Unmanned aerial vehicle control system and unmanned aerial vehicle control method

Abstract

The invention relates to the technical field of unmanned aerial vehicle flight control, and discloses an unmanned aerial vehicle control system and an unmanned aerial vehicle control method.The unmanned aerial vehicle control system comprises a body sensing module used for directly measuring external acting force, a main control module and a reflex arc control module, and the main control module and the reflex arc control module are arranged in parallel; and the main control module carries out rapid disturbance feed-forward compensation by using ontology sensing information, an online dynamic identification module fuses a control instruction, a motion response and a direct disturbance observed quantity provided by the ontology sensing module, and updates a control gain of the main control module in real time by using the parameter, so that high-precision control self-adaption is realized. According to the invention, the body sensing module for directly collecting the mechanical distribution data of the body structure is arranged, and the reflex arc control module for processing the data is combined, so that a rapid disturbance suppression path independent of traditional motion state feedback is constructed, the timeliness and the directness of the system to sudden and local disturbance response are improved, and the reliability of the system is improved. And the flight stability of the unmanned aerial vehicle is enhanced.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) flight control technology, specifically to a UAV control system and a UAV control method. Background Technology

[0002] Unmanned aerial vehicles (UAVs), especially multi-rotor UAVs, have been widely used in many fields such as aerial photography, logistics transportation, agricultural plant protection, and power line inspection due to their advantages of simple structure, maneuverability, and vertical takeoff and landing capabilities. As application scenarios continue to expand, the flight environments faced by UAVs are becoming increasingly complex, including turbulence between urban buildings, gusts of wind during field operations, and changes in their own characteristics caused by variations in mission payloads or delivery methods. These factors place higher demands on the robustness and adaptability of UAV flight control systems.

[0003] In existing technologies, the design of UAV flight control systems is mostly based on a predetermined, idealized mathematical model. The parameters of the controller (such as a common three-loop PID controller) are tuned and optimized according to this nominal model in order to obtain the best flight performance under ideal conditions. This control method relies on traditional navigation sensors such as inertial measurement units (IMUs) and global navigation satellite systems (GNSS) to acquire motion state information such as attitude, velocity, and position of the UAV, and performs feedback control based on the error between this information and the desired state.

[0004] However, traditional control methods based on fixed-parameter models reveal inherent limitations in practical applications. Firstly, this control strategy is essentially a form of "post-event compensation." When the UAV encounters external aerodynamic disturbances, the controller can only detect these errors through sensors and drive the motors to compensate after the disturbance has already caused an observable deviation in the UAV's flight attitude or trajectory. This response delay makes it ineffective in dealing with sudden, intense disturbances, failing to effectively suppress instantaneous attitude deviations and affecting flight stability and safety.

[0005] Secondly, the dynamic characteristics of UAVs during missions are time-varying, not static. For example, battery depletion leads to a continuous decrease in the overall mass of the UAV; mounting or dropping different mission payloads causes a step change in its mass and moment of inertia matrix; aging or minor damage to the airframe structure also affects its aerodynamic characteristics. Fixed controller parameters cannot adapt to these changes in inherent physical characteristics, resulting in suboptimal performance of the control system throughout the mission cycle, and even stability degradation when characteristics change significantly.

[0006] To address the time-varying dynamics problem, existing technologies have proposed several adaptive control schemes. The core idea is to identify the dynamic model parameters of the UAV online and adjust the controller gain in real time based on the identification results. However, the performance of these online identification algorithms is largely limited by external disturbances. During the identification process, the algorithms struggle to effectively distinguish between the motion response caused by unknown external disturbances / torques and the motion response caused by controller commands. External disturbances are considered "colored noise," which severely contaminates the data used for parameter identification, leading to inaccurate or even divergent identification results, thus affecting the reliability and effectiveness of the entire adaptive control closed loop.

[0007] Furthermore, both traditional fixed-parameter control and existing adaptive control ultimately limit their control methods to adjusting the rotor thrust output by regulating the motor speed. When UAVs face extremely harsh environments, such as continuous strong winds or periodic excitations close to their structural natural frequency, the actuators (motors) overheat due to prolonged high-load operation, or lose control due to the thrust reaching its physical limit (saturation). Especially under resonant conditions, relying solely on aerodynamic compensation is often insufficient to suppress catastrophic vibration accumulation. This indicates that existing control frameworks lack a physical dimension of control when dealing with extreme disturbances. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides a drone control system and a drone control method, which solves the problem of delayed response to external disturbances in existing drone control systems due to their reliance on motion state feedback.

[0009] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of the present invention provides a control system for an unmanned aerial vehicle (UAV), which includes: a body perception module, a main control module, a reflection arc control module, and a control command fusion module.

[0010] The physical carrier of the body perception module is one or more sets of sensors integrated on the body structure of the UAV, which are used to collect mechanical distribution data at multiple predetermined positions on the body structure in real time. The mechanical distribution data objectively reflects the real-time, distributed force characteristics generated on the body by external airflow or physical contact factors.

[0011] The main control module is used to receive motion state data that characterizes the overall motion state of the UAV, and generate main control commands to maintain or change the long-term flight attitude and trajectory of the UAV based on the deviation between the motion state data and the preset flight commands.

[0012] The input terminal of the reflection arc control module is connected to the output terminal of the body sensing module, and is used to directly receive and process the mechanical distribution data. The function of this module is to generate a differential compensation control command based on the disturbance pattern presented by the mechanical distribution data. The purpose of the differential compensation control command is to generate an instantaneous counteracting torque that is opposite to the external disturbance force or torque.

[0013] The control command fusion module has its input terminals connected to the output terminals of the main control module and the reflection arc control module, respectively. This module is used to merge the received main control commands and differential compensation control commands to generate the final motor control commands that drive the UAV motors.

[0014] Preferably, the body sensing module may specifically include at least one of a pressure sensor array, a strain gauge array, or a piezoelectric thin film sensor array disposed on the UAV arm, central frame, or motor base to obtain high-resolution mechanical distribution information.

[0015] Preferably, the internal processing unit of the reflection arc control module may include a pre-trained temporal neural network model, which is trained to process the temporal mechanical distribution data input by the ontology perception module and establish a direct mapping relationship from specific mechanical data patterns to corresponding differential compensation control commands.

[0016] Preferably, the system may further include an online dynamics identification module, whose input terminals are connected to the main control module, the body perception module, and sensors for providing motion state data. Its function is to comprehensively analyze the command output of the control system, the mechanical response of the aircraft, and the final motion response, thereby identifying and updating the dynamics model parameters of the UAV in real time and online. These dynamics model parameters include the UAV's mass, moment of inertia, or aerodynamic coefficients.

[0017] Preferably, there is an information feedback path between the main control module and the online dynamics identification module. The main control module receives the latest dynamics model parameters output by the online dynamics identification module and adaptively adjusts its internal control law based on the updated parameters, such as adjusting the gain parameter of its PID controller.

[0018] Preferably, the time-series neural network model in the reflection arc control module, in addition to generating differential compensation control commands, is also configured to output disturbance feature information that characterizes key attributes of external disturbances. The disturbance feature information may include the estimated intensity, dominant frequency, or direction of action of the disturbance.

[0019] Preferably, the system may further include an active structure adjustment module, the input of which is connected to the reflection arc control module, for receiving the disturbance characteristic information. When the disturbance characteristic information meets preset conditions, the module generates and outputs a structure adjustment command. The structure adjustment command is used to control variable physical characteristic elements integrated in the UAV's airframe structure.

[0020] Preferably, the variable physical characteristic element may specifically include at least one of a variable damping element or a variable stiffness element. By executing the structural adjustment command, the UAV can actively change its own physical damping or stiffness characteristics to dissipate or avoid specific disturbance energy from a physical perspective.

[0021] Preferably, the control command fusion module fuses the main control command and the differential compensation control command by linearly superimposing them to form the final motor control command.

[0022] A second aspect of the present invention provides a method for controlling an unmanned aerial vehicle (UAV), which achieves stable control of the UAV by performing the following steps: First, the body perception module installed on the UAV's body structure collects mechanical distribution data at multiple locations on the body structure in real time.

[0023] In parallel, the main control module generates main control commands to maintain the long-term flight attitude and trajectory of the UAV based on the difference between the UAV's motion state data and the preset flight commands.

[0024] Meanwhile, the mechanical distribution data is directly processed through an independent reflection arc control module, and differential compensation control commands are generated to instantaneously counteract external disturbances.

[0025] Finally, the main control command and the differential compensation control command are fused together to obtain the final motor control command, and the various motors of the UAV are controlled according to the final motor control command.

[0026] Preferably, the method may further include: using an online dynamics identification module to comprehensively analyze the output of the main control module, the motion state data of the UAV, and the mechanical distribution data to identify the dynamics model parameters of the UAV online; and using the identified dynamics model parameters to adaptively adjust the control law that generates the main control command.

[0027] Preferably, the method may further include: while generating differential compensation control commands, identifying and outputting disturbance characteristic information that characterizes external disturbance attributes, and deciding whether to generate and execute a structural adjustment command based on whether the disturbance characteristic information meets preset conditions, so as to control the variable physical characteristic elements in the UAV body and change the physical damping or stiffness characteristics of the body.

[0028] This invention provides a drone control system and a drone control method. It has the following beneficial effects: 1. This invention constructs a rapid disturbance suppression path independent of traditional motion state feedback by setting up a body perception module that directly collects mechanical distribution data of the airframe structure and combining it with a reflection arc control module that processes the data. This path can generate and apply differential compensation control commands to counteract external disturbances before they significantly affect the overall flight attitude of the UAV, thereby improving the timeliness and directness of the system's response to sudden and local disturbances and enhancing the flight stability of the UAV.

[0029] 2. This invention, by adding an online dynamics identification module and using the mechanical distribution data collected by the body perception module as one of its inputs, enables the system to effectively separate external disturbance forces from the UAV's own dynamic response. This makes the online identification results of the UAV's mass and moment of inertia intrinsic dynamic model parameters more accurate, thereby allowing the main control module to make adaptive adjustments based on a more accurate model, improving the control accuracy and robustness of the UAV under load changes or model uncertainties.

[0030] 3. This invention provides a new adjustment dimension for UAVs in addition to motor power compensation by setting an active structural adjustment module that can identify disturbance characteristics and working in conjunction with variable physical characteristic components integrated on the body. In response to continuous or periodic external disturbances, the system can actively change the physical damping or stiffness characteristics of the body to dissipate or avoid disturbance energy at the physical level. This helps to suppress resonance phenomena at specific frequencies and can reduce energy consumption caused by continuous resistance to disturbances under specific operating conditions. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the overall functional architecture of the unmanned aerial vehicle control system of the present invention; Figure 2 This is a schematic diagram of the physical layout of the body sensing module of the present invention; Figure 3 This is a block diagram illustrating the working principle of the online dynamics identification module of the present invention; Figure 4 This is a block diagram illustrating the working principle of the active structure adjustment module of the present invention. Figure 5 This is a flowchart of the control method of the present invention. Detailed Implementation

[0032] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] Example 1 This invention provides a UAV control system that integrates a parallel control path for rapid disturbance suppression. The system includes: a body perception module, a main control module, a reflection arc control module, and a control command fusion module.

[0034] Propriometry Module The body perception module functions to provide real-time, distributed measurement of external forces directly acting on the UAV's airframe structure. Physically, this module consists of one or more sensor arrays. These sensors are pre-integrated or installed at key stress locations on the UAV's airframe structure, such as the upper and lower surfaces of the four or more arms, the central frame, and the motor base. Depending on the specific application requirements, the sensor array can be selected from: Strain gauge array: used to measure the minute deformation, i.e. strain, that occurs when a body structure is subjected to force.

[0035] Pressure sensor array: used to directly measure the pressure distribution generated by external airflow on the surface of the machine body.

[0036] Piezoelectric thin film sensor array: Utilizes the piezoelectric effect to convert dynamic pressure changes or vibrations into electrical signals.

[0037] These sensors collectively form a distributed body sensing matrix (DPSM). This module operates continuously at a preset high sampling frequency and combines multiple sensor readings, which characterize the real-time force conditions at various points on the body, into a mechanical distribution data vector. , as its output.

[0038] Main Control Module (PFC) The main control module is the core of the UAV's conventional flight control. Its function is to generate main control commands to maintain or change the flight attitude and trajectory based on high-level mission instructions (such as waypoint flight and target tracking) and the UAV's overall motion state. The operation of this module mainly includes the following steps: Motion status reception: Receives motion status data from inertial measurement unit (IMU) and traditional sensors of global navigation satellite system (GNSS), including the UAV's three-axis acceleration, angular velocity, absolute geographical location and speed.

[0039] State estimation: The extended Kalman filter (EKF) data fusion algorithm is used to process the received multi-source motion state data to obtain the optimal estimates of the UAV's current attitude, velocity, and position. .

[0040] Feedback control law calculation: The state estimate is used as the basis for the calculation. Desired flight commands from external input Compare and calculate the error vector. Subsequently, the main control command is calculated using a feedback controller (e.g., a proportional-integral-derivative PID controller). In a preferred embodiment, the controller is adaptive, and its control parameters are adjusted in real time based on the output of the online dynamics identification module described later.

[0041] The Reflection Arc Control (RAC) module constitutes one of the core pathways that distinguishes this invention from existing technologies. Its function is to perform proactive feedforward compensation before external disturbances cause macroscopic attitude deviations in the machine. This module is directly connected to the body sensing module, and its workflow is as follows: Input data processing: Receives real-time mechanical distribution data streams output by the body perception module. To capture the spatiotemporal characteristics of the disturbance, this module constructs a historical state matrix from the data within the most recent time window. .

[0042] Perturbation pattern recognition and compensation quantity generation: The core processing unit of this module is a pre-trained temporal neural network model. Examples include Temporal Convolutional Networks (TCNs) or Lightweight Recurrent Neural Networks (RNNs). This network uses a historical state matrix... As input, its output is a parallel vector pair. .

[0043] It is a differential compensation control command. It is an instantaneous, small-amplitude motor speed adjustment, the value and direction of which are trained by a neural network to directly generate a counteracting torque that is the same in magnitude and opposite in direction to the current disturbance torque.

[0044] It is a perturbation feature information vector. It encodes key attributes of the current perturbation, such as the estimated perturbation strength, dominant frequency, and direction of action. This information can be used by other high-level modules in the system.

[0045] Control command fusion module The control command fusion module is the intersection of the main control path and the reflection arc path. It is responsible for merging the commands generated by the two modules to form the final command to drive the motor. Its input terminals are connected to the main control module and the reflection arc control module, respectively. In one specific implementation, this module uses a linear superposition method for command fusion:

[0046] in, It is the final motor control command. It's an instruction from the main control module. This is an instruction from the reflection arc control module. The principle behind this fusion method is that... It is primarily responsible for the macroscopic control of flight trajectory and attitude in the low-frequency band, while This allows for precise and rapid correction of the control signal in the high-frequency band to combat disturbances; the two are complementary in function and frequency domain. Ultimately... After processing, it is converted into a pulse width modulation (PWM) signal and sent to the various electronic speed controllers (ESCs) of the drone.

[0047] Example 2 This embodiment, based on Embodiment 1, further describes the additional modules in the system used to achieve depth adaptation and multi-dimensional adjustment.

[0048] Online Dynamics Identification Module (OIM) To enable the main control module to adapt to changes in the UAV's own state (such as increased load or component aging), this embodiment also includes an online dynamics identification module. This module's function is to estimate the current dynamics model parameters of the UAV in real time. Its input data is multi-source: Main control commands output by the main control module This represents the control torque applied to the system.

[0049] IMU sensor provides the actual motion response of the drone. This represents the system's output.

[0050] Mechanical distribution data collected by the body perception module This data was used to estimate the external disturbance torque acting on the body.

[0051] Internally, this module employs an online parameter identification algorithm using recursive least squares (RLS) with a forgetting factor. By continuously comparing the input-output relationship and using mechanical distribution data to remove disturbance terms from the model, this module can accurately identify a dynamic parameter vector containing information about the UAV's mass, moment of inertia, and aerodynamic drag coefficient online. The identified parameter vector is fed back to the main control module in real time to adjust the gain of its PID controller, thereby achieving adaptive control.

[0052] Active Structural Adjustment Module (ASRM) To cope with persistent or periodic strong disturbances, the system in this embodiment may also include an active structure adjustment module. The input of this module is connected to the reflection arc control module, receiving the disturbance characteristic information vector output by the module. This module contains a decision-making unit that detects... When the perturbation energy exceeds a preset threshold, or the perturbation frequency approaches the known resonant frequency of the organism's structure, the adjustment mechanism will be triggered. At this time, the module will generate and output a structural adjustment command. The command is sent to a variable physical property element integrated into the UAV's airframe structure. This element may be: Variable damping elements: For example, miniature magnetorheological or electrorheological dampers installed at the connection between the arm and the frame or under the motor base. By changing the applied magnetic or electric field, their damping coefficient can be rapidly changed, thereby dissipating vibrational energy at a specific frequency.

[0053] Variable stiffness elements: For example, tensioning mechanisms driven by piezoelectric materials or micro servo motors can change the structural stiffness of a machine arm or frame.

[0054] In this way, the system expands the control dimension from simply the motor power output to the active adjustment of the physical characteristics of the machine body, thereby suppressing or avoiding the impact of specific disturbances at the source.

[0055] Example 3 This invention provides a method for controlling an unmanned aerial vehicle (UAV). This method is implemented through the aforementioned system, and its complete execution flow can be summarized as the following parallel and coordinated steps: Step S1: Parallel acquisition of multimodal data. After the system starts, it continuously and synchronously acquires the UAV's motion state data, expected command data from mission planning, mechanical distribution data of the airframe structure, and current state data of variable structural components.

[0056] Step S2: Dual-loop parallel control processing.

[0057] In the main control loop, the main control command is calculated based on the motion state data and the desired command data using an adaptive feedback control law. .

[0058] In the reflection arc loop, differential compensation control commands are calculated in parallel using a time-series neural network model based on mechanical distribution data. and disturbance feature information .

[0059] Step S3: Online Adaptation and Adjustment.

[0060] The online dynamics identification module runs continuously, integrating main control commands, motion state data, and mechanical distribution data, updating the dynamic parameter vector in real time, and feeding it back to the main control loop to adjust the control law.

[0061] The active structure adjustment module continuously monitors disturbance characteristic information. When the triggering conditions are met, structural adjustment instructions are generated and executed. It alters the physical properties of the organism.

[0062] Step S4: Multi-dimensional instruction fusion and execution. This involves fusing and executing the main control instructions. Differential compensation control command Linear superposition is performed to form the final motor control command. This drives the drone's motors to perform actions. Simultaneously, it transmits structural adjustment commands. Send it to the corresponding variable physical characteristic element for execution.

[0063] By cyclically executing the above steps, this method can achieve rapid feedforward suppression of external disturbances, while ensuring high-precision adaptation of the control system to the time-varying characteristics of the UAV itself. It can also actively adjust the physical characteristics of the airframe to cope with harsh environments, thus systematically improving the overall control performance of the UAV.

[0064] In this embodiment, the body perception module is a fundamental component of the UAV control system of the present invention, enabling it to proactively perceive and suppress external disturbances. Its purpose is to overcome the inherent delay in existing technologies that rely on inertial measurement units (IMUs) for "post-event" feedback control, and instead provide the system with real-time, high-dimensional information about external disturbances by directly measuring the mechanical effects acting on the UAV's structure.

[0065] Specifically, the proprioceptive module is not a single sensor in physical implementation, but a distributed sensor network integrated on the UAV's airframe structure, namely a distributed proprioceptive sensor matrix.

[0066] In a preferred embodiment, the matrix comprises one or more types of micro-sensor arrays strategically positioned on the UAV fuselage at critical locations sensitive to external airflow or subjected to major structural stresses during flight. These locations may include, but are not limited to, the windward upper and leeward lower surfaces of the UAV's various arms, the central frame carrying core electronics, and the area surrounding the motor base connecting to the power source. The specific type of sensor can be selected based on cost, sensitivity, and application scenario, for example: Strain gauge array: When the body undergoes slight deformation due to external airflow or physical contact, the strain gauges placed on the surface of the structure will produce corresponding changes in resistance. By measuring these changes in resistance, the stress and strain distribution at different locations on the body structure can be accurately determined.

[0067] Pressure sensor array: Directly measures the absolute or relative pressure value at a specific point. By arranging these sensors in a matrix on the surface of the aircraft, it is possible to capture in real time the complex airflow field caused by gusts, ground effect, or nearby obstacles, and form a dynamic "pressure cloud map".

[0068] Piezoelectric thin-film sensor array: This type of sensor is extremely sensitive to dynamic pressure changes or high-frequency vibrations, and can directly convert mechanical energy into electrical signals. It is particularly suitable for detecting sudden impacts or structural vibrations caused by eddy currents.

[0069] The body sensing module continuously acquires data at a preset high sampling frequency (preferably on the order of kilohertz). At any sampling moment... This module will come from the entire machine. The measurements from individual sensor nodes are combined to form a high-dimensional mechanical distribution data vector. :

[0070] in, Representing the Each sensing node in The reading at time. This vector. This constitutes a comprehensive, multi-point objective description of the instantaneous force state of the organism.

[0071] The mechanical distribution data vector generated by this module has two key, parallel data flows in the system, demonstrating its core role: Firstly, as the primary data flow direction, The data is transmitted in real time to the Reflection Arc Control (RAC) module described in this invention. To enable the RAC module to understand the temporal evolution characteristics of the disturbance, and not just its instantaneous state, the data input to the module is typically constructed to include past... Historical state matrix of continuously sampled data within the time window :

[0072] in, This refers to the sampling time interval. This data structure enables subsequent processing units (such as temporal neural networks) to identify specific disturbance patterns, such as the dynamic process of a crosswind sweeping across the aircraft from right to left, rather than just a static pressure value. This is the technical prerequisite for achieving "feedforward" compensation and counteracting disturbances before they fully affect the aircraft.

[0073] Secondly, as a supplementary but equally important data flow, It is also transmitted to the Online Dynamics Identification (OIM) module described in the preferred embodiment of the present invention. Its purpose is to provide this module with information about the external disturbance torque. Directly observable or highly accurate estimators. In traditional control systems, The unknown nature of the dynamic parameters, combined with the effects of model uncertainty, makes accurate identification of the UAV's own dynamic parameters difficult. This invention introduces a ontology perception module, enabling the online identification module to effectively decouple the motion response caused by external disturbances from the motion response caused by control commands, thereby significantly improving the accuracy and convergence speed of dynamic parameter identification.

[0074] In summary, the body perception module transforms the UAV's airframe structure itself into a dynamic, distributed mechanical sensor, providing the entire control system with a completely new input dimension independent of traditional motion state perception. It is not only the data foundation for achieving rapid reflective control but also a key information source for achieving high-precision adaptive control.

[0075] In this embodiment, the Primary Flight Control Module (PFC) is one of the core components of the dual-loop parallel architecture of the UAV control system of this invention. Its fundamental purpose is to be responsible for the attitude maintenance and flight trajectory tracking of the UAV over a longer time scale, execute high-level flight commands from the ground station or the onboard mission computer, and constitute a prudent control path for the system to achieve stable flight.

[0076] Specifically, the functionality of this main control module relies on accurate perception of the overall motion state of the UAV and a self-optimizing control law. Its internal workflow can be broken down into the following closely linked stages.

[0077] First, during the data reception and state estimation phase, the module's input is configured to receive multi-source heterogeneous data from airborne conventional navigation sensors. This data primarily includes raw triaxial acceleration measurements provided by the inertial measurement unit (IMU). Compared with triaxial angular velocity measurements ; and the UAV's position information in the global coordinate system provided by a Global Navigation Satellite System (GNSS) receiver. With speed information .

[0078] Since raw sensor data inevitably contains noise and the update frequencies of each sensor differ, a state estimation algorithm is integrated into the main control module to obtain an accurate, continuous, and reliable estimate of the UAV's current state. In a preferred embodiment, this algorithm is an Extended Kalman Filter (EKF). This filter defines the core state of the UAV as a state vector. ,For example:

[0079] in, and These represent the drone's position and velocity in the world coordinate system, respectively. A quaternion used to describe the body's attitude. This refers to the three-axis angular velocities in the body coordinate system. The EKF uses a recursive predict-update loop to adjust the state vector... Perform real-time optimal estimation to obtain the estimated value. .

[0080] Subsequently, in the error calculation and feedback control stage, the main control module uses the state estimate obtained in the previous stage. With externally given expected flight commands Real-time comparisons are performed to calculate the system's current error vector across various control dimensions. :

[0081] The error vector It is the core basis for driving subsequent feedback control.

[0082] To generate control commands based on this error vector, the main control module in this embodiment employs a feedback controller. In a particularly preferred embodiment, this controller is an adaptive proportional-integral-derivative (PID) controller. Its "adaptive" characteristic is one of the key technical features that distinguishes this invention from traditional fixed-gain controllers. Its purpose is to enable the control system to actively adapt to changes in the UAV's own dynamic characteristics, such as changes in mass and moment of inertia caused by the addition of different loads, or changes in aerodynamic characteristics caused by changes in the flight environment.

[0083] Therefore, the gain parameter of this adaptive PID controller, namely the proportional gain matrix... Integral gain matrix and differential gain matrix This is not a pre-set fixed value, but rather it is linked in real time with the Online Dynamics Identification (OIM) module described in Embodiment 2 of this invention. This identification module continuously outputs an updated dynamics parameter vector. This vector represents the current physical characteristics of the UAV. The main control module then adjusts its control gain in real time based on this vector. Its control law can be expressed as:

[0084] In this way, the main control module can always calculate the control quantity based on the most accurate "cognition" of the UAV's own state, thereby ensuring the accuracy and robustness of control under different operating conditions.

[0085] Finally, the module will calculate the control command vector. As its output, this command is a relatively low-frequency control quantity used for macroscopic adjustment, representing the target force and torque required to eliminate flight errors. This output command is transmitted to the control command fusion module described in this invention for merging with the high-frequency differential compensation command from the reflection arc control module.

[0086] In summary, within the overall framework of this invention, the main control module serves as the "brain" for tracking advanced commands and maintaining long-term stable flight. It perceives its own motion through precise state estimation, achieves self-adaptation of its own model through linkage with the online identification module, and ultimately generates macroscopic, goal-oriented main control commands. This forms a complementary parallel control structure with the rapid disturbance suppression function of the reflex arc control module.

[0087] In this embodiment, the Reflex Arc Control Module (RAC) is the innovative core of the UAV control system of this invention. It constitutes a fast feedforward control path, independent of the main control module, for achieving instantaneous disturbance suppression. Its purpose is to directly utilize the mechanical sensing information from the airframe structure itself to pre-generate and apply an adversarial control compensation before external disturbances cause a considerable error in the overall flight attitude of the UAV, thereby simulating the "reflex arc" function in a biological organism.

[0088] Specifically, the input of the reflection arc control module is directly connected to the output of the aforementioned body sensing module, and processes the real-time mechanical distribution data stream from the body sensing module at an extremely high operating frequency. The internal function of this module can be understood as the following sequential processing steps.

[0089] First, in the input data processing stage, this module receives the mechanical distribution data vector output in real time by the body perception module. Considering that a single snapshot of data cannot fully characterize the dynamic evolution of a disturbance (such as the process of a gust of wind sweeping across an aircraft), in a preferred embodiment, the module will use the most recent preset time window. A historical state matrix is ​​constructed by continuously sampling data within the range. .

[0090]

[0091] in, This represents the time interval for data sampling. This is the historical state matrix. As a richer data structure, it not only contains information on the spatial distribution of perturbations (at different sensor locations) but also their temporal trends. This processing method provides the necessary foundation for subsequent accurate pattern recognition.

[0092] Secondly, in the disturbance identification and compensation generation stage, the core processing unit of this module is a pre-trained offline computational model, preferably a time-series neural network model capable of efficiently processing time-series data. Examples include Temporal Convolutional Networks (TCNs) or lightweight Recurrent Neural Networks (RNNs). This model is chosen because it addresses the challenges of handling high-dimensional, complex mechanical distribution data. The mapping relationship between the control and compensation commands for motors is highly nonlinear and difficult to describe accurately using traditional analytical models. Through offline training, this neural network model can learn and solidify this complex mapping relationship.

[0093] During system operation, this neural network model With historical state matrix As input, and in parallel, a pair of two-component output vectors is generated:

[0094] The two components in this output vector pair have clear and distinct technical meanings and uses: First component This is defined as a differential compensation control command. It is an instantaneous and typically small-amplitude motor speed adjustment vector. The dimension of this vector corresponds to the number of motors in the UAV, and the sign and magnitude of each element are trained by a neural network model to precisely generate an antagonistic control effect that can directly counteract the external disturbance force or torque characterized by the current mechanical distribution.

[0095] Second component This is defined as a disturbance feature information vector. This vector is a structured description of the key attributes of the current disturbance, encoding the estimated intensity, dominant frequency, and spatial direction of the disturbance. In this embodiment, this feature vector can be used for system status monitoring; in other preferred embodiments of the invention, it serves as a key basis for driving the active structural adjustment module to make decisions.

[0096] Finally, the reflection arc control module applies its calculated differential compensation control commands. Its main output is transmitted to the control command fusion module described in this invention. The generation and transmission path of this command is completely independent of the main control module, and due to its highly simplified processing flow and reliance on a dedicated calculation model, its end-to-end latency is extremely low.

[0097] In summary, the reflex arc control module plays the role of the UAV's "neural reflex system" within the overall architecture of this invention. By directly analyzing the aircraft's mechanical sensing data, it constructs an ultra-short path from disturbance perception to compensation execution, achieving feedforward suppression of external disturbances. The high-frequency operating characteristics of this module perfectly complement the low-frequency prudent control characteristics of the main control module, jointly ensuring the speed and stability of UAV control.

[0098] In this embodiment, the control command fusion module is a key hub node in the dual-loop parallel control architecture of the UAV control system of the present invention. Its fundamental purpose is to effectively merge commands from two control modules with different functions and operating frequency domains to generate a unified final control signal that can be directly parsed and executed by the physical actuator.

[0099] Specifically, the input terminals of this control command fusion module are connected to the output terminals of the aforementioned main control module (PFC) and reflex arc control module (RAC), respectively. Therefore, this module will synchronously receive two control command vectors of different natures at any given time: One is the main control command from the main control module. This command is a macroscopic control quantity carefully calculated by the system to achieve long-term attitude maintenance and mission trajectory tracking. Its change frequency is relatively low, reflecting the resultant force and resultant torque required by the UAV to achieve the predetermined flight target.

[0100] Secondly, there are differential compensation control commands from the reflection arc control module. This instruction is a micro-compensation quantity generated by the system through a fast feedforward path to counteract instantaneous external disturbances. It changes at an extremely high frequency, but usually has a small amplitude, and is designed to make fine and real-time corrections to the main control instruction.

[0101] The core function of this module is to merge two instruction vectors from different sources into a single final motor control instruction, based on a clear fusion strategy. In a preferred embodiment of the present invention, the fusion strategy employed is direct linear superposition. This fusion process can be precisely described by the following formula:

[0102] The underlying technical principle of this linear superposition strategy lies in the fact that the two control commands are naturally complementary in both the functional and frequency domains. (Main control command) The basic "intent" of the UAV flight and the main power output baseline were established, while the differential compensation control commands... It acts like a high-frequency, adaptive "noise canceller," rapidly and precisely fine-tuning the control baseline without interfering with the basic flight intentions. Because... Its instantaneous and small-amplitude characteristics, combined with its superposition effect, result in fine modulation of the main control signal, thereby achieving a harmonious unity between macroscopic target tracking and microscopic disturbance suppression at the execution level.

[0103] Finally, the control command fusion module calculates the final motor control command. As its output, the dimension of this output vector corresponds to the number of motors on the drone, with each element representing the drive signal strength to be applied to the corresponding motor. Before being sent to the physical actuators, this command vector typically undergoes further processing, namely, being converted into hardware control protocol signals suitable for the various electronic speed controllers (ESCs), such as pulse width modulation (PWM) signals.

[0104] In summary, the control command fusion module plays the role of final decision-making arbitration and command distribution in the overall architecture of this invention. Through an efficient and clearly defined fusion method, it ensures that commands from two parallel control loops can be integrated without conflict. This allows the UAV to simultaneously benefit from the stable planning capabilities provided by the main control module and the agile response capabilities provided by the reflex arc control module, thereby systematically guaranteeing the integrity and effectiveness of the control scheme.

[0105] In this embodiment, the Online Dynamics Identification Module (OIM) is a key technical component for achieving depth adaptive control in the UAV control system of this invention. Its fundamental purpose is to endow the UAV control system with a "self-awareness" capability, enabling it to identify its own dynamic model parameters that change over time in real time and online, thereby providing an accurate and reliable basis for the adaptive adjustment of the upper-level control law.

[0106] This module is necessary because the dynamic characteristics of a UAV are not constant during actual flight missions. For example, battery depletion leads to a continuous reduction in total mass, adding or dropping different payloads causes abrupt changes in mass and moment of inertia, and airframe aging or damage also alters its aerodynamic characteristics. Traditional fixed-parameter controllers inevitably experience a decline in control performance when faced with these changes. This module is designed to address this technical problem.

[0107] In a preferred embodiment, the online dynamics identification module acts as a parallel background processing unit, and its input is configured to receive multi-source information streams from multiple different modules in the system. This is a technical prerequisite for achieving its accurate identification function. First, it receives output commands from the main control module (PFC). This command can be converted into a control torque applied to the UAV body via a known control allocation matrix. This constitutes the "known input" in the identification process.

[0108] Secondly, it receives motion state information from the UAV's inertial measurement unit (IMU) and processes it through a state estimation algorithm, especially the three-axis angular velocities of the aircraft. With triaxial angular acceleration This constitutes the "system response output" in the identification process.

[0109] Thirdly, and also one of the key technical features of this invention, this module receives real-time mechanical distribution data from the body perception module. By processing this data, the unknown disturbance torque exerted on the fuselage by the external environment can be estimated. .

[0110] By introducing the disturbance torque Based on direct observations, this module can effectively decompose the system's motion response into "the part caused by known control inputs" and "the part caused by unknown external disturbances," thereby decoupling the intrinsic dynamic parameters to be identified from the influence of external disturbances. This is the technical guarantee for achieving high-precision online identification.

[0111] The identification process of this module is based on the rigid body dynamics model of the UAV. Taking its rotational motion as an example, this model can be described by Euler's equations:

[0112] in, The moment of inertia matrix is ​​the key dynamic parameter characterizing the rotational properties of the UAV. To facilitate the use of linear identification algorithms, this nonlinear equation can be reconstructed. Assume the moment of inertia matrix... diag is a diagonal matrix Then the above equation can be rearranged to be about the parameter vector to be identified. Linear form: In this standard linear regression model Known as the regression matrix or observation matrix, its elements consist of measurable angular velocities and angular accelerations; while The observation vector consists of known control torques and estimated disturbance torques.

[0113] To solve this equation and update the parameter estimates online, this module integrates a recursive parameter identification algorithm. In a particularly preferred embodiment, Recursive Least Squares (RLS) with a forgetting factor is used. The advantage of this algorithm is that it does not require storing all historical data; it only needs to iteratively update the parameters based on the current measurement and the estimation result from the previous time step, thus having low computational overhead and being very suitable for airborne real-time systems. Its core iterative steps are as follows: first, calculate the prediction error; then, update the gain matrix; and finally, update the parameter estimates and covariance matrix. This is achieved by introducing a forgetting factor less than 1. This algorithm can gradually reduce the weight of old data in parameter estimation, thereby enabling it to effectively track time-varying parameters.

[0114] The final output of this online dynamics identification module is a vector of dynamic parameters that is updated in real time. This output vector is fed back to the main control module (PFC) in real time and continuously. The main control module receives the latest... Then, it will be used to adjust the gain matrix of its internal adaptive PID controller. Based on this, the entire adaptive control closed loop is completed.

[0115] In summary, the online dynamics identification module, by integrating control commands, motion responses, and innovative direct disturbance perception information, constructs a precise and robust self-physical characteristic recognition engine. It transforms the UAV from a static, passively responding control object into an intelligent agent capable of proactively recognizing its own changes and adjusting its behavior. This is the theoretical and technological foundation for the high-precision, highly robust adaptive control achieved in this invention.

[0116] In this embodiment, the Active Structure Regulation Module (ASRM) is an advanced control strategy execution unit employed by the UAV control system of this invention when responding to severe or persistent external disturbances. Its fundamental purpose is to expand control methods from the traditional, purely rotor dynamics-dependent approach to actively adjusting the physical structural characteristics of the UAV itself, thereby fundamentally altering the system's response characteristics to specific disturbances.

[0117] This module was introduced because when a drone encounters persistent strong winds, periodic gusts of a specific frequency, or external excitations close to the structural resonant frequency, relying solely on the main control module and the reflection arc control module for aerodynamic compensation can lead to the motors operating at high loads or even saturation for extended periods, or failing to effectively suppress harmful vibrations caused by resonance. This module aims to address these issues from a completely new perspective by altering the physical properties of the drone's airframe.

[0118] Specifically, the input of the active structure adjustment module is connected to the output of the Reflection Arc Control (RAC) module described in this invention, and is specifically designed to receive the disturbance feature information vector generated in parallel by the module. This vector is a quantitative description of the essential attributes of the current disturbance (such as estimated intensity, dominant frequency, and direction of action), and constitutes the sole and sufficient information basis for this module's decision-making.

[0119] The core of this module is a decision-making unit. This unit operates continuously at a low frequency, processing the input perturbation feature information vector. Real-time monitoring and analysis are performed. In a preferred embodiment, the decision-making unit has a pre-set set of trigger rules. When a detected disturbance meets one or more conditions, the adjustment mechanism is activated. These conditions may include: Disturbance energy exceeding limits: When the integral or average value of the disturbance intensity or energy represented by the disturbance feature vector exceeds a preset safety threshold over a period of time, it indicates that the UAV is in a harsh flight environment.

[0120] Imminent risk of resonance: when the dominant frequency decoded from the perturbation eigenvector Gradually approaching one or more known inherent natural resonant frequencies of the UAV structure When, that is, the condition is met. ,in This represents a very small frequency safety margin. This is a crucial forward-looking judgment for avoiding catastrophic structural failure.

[0121] Perturbation pattern solidification: When the perturbation feature vector exhibits a stable, non-transient pattern over a period of time, such as persistent crosswinds or periodic ground effects, it indicates that the perturbation source is environmental rather than accidental and is suitable for structural adaptation.

[0122] Once any or a combination of the above conditions are met, the decision-making unit will generate and output a structural adjustment instruction. This instruction is a vector encoding the adjustment target and the adjustment amount, which will be sent to one or more variable physical characteristic elements pre-integrated into the UAV's airframe structure.

[0123] The variable physical characteristic elements are the physical basis of this invention; they are electromechanical devices capable of responding to electrical signals and changing their own physical parameters. Different types of elements may be used in different embodiments of this invention: Variable damping element: Preferably, a miniature magnetorheological (MR) or electrorheological (ER) damper can be installed at the connection between the boom and the central frame, or on the landing gear or motor base, along a critical vibration transmission path. When structural adjustment commands are issued... When a specific magnetic or electric field is applied, the viscosity coefficient of the fluid inside these dampers changes on a millisecond scale, thus precisely altering the damping ratio of the structure. This adjustment method is particularly suitable for dissipating vibrational energy at specific frequencies to suppress resonance.

[0124] Variable stiffness element: Preferably, this can be achieved by embedding a tensioning mechanism driven by piezoelectric material or a micro servo motor inside the key load-bearing components of the arm or frame. Structural adjustment command. These components can be driven to generate internal stress or change preload, thereby macroscopically altering the effective stiffness of the parts. The main purpose of this adjustment method is to actively change the inherent resonant frequency of the body structure. When a disturbance frequency is detected to be close to a certain resonant frequency, it can be "pushed away" by adjusting the stiffness, thereby avoiding resonance.

[0125] The module's output instructions With the final motor control command They operate on two completely different levels. The former adjusts the "factory settings" of the drone as a physical system, while the latter adjusts the instantaneous power output of the system under the current settings.

[0126] In summary, the active structural adjustment module constitutes a deeper, longer-term adaptive control closed loop within the overall architecture of this invention. This transforms the UAV from a passive, physically fixed controlled object into an intelligent system capable of proactively "self-reconfiguring" based on environmental feedback. Consequently, it exhibits adaptability and survivability exceeding traditional control frameworks when facing extreme or persistently harsh environments.

[0127] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A drone control system, characterized in that, include: The body perception module is installed on the UAV's body structure and is used to collect mechanical distribution data at multiple locations on the body structure in real time. The main control module is used to generate main control commands to maintain the flight attitude and trajectory of the UAV based on the motion state data of the UAV. The reflection arc control module, connected to the body sensing module, is used to generate differential compensation control commands to counteract external disturbances based on the mechanical distribution data. The control command fusion module is connected to both the main control module and the reflection arc control module, and is used to fuse the main control command with the differential compensation control command to obtain the final motor control command.

2. The system according to claim 1, characterized in that, The body sensing module includes at least one of a pressure sensor array, a strain gauge array, or a piezoelectric thin film sensor array disposed on the body structure.

3. The system according to claim 1, characterized in that, The reflection arc control module incorporates a time-series neural network model for identifying disturbance patterns in the mechanical distribution data.

4. The system according to claim 1, characterized in that, Also includes: The online dynamics identification module is used to identify the dynamics model parameters of the UAV online based on the output of the main control module, the motion state data of the UAV, and the mechanical distribution data collected by the body perception module.

5. The system according to claim 4, characterized in that, The main control module is connected to the online dynamics identification module, and adaptively adjusts its control law according to the dynamics model parameters identified by the online dynamics identification module.

6. The system according to claim 3, characterized in that, The temporal neural network model is also used to output disturbance feature information characterizing the intensity, frequency, or direction of external disturbances based on the mechanical distribution data.

7. The system according to claim 6, characterized in that, Also includes: An active structural adjustment module is used to generate structural adjustment commands for controlling variable physical characteristic elements in the UAV body structure based on the disturbance characteristic information.

8. The system according to claim 7, characterized in that, The variable physical property element includes at least one of a variable damping element or a variable stiffness element.

9. The system according to claim 1, characterized in that, The control command fusion module linearly superimposes the differential compensation control command and the main control command to form the final motor control command.

10. A method for controlling an unmanned aerial vehicle (UAV), characterized in that, Includes the following steps: By using a body perception module installed on the drone's body structure, mechanical distribution data at multiple locations on the body structure can be collected in real time. The main control module generates main control commands to maintain the flight attitude and trajectory of the UAV based on the motion state data of the UAV. The reflection arc control module generates differential compensation control commands to counteract external disturbances based on the mechanical distribution data. The main control command and the differential compensation control command are fused together to obtain the final motor control command, which is then used to control the UAV.

Images