A model predictive contour control system and method based on neural network enhancement

Abstract

The application belongs to the technical field of unmanned aerial vehicle control, and discloses a model prediction contour control system and method based on neural network enhancement, wherein the system processes multi-source sensing data of the unmanned aerial vehicle through an S4-Conv adaptive network architecture, including an odometer, thrust output and a PVDF flexible collision sensor, and accurately predicts the next moment position and attitude of the unmanned aerial vehicle. Based on the prediction model, a neural network enhanced model prediction contour control framework is proposed, which is used for optimizing the thrust output of the unmanned aerial vehicle and performing real-time attitude adjustment. The method can quickly respond to external collision and dynamic disturbance in a complex environment, significantly improves the trajectory tracking accuracy and stability of task execution of the unmanned aerial vehicle, and is suitable for unmanned aerial vehicle flight control tasks in complex environments such as urban space and disaster rescue.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) control technology, and particularly relates to a model prediction contour control system and method based on neural network enhancement. Background Technology

[0002] When performing complex missions, unmanned aerial vehicles (UAVs) typically face external disturbances and potential collision risks, especially in dynamic or confined environments such as urban spaces and disaster relief operations, where external collisions are almost unavoidable. Existing control methods are primarily based on fixed model designs, neglecting real-time response to sudden external collisions and adaptability to dynamic environments. While traditional model predictive contour control frameworks can optimize trajectory, they often struggle to adjust quickly in the face of sudden collisions and complex disturbances, causing the UAV to deviate from its intended trajectory or even become uncontrollable. Furthermore, existing control algorithms lack sufficient adaptability, failing to adjust control parameters based on real-time sensor data, resulting in poor UAV performance in dynamically changing environments. Therefore, a more adaptive UAV control method is needed, capable of handling trajectory tracking problems in complex environments and rapidly responding to external collisions and dynamic disturbances. Summary of the Invention

[0003] The purpose of this invention is to provide a model prediction contour control system and method based on neural network enhancement, so as to solve the above-mentioned technical problems.

[0004] To address the aforementioned technical problems, the present invention provides a specific technical solution for a model prediction contour control system and method based on neural network enhancement, as follows:

[0005] A model prediction contour control system based on neural network enhancement includes: an S4Conv adaptive network architecture for receiving and processing multi-source sensor data to predict the future position and attitude of an unmanned aerial vehicle;

[0006] A PVDF flexible collision sensor is used to detect information about collisions between a drone and external objects. The PVDF flexible collision sensor senses collision information through voltage signals and inputs it into an S4-Conv adaptive network architecture for processing.

[0007] A neural network-enhanced model prediction contour control framework is used to optimize the thrust output of the UAV and adjust its attitude in real time based on the prediction results.

[0008] This invention also discloses a control method for a model prediction contour control system based on neural network enhancement, comprising the following steps:

[0009] Step 1: Synchronization and fusion of multi-source sensor data;

[0010] Step 2: Data preprocessing and augmentation;

[0011] Step 3: Trajectory prediction using the Sana prediction model;

[0012] Step 4: Trajectory optimization for model-predicted contour control.

[0013] Furthermore, step 1 includes data sourcing and data synchronization. The data source is data received by the Sana prediction model from multiple sensors, including information from odometry, thrust sensors, and PVDF flexible collision sensors. This data is fused to sense and predict the UAV's flight status. The data synchronization employs a timestamp-based alignment and synchronization method.

[0014] Assuming different sensors (S1, S2, ..., S...) n The sampling times are (t1, t2, ..., t). n The data synchronization formula, based on timestamp alignment, is as follows:

[0015]

[0016] This method aligns all data to a uniform time step t. aligned This ensures the consistency of multidimensional sensor data.

[0017] Furthermore, step 2 includes:

[0018] Preprocessing: For sensor data x t The Exponentially Weighted Moving Average (EWMA) filtering formula is used:

[0019]

[0020] Where α is the smoothing coefficient, used to control the weight of historical data;

[0021] Standardize the sensor data to ensure that data of different dimensions have consistent mean and variance:

[0022]

[0023] Where μ is the mean of the data and σ is the standard deviation. Standardization makes all sensor data have the same mean of 0 and variance of 1.

[0024] Data augmentation includes voltage noise simulation, thrust disturbance, random rotation, and displacement.

[0025] Furthermore, step 3 includes:

[0026] S4-Conv Adaptive Network Architecture: The Sana prediction model is based on the S4D state-space sequence modeling method. Through the fusion of multi-source sensor data, it generates high-precision future trajectory predictions, providing the future position, velocity, and attitude of the UAV.

[0027] Suppose X t Given multi-source sensor data at time t, predicting the trajectory over the next k time steps, the output trajectory of the Sana prediction model is:

[0028]

[0029] Among them, f θ It is a neural network function based on the S4D state-space model, where θ represents the network parameters. Further, step 4 includes:

[0030] Trajectory Generation and Optimization: Based on the results of UAV path search and path optimization, the Mina control framework first calculates the ideal trajectory of the UAV to ensure that the UAV can follow the target trajectory with the optimal path. The current path is p(t) = {p1, p2, ..., p m The target path is p. target (t)={p1,p2,…,p m The position error is defined as:

[0031] e(t) = p target (t)-p(t)

[0032] PID control is used during the calculation process:

[0033]

[0034] Among them, K p K i K d These are the proportional, integral, and derivative gains of the PID control, respectively.

[0035] Furthermore, step 5 includes:

[0036] Collision detection: When the PVDF flexible collision sensor detects a collision between the UAV and an external object, the Sana prediction model immediately updates the post-collision state, and the Mina control framework adjusts based on this real-time feedback information;

[0037] Adaptive adjustment: The Mina control framework uses the MPCC mechanism to adjust thrust output and attitude control after a collision, restoring the UAV to normal flight status.

[0038] Furthermore, step 6 includes:

[0039] Dynamic environment adaptation: The MPCC framework monitors external disturbances in real time through the Sana prediction model and adjusts the trajectory based on this data. Through online optimization, the control system can quickly respond to external disturbances.

[0040] Furthermore, step 7 includes:

[0041] Under the Model Predictive Contour Control (MPCC) framework, the optimization objectives for the control input include the following four items: minimizing position error, minimizing velocity error, smoothing thrust change, and minimizing thrust power consumption. Assume: p(t) represents the current position of the UAV, v(t) represents the current velocity of the UAV, T(t) represents the current thrust of the UAV, and T... prev (t-1) represents the thrust at the previous moment, and P(t) represents the current thrust power consumption. Multiple objective functions are defined and combined into a single overall optimization problem:

[0042] (1) Minimize position error: The goal is to minimize the difference between the current position of the UAV and the ideal target position p. target The error between (t):

[0043] J1=‖p target (t)-p(t)‖ 2

[0044] (2) Minimize velocity error: The objective is to minimize the difference between the current velocity of the UAV and the ideal velocity v. target The error between (t):

[0045] J2 = ||v target (t)-v(t)‖ 2

[0046] (3) Smoothness of thrust changes: The goal is to minimize the drasticness of thrust changes and maintain the smoothness of thrust adjustment.

[0047] J3=‖T(t)-T prev (t-1)‖ 2

[0048] (4) Minimize thrust power consumption: The goal is to reduce energy consumption and optimize thrust power.

[0049]

[0050] Where P(t) is the power consumption of thrust T(t), which is usually related to thrust and the UAV's load, and is calculated using the following formula:

[0051] P(t) = T(t)·v(t)

[0052] (5) Combine these objective functions to construct a weighted comprehensive optimization objective function:

[0053] J = λ1J1 + λ2J2 + λ3J3 + λ4J4

[0054] Where λ1, λ2, λ3, and λ4 are the weights of each optimization objective, used to balance the relative importance of each optimization objective.

[0055] Finally, by optimizing the above objective function, taking into account trajectory error, external disturbances, and energy optimization factors, the thrust T(t) and attitude quaternion q(t) of the UAV are adjusted and output to the controller.

[0056] Furthermore, step 8 includes:

[0057] Offline training: Based on historical flight data D={X1,X2,...,X N The loss function of the neural network model is:

[0058]

[0059] By minimizing the loss function L(θ), the offline training model learns typical flight patterns and solutions for dealing with emergencies;

[0060] Online optimization: During flight, the MPCC framework continuously updates model parameters based on sensor data.

[0061] Improve adaptive capabilities and control precision during flight;

[0062] θ t+1 =θ t -η▽ θ L t

[0063] Where η is the learning rate, L t It is the loss function for the current time step.

[0064] The model prediction contour control system and method based on neural network enhancement of the present invention have the following advantages:

[0065] 1. Strong real-time response capability: Through the S4-Conv adaptive network architecture in the Sana prediction model, it can quickly process multi-source sensor data of UAVs, accurately predict position and attitude, and adjust the control strategy in real time according to external collisions or disturbances.

[0066] 2. High adaptability and robustness: The Sana prediction model integrates multi-dimensional information from thrust, odometry, and flexible collision sensors, combined with data augmentation technology, to improve the UAV's adaptability in complex and dynamic environments, making it suitable for efficient mission execution in urban, forest, and disaster areas.

[0067] 3. Adaptive adjustment after collision: The Mina control framework integrates neural network enhancement, which can sense and make adaptive control adjustments in real time after the UAV collides, significantly improving the trajectory tracking accuracy and stability of the UAV in complex environments.

[0068] 4. High precision and energy saving: By combining model prediction contour control with high-precision prediction of neural networks, this method optimizes the thrust output of UAVs, ensuring flight efficiency and safety while effectively saving energy consumption.

[0069] 5. Improved mission success rate: This invention can maintain flight stability when there is interference in the external environment, reducing the possibility of mission failure due to trajectory deviation or loss of control, and is applicable to various complex flight missions. Attached Figure Description

[0070] Figure 1 This is a schematic diagram of the front structure of the quadcopter drone with a spherical shell according to the present invention.

[0071] Figure 2 This is a top view schematic diagram of the quadcopter drone with a spherical shell according to the present invention;

[0072] Figure 3 This is a diagram of the model prediction contour control framework of the present invention, which incorporates collision-based neural network enhancement.

[0073] Figure 4 This is a network architecture diagram of the Sana prediction model of the present invention. Detailed Implementation

[0074] To better understand the purpose, structure, and function of this invention, the following detailed description of a model prediction contour control system and method based on neural network enhancement is provided in conjunction with the accompanying drawings.

[0075] This invention utilizes an S4-Conv adaptive network architecture, fusing data from odometry, thrust, and a PVDF flexible collision sensor. Leveraging the learning and generalization capabilities of neural networks, it accurately predicts the future position and attitude of a UAV. The UAV can employ methods such as... Figure 1 Figure 2 The structure is shown. By combining a neural network-enhanced model prediction contour control framework, this invention can quickly respond to sudden collision events and adjust the UAV's attitude and thrust output in real time, improving the trajectory tracking accuracy and stability of the UAV in complex environments. It consists of two core parts: 1) a prediction model; 2) a control framework.

[0076] The first core component of this invention proposes the Sana Prediction Model (S4-Conv Adaptive Net Architecture), which aims to leverage the powerful learning and generalization capabilities of neural networks to achieve high-precision prediction of UAV position and velocity by fusing multi-source information such as flexible collision sensor PVDF, thrust, and odometry, thereby adapting to dynamic changes in complex environments. Unlike traditional fixed models, the Sana Prediction Model can adapt to external disturbances in real time, improving the accuracy of trajectory prediction.

[0077] The second core component of this invention proposes the Mina control framework (MPCC Integrated with Neural Augmentation), which introduces neural network enhancements to traditional model predictive contour control, enabling it to handle unexpected situations, such as collisions between UAVs and external objects. By integrating the results of the Sana prediction model, the Mina control framework achieves real-time perception and control adjustment of the post-collision state, solving the problems of slow response and inaccurate adjustment in complex environments of traditional control methods. This significantly improves the trajectory tracking capability and autonomous adaptability of UAVs in complex tasks.

[0078] Specifically, the present invention provides a model prediction contour control system based on neural network enhancement, comprising:

[0079] The S4Conv adaptive network architecture is used to receive and process multi-source sensor data to predict the future position and attitude of the UAV. The S4-Conv adaptive network architecture integrates odometry data, thrust output data and collision sensor data, and improves the robustness of data processing through standardization and data augmentation techniques.

[0080] The PVDF flexible collision sensor is used to detect information about collisions between a drone and external objects. The PVDF flexible collision sensor senses collision information through voltage signals and inputs it into an S4-Conv adaptive network architecture for processing, so as to realize real-time state adjustment of the drone after the collision.

[0081] A model prediction contour control framework based on neural network enhancement is used to optimize the thrust output of the UAV and adjust its attitude in real time based on the prediction results. By combining neural network enhancement, the model prediction contour control framework can quickly respond to changes in the external environment and collision situations, improving the trajectory tracking accuracy and stability of the UAV.

[0082] like Figure 3 As shown, the present invention provides a model prediction contour control method based on neural network enhancement, comprising the following steps:

[0083] Step 1: Synchronization and Fusion of Multi-Source Sensor Data

[0084] like Figure 4 As shown, the data source: The Sana prediction model receives data from multiple sensors, including odometry, thrust sensors, and flexible collision sensors (PVDF). This data is fused to accurately sense and predict the UAV's flight status. Unlike traditional fixed models, the Sana prediction model can adapt to external disturbances in real time, improving the accuracy of trajectory prediction.

[0085] Data synchronization: Since different sensors have different sampling frequencies and delays, a timestamp-based alignment and synchronization method is adopted to ensure that the multidimensional data at each time step is consistent, so that the prediction model can process sequentially.

[0086] Assuming different sensors (S1, S2, ..., S...) n The sampling times are (t1, t2, ..., t). n The data synchronization formula, based on timestamp alignment, is as follows:

[0087]

[0088] This method aligns all data to a uniform time step t. aligned This ensures the consistency of multidimensional sensor data.

[0089] Step 2: Data Preprocessing and Augmentation

[0090] Preprocessing: For sensor data x t The Exponentially Weighted Moving Average (EWMA) filtering formula is used:

[0091]

[0092] Where α is the smoothing coefficient, used to control the weight of historical data.

[0093] Further standardization of the sensor data ensures that data of different dimensions have consistent mean and variance, thereby avoiding negative impacts on model training.

[0094]

[0095] Where μ is the mean of the data and σ is the standard deviation. Standardization ensures that all sensor data have the same mean of 0 and variance of 1.

[0096] Data augmentation: Data augmentation techniques are introduced to simulate complex situations that may be encountered under different flight scenarios, enhancing the robustness of the model. This includes voltage noise simulation, thrust disturbance, random rotation, and displacement simulation to enhance the model's adaptability to sensor instability and external disturbances.

[0097] Step 3: Trajectory Prediction by the Sana Prediction Model

[0098] S4-Conv Adaptive Network Architecture: The Sana prediction model is based on the S4D state-space sequence modeling method. Through the fusion of multi-source sensor data, it generates high-precision future trajectory predictions, providing the future position, velocity, and attitude of the UAV.

[0099] Suppose X t Given multi-source sensor data at time t, predicting the trajectory over the next k time steps, the output trajectory of the Sana prediction model is:

[0100]

[0101] Among them, f θ It is a neural network function based on the S4D state-space model, where θ represents the network parameters.

[0102] Step 4: Trajectory optimization for model-predicted contour control

[0103] Trajectory Generation and Optimization: Based on the results of UAV path search and optimization, the Mina control framework first calculates the ideal trajectory of the UAV to ensure that the UAV can follow the target trajectory with the optimal path. The current path is p(t) = {p1, p2, ..., p m The target path is p. target (t)={p1,p2,…,p m Position error is defined as follows:

[0104] e(t) = p target (t)-p(t)

[0105] In the calculation process, traditional PID control is used.

[0106]

[0107] It adapts to changes in the external environment in real time, generating initial values ​​for thrust and attitude to provide a numerical basis for subsequent optimization, where K p K i K d These are the proportional, integral, and derivative gains of the PID control, respectively.

[0108] Step 5: Collision Sensing and Adaptive Control

[0109] Collision Detection: When the PVDF flexible collision sensor detects a collision between the drone and an external object, the Sana prediction model immediately updates the post-collision state, and the Mina control framework adjusts based on this real-time feedback information.

[0110] Adaptive adjustment: The Mina control framework uses the MPCC mechanism to quickly adjust thrust output and attitude control after a collision, restoring the UAV to normal flight status and preventing deviation from the trajectory or loss of control.

[0111] Step 6: Trajectory Tracking in Disturbed Environments

[0112] Dynamic Environment Adaptation: The MPCC framework uses the Sana predictive model to monitor external disturbances (such as wind speed changes or sudden obstacles) in real time and adjusts the trajectory based on this data. Through online optimization, the control system can respond quickly to external disturbances, ensuring the flight stability of the UAV in complex environments.

[0113] Step 7: Real-time optimization of control input

[0114] Within the Model Predictive Contour Control (MPCC) framework, the optimization objectives for the control input include the following four items: minimizing position error, minimizing velocity error, smoothing thrust variation, and minimizing thrust power consumption. Assumptions: p(t) represents the current position of the UAV, v(t) represents the current velocity of the UAV, T(t) represents the current thrust of the UAV, and T... prev (t-1) represents the thrust at the previous moment, and P(t) represents the current thrust power consumption.

[0115] We can define multiple optimization objective functions and combine them into a single overall optimization problem:

[0116] (1) Minimize position error: The goal is to minimize the difference between the current position of the UAV and the ideal target position p. target The error between (t)

[0117] J1=‖p target (t)-p(t)‖ 2

[0118] (2) Minimize velocity error: The objective is to minimize the difference between the current velocity of the UAV and the ideal velocity v. target The error between (t)

[0119] J2 = ||v target (t)-v(t)‖ 2

[0120] (3) Smoothness of thrust changes: The goal is to minimize the drasticness of thrust changes and maintain the smoothness of thrust adjustment.

[0121] J3=‖T(t)-T prev (t-1)‖ 2

[0122] (4) Minimize thrust power consumption: The goal is to reduce energy consumption and optimize thrust power.

[0123]

[0124] Where P(t) is the power consumption of thrust T(t), which is usually related to thrust and the UAV's load, and is calculated using the following formula:

[0125] P(t) = T(t)·v(t)

[0126] (5) Combine these objective functions to construct a weighted comprehensive optimization objective function:

[0127] J = λ1J1 + λ2J2 + λ3J3 + λ4J4

[0128] Wherein, λ1, λ2, λ3, and λ4 are the weights of each optimization objective, used to balance the relative importance of each optimization objective.

[0129] Finally, by optimizing the above objective function and taking into account factors such as trajectory error, external disturbance, and energy optimization, the thrust T(t) and attitude quaternion q(t) of the UAV are adjusted and output to the controller.

[0130] Step 8: Offline and Online Training

[0131] Offline training: Through offline training, the neural network learns typical flight patterns and contingency response plans to enable rapid reactions in actual flight. Based on historical flight data D={X1,X2,…,X… N The loss function of the neural network model is:

[0132]

[0133] By minimizing the loss function L(θ), the offline training model learns typical flight patterns and solutions for dealing with emergencies.

[0134] Online optimization: During flight, the MPCC framework continuously updates model parameters based on sensor data, improving the adaptive capability and control accuracy during flight.

[0135] θ t+1 =θ t -η▽ θ L t

[0136] Where η is the learning rate, L t It is the loss function for the current time step.

[0137] Step 9: Final Control Output

[0138] Control Output: After completing trajectory optimization, the MPCC framework outputs the final control commands to the UAV's flight control system, ensuring that the UAV continuously performs high-precision trajectory tracking tasks in complex environments. Throughout the process, the Mina control framework, combined with feedback from the Sana predictive model, continuously performs real-time optimization and adjustments.

[0139] Through the above steps, this invention combines the Sana prediction model with the Mina control framework to form a model prediction contour control system based on neural network enhancement. This system can perform high-precision trajectory tracking and adaptive control adjustments in real time in complex and dynamic environments, effectively improving the execution efficiency and stability of UAVs in complex tasks.

[0140] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A model prediction contour control system based on neural network enhancement, characterized in that, include: The S4Conv adaptive network architecture is used to receive and process multi-source sensor data to predict the future position and attitude of the drone. A PVDF flexible collision sensor is used to detect information about collisions between a drone and external objects. The PVDF flexible collision sensor senses collision information through voltage signals and inputs it into an S4Conv adaptive network architecture for processing. A model prediction contour control framework based on neural network enhancement is used to optimize the thrust output of the UAV and adjust its attitude in real time based on the prediction results.

2. A control method for a model prediction contour control system based on neural network enhancement as described in claim 1, characterized in that, Includes the following steps: Step 1: Synchronization and fusion of multi-source sensor data; Step 2: Data preprocessing and augmentation; Step 3: Trajectory prediction using the S4Conv adaptive network architecture; Step 4: Trajectory optimization for model-predicted contour control.

3. The control method according to claim 2, characterized in that, Step 1 includes data sourcing and data synchronization. The data source is data received from multiple sensors via the S4Conv adaptive network architecture, including information from odometry, thrust sensors, and PVDF flexible collision sensors. This data is fused to sense and predict the UAV's flight status. The data synchronization employs a timestamp-based alignment and synchronization method. Assuming different sensors (S1, S2, ..., S...) n The sampling times are (t1, t2, ..., t). n The data synchronization formula, based on timestamp alignment, is as follows: This method aligns all data to a uniform time step t. aligned This ensures the consistency of multidimensional sensor data.

4. The control method according to claim 2, characterized in that, Step 2 includes: preprocessing: for sensor data x t The exponentially weighted moving average (EWMA) filter formula is used: Where α is the smoothing coefficient, used to control the weight of historical data; Standardize the sensor data to ensure that data of different dimensions have consistent mean and variance: Where μ is the mean of the data and σ is the standard deviation. Standardization makes all sensor data have the same mean of 0 and variance of 1. Data augmentation includes voltage noise simulation, thrust disturbance, random rotation, and displacement.

5. The control method according to claim 2, characterized in that, Step 3 includes: S4Conv Adaptive Network Architecture: The S4Conv adaptive network architecture is based on the S4D state-space sequence modeling method. It generates high-precision future trajectory predictions through multi-source sensor data fusion, providing the future position, velocity, and attitude of the UAV. Suppose X t Given multi-source sensor data at time t, predicting the trajectory over the next k time steps, the output trajectory of the S4Conv adaptive network architecture is: Among them, f θ It is a neural network function based on the S4D state-space model, where θ represents the network parameters.

6. The control method according to claim 2, characterized in that, Step 4 includes: trajectory generation and optimization: Based on the results of UAV path search and path optimization, a model prediction contour control framework based on neural network enhancement first calculates the ideal trajectory of the UAV to ensure that the UAV can follow the target trajectory with the optimal path. The current path is p(t) = {p1, p2, ..., p...} m The target path is p. target (t)={p1′,p2′,...,p m The position error is defined as: e(t)=p target (t)-p(t) PID control is used during the calculation process: Among them, K p K i K d These are the proportional, integral, and derivative gains of the PID control, respectively.

7. The control method according to claim 2, characterized in that, It also includes step 5: collision detection: when the PVDF flexible collision sensor detects that the UAV collides with an external object, the S4Conv adaptive network architecture immediately updates the state after the collision, and the model prediction contour control framework based on neural network enhancement adjusts according to these real-time feedback information; Adaptive Adjustment: The model prediction profile control framework based on neural network enhancement adjusts thrust output and attitude control after a collision to restore the UAV to normal flight status.

8. The control method according to claim 2, characterized in that, It also includes step 6: Dynamic environment adaptation: The model prediction contour control framework based on neural network enhancement monitors external disturbances in real time through the S4Conv adaptive network architecture and adjusts the trajectory based on these data. Through online optimization, the control system can quickly respond to external disturbances.

9. The control method according to claim 2, characterized in that, Step 7 is also included: Under the model prediction contour control framework based on neural network enhancement, the optimization objectives of the control input include the following four items: minimizing position error, minimizing velocity error, smoothness of thrust change, and minimizing thrust power consumption. Assume: p(t) represents the current path of the UAV, v(t) represents the current velocity of the UAV, T(t) represents the current thrust of the UAV, and T... prev (t-1) represents the thrust at the previous moment, and P(t) represents the current thrust power consumption. Define multiple optimization objective functions and combine them into a single overall optimization problem: (1) Minimize position error: The objective is to minimize the difference between the current path of the UAV and the ideal target path p. target The error between (t): J1=‖p target (t)-p(t)‖ 2 (2) Minimize speed error: The objective is to minimize the difference between the current speed of the UAV and the ideal speed v. target The error between (t): J2=‖v target (t)-v(t)‖ 2 (3) Smoothness of thrust changes: The goal is to minimize the drasticness of thrust changes and maintain the smoothness of thrust adjustment. J3=‖T(t)-T prev (t-1)‖ 2 (4) Minimize thrust power consumption: The goal is to reduce energy consumption and optimize thrust power. Where P(t) is the power consumption of thrust T(t), which is usually related to thrust and the UAV's load, and is calculated using the following formula: P(t) = T(t)·v(t) (5) Combine these objective functions to construct a weighted comprehensive optimization objective function: J = λ1J1 + λ2J2 + λ3J3 + λ4J4 Where λ1, λ2, λ3, and λ4 are the weights of each optimization objective, used to balance the relative importance of each optimization objective. Finally, by optimizing the above objective function, taking into account trajectory error, external disturbances, and energy optimization factors, the thrust T(t) and attitude quaternion q(t) of the UAV are adjusted and output to the controller.

10. The control method according to claim 5, characterized in that, It also includes step 8: Offline training: based on historical flight data D={X1,X2,...,X... N The loss function of the neural network model is: By minimizing the loss function L(θ), the offline training model learns typical flight patterns and solutions for dealing with emergencies; Online optimization: During flight, the model prediction profile control framework based on neural network enhancement continuously updates the model parameters according to the sensor data, improving the adaptive capability and control accuracy during flight; i t+1 =θ t -η▽ θ L t Where η is the learning rate, L t It is the loss function for the current time step.

Images