3…, M, and initially locating the unmanned towing vehicle at the starting point of the measured path, obtaining a preview distance S', and initially setting the lateral distance between the front landing gear of the to-be-towed aircraft and the path segment to 0 and the steering angle of the front axle tire of the unmanned towing vehicle to 0; B2: calculating the minimum lateral distance RN between the front landing gear of the to-be-towed aircraft and the measured path segment and the included angle SN between the heading angle of the unmanned towing vehicle and the tangent at the preview distance S' according to the positioning information of the unmanned towing vehicle, obtaining the steering angle δ of the unmanned towing vehicle based on RN, SN and the set of measured control parameters through the PID-FF algorithm, and calculating the difference between δ and the standard steering angle corresponding to the current path segment, and taking the maximum steering angle difference of the set of measured control parameters in each measured path segment as the first measured error; B3: obtaining the actual driving route of the current to-be-towed aircraft based on the positioning information, and taking the maximum difference between the actual driving route and the measured path in the lateral distance as the second measured error; B4: obtaining the second lateral route error based on the first measured error and the second measured error.

8. The aircraft-unmanned tug operation control method according to claim 7, characterized by, If no vehicle control model corresponding to all measured control parameters is obtained after the field test, the measured control parameters are fine-tuned to obtain the vehicle control model, and the method comprises the following steps: sorting the second lateral route errors corresponding to all measured control parameters in ascending order, and selecting at least two sets of measured control parameters with the smallest errors as target control parameters for fine-tuning; The first measured error and the second measured error corresponding to the target control parameter are acquired, it is judged whether the quotient of the first measured error and the second measured error is greater than 1, if greater than 1, the parameter K ffy in the PID+FF algorithm is fine-tuned, otherwise, the parameter K Py in the PID+FF algorithm is fine-tuned.

9. The aircraft-unmanned tug operation control method according to claim 8, wherein For K ffy adjustment is made with a first preset step value A, for K Py adjustment is made with a second preset step value B, wherein A and B are constants The K ffy The fine tuning process comprises: D1: with K ffy +A reconstruct the target control parameter for field test, determine whether the second lateral path error corresponding to the reconstructed target control parameter is less than the second preset threshold, if less than, the corresponding target control parameter is used as the vehicle control model, if greater than, continue to execute D2; D2: with K ffy A reconfigure the target control parameter for field test, judge the second lateral route error corresponding to the reconfigured target control parameter is less than the second preset threshold, if less than, the corresponding target control parameter is used as the vehicle control model, if greater than, continue to execute D3; D3: if K ffy +A is greater than K ffy -A, then the fine-tuning direction is determined to be decreasing, and the target control parameter is reconstructed as K ffy -2A, then the fine-tuning direction is determined to be increasing, and the target control parameter is reconstructed as K ffy +2A, and the target control parameter is reconstructed as K D4: if the second lateral route error corresponding to the target control parameter re-constructed is less than a second preset threshold, the corresponding target control parameter is taken as the vehicle control model, and if it is greater than the second preset threshold, the target control parameter is continuously constructed in the fine-tuning direction determined in D3 for field test until the second lateral route error obtained is less than the second preset threshold, and the target control parameter at this time is taken as the vehicle control model; The K Py fine tuning process is the same as the K ffy fine tuning process.

10. The aircraft-unmanned tow vehicle operation control method according to claim 9, characterized by, the heading angle θ N the steering angle δ based on the previous path segment N-1 obtaining, comprising: θ N = a t2 T, where F xf and F yf are the longitudinal force and the cornering force of the front wheels, respectively, l tf is the distance between the center of mass of the unmanned tractor and the front axle, l n is the distance between the center of mass of the unmanned tractor and the steering joint, F yr is the cornering force of the rear wheels, l tr is the distance between the center of mass of the unmanned tractor and the rear axle, a ty is the lateral acceleration of the unmanned tractor, I t is the moment of inertia of the unmanned tractor, m is the mass of the unmanned tractor, δ is the steering angle of the unmanned tractor, a t2 is the yaw angular acceleration of the unmanned tractor, and T is the current driving time.

Images