Intelligent cooperative control system and control method of deformable folding machine arm structure of multi-rotor unmanned aerial vehicle
Patent Information
- Application Number
- CN202611061865.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-16
- Publication Date
- 2026-10-13
Smart Images
Figure CN122902665A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to an intelligent collaborative control system and control method for a deformable and foldable arm structure of a multi-rotor unmanned aerial vehicle (UAV), and particularly to the fields of multi-rotor UAV technology and intelligent control technology. Background Technology
[0002] Multirotor drones have been widely used in aerial photography, surveying, logistics delivery, and emergency rescue due to their superior performance in vertical takeoff and landing, hovering, and low-speed maneuverability. However, the arm length and rotor position of traditional multirotor drones are fixed before takeoff and cannot be changed during flight. This inherent limitation severely restricts the drone's ability to navigate narrow spaces.
[0003] In recent years, academia and industry have begun to focus on the research of deformable multi-rotor drones. Patent applications for drones with automatically folding arms have been filed, aiming to reduce storage volume through automated longitudinal folding. However, existing technologies still have the following shortcomings: First, the folding / unfolding process relies on ground operation or simple timing control, lacking dynamic deformation capabilities during flight and unable to actively adjust the arm configuration according to environmental changes; second, the coordinated deformation of multiple arms lacks synchronization and consistency guarantees, leading to center of gravity shift, attitude jitter, and even loss of control; third, the changes in dynamic characteristics during deformation (such as moment of inertia and center of gravity position) lack real-time modeling and adaptive control, and existing fixed-parameter controllers cannot track parameter changes, resulting in significant attitude tracking errors during rapid deformation; fourth, there is a lack of intelligent decision-making capabilities for flight strategies based on arm state perception, and the system cannot autonomously determine whether deformation is necessary or which deformation mode to select. Summary of the Invention
[0004] This invention provides an intelligent collaborative control system and control method for a deformable and foldable arm structure of a multi-rotor unmanned aerial vehicle (UAV) to overcome the shortcomings of existing technologies, such as the inability to dynamically deform during flight, poor synchronization of multi-arm collaboration, loss of control of attitude disturbances during deformation, and lack of intelligent decision-making.
[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention discloses an intelligent collaborative control system and method for a deformable and foldable arm structure of a multi-rotor unmanned aerial vehicle (UAV), comprising: determining a target arm configuration based on the UAV's flight mission information and / or perceived environmental information; controlling each arm of the UAV to collaboratively operate towards its corresponding target angle position; performing attitude control on the UAV based on its current attitude information during the collaborative arm operation; performing master-slave synchronization control and correcting synchronization deviations based on real-time angle feedback from each arm during the collaborative arm operation and attitude control; identifying UAV dynamic parameters that change due to the arm configuration change online during the collaborative arm operation and attitude control, and adaptively adjusting the attitude control law based on the dynamic parameters; wherein, determining the target arm configuration includes: making a matching decision based on the environmental information and / or the flight mission information according to a pre-established arm configuration library to determine the target arm configuration.
[0006] Furthermore, the control of the coordinated operation of each arm includes: the central controller generating an S-shaped acceleration and deceleration target angle trajectory and broadcasting it to the drive unit of each arm, and each drive unit performing local closed-loop control to track the target angle trajectory.
[0007] Furthermore, the local closed-loop control is a position-speed-current three-loop PID control; the execution of master-slave synchronization control and correction of synchronization deviation includes: the central controller calculates the angle deviation between each arm in real time, and when the angle deviation exceeds a preset threshold, outputs an acceleration fine-tuning signal to the drive unit of the lagging arm and an acceleration fine-tuning signal to the drive unit of the leading arm.
[0008] Furthermore, the online identification employs the recursive least squares (RLS) algorithm; the adaptive adjustment of the attitude control law employs the parameter adaptive active disturbance rejection control (ADRC) algorithm.
[0009] Furthermore, the forgetting factor in the recursive least squares algorithm is dynamically adjusted based on the sliding variance of the current information residual and the deformation rate of the robotic arm.
[0010] As one implementation method, the formula for calculating the forgetting factor λ(k) is: λ(k)=λ_min + (λ_max - λ_min)·exp(-β·σ²(k) - γ·v_m(k)), where σ²(k) is the moving variance of the new residual, v_m(k) is the deformation rate, and λ_max, λ_min, β, and γ are preset coefficients.
[0011] Furthermore, the arm configuration library includes at least one of the following: fully deployed configuration, semi-folded crossing configuration, fully folded storage configuration, and asymmetric folding wind-resistant configuration; the environmental information is obtained based on at least one of the following: the width of the passage ahead, the current wind speed, and the remaining battery power of the UAV.
[0012] As one implementation, the matching decision based on the environmental information and / or the flight mission information includes: calculating the deformation demand integral S(k) by integrating the trigger indicators of each sensing channel; updating the deformation demand integral S(k) recursively through an integrator with an asymmetric hysteresis function; and outputting a valid deformation request when S(k) is greater than the activation threshold.
[0013] Specifically, the asymmetric hysteresis function H(δ(k)) is defined as follows: when S(k-1)>S_on, H(δ(k))=-h_off; when S(k-1)≤S_on, H(δ(k))=h_on; where δ(k) is the trigger deviation, h_on>h_off>0, and S_on is the activation threshold.
[0014] In one implementation, at least one folding joint of the arm includes at least one of a wear adaptive compensation structure, an off-center load torque balancing structure, and an integrated buffer locking structure.
[0015] Furthermore, the wear-adaptive compensation structure includes a locking pin with a variable cross-section helical groove and a radially micro-moving elastic bushing.
[0016] Furthermore, the off-center load torque balancing structure includes a pair of symmetrically arranged wedge sliders and a double-headed wedge block disposed therebetween.
[0017] Furthermore, the integrated buffer locking structure includes a rotating chuck with a variable cross-section U-shaped buffer groove and a cooperating elastic pawl.
[0018] In one implementation, the control is executed through a central controller, which is connected to the modular drive units of each arm; the modular drive unit includes a brushless DC servo motor, a high-precision magnetic encoder, a current sensor, and a microcontroller.
[0019] Furthermore, the central controller communicates with the microcontrollers of each drive unit via a CAN bus.
[0020] This application provides an intelligent collaborative control device for the arms of a multi-rotor unmanned aerial vehicle (UAV), comprising: a configuration decision module for determining a target arm configuration based on the UAV's flight mission information and / or perceived environmental information; wherein, determining the target arm configuration includes: making a matching decision based on the environmental information and / or the flight mission information according to a pre-established arm configuration library to determine the target arm configuration; a collaborative drive module for controlling each arm of the UAV to collaboratively operate towards its corresponding target angle position; an attitude control module for performing attitude control on the UAV based on the current attitude information of the UAV during the collaborative operation of the arms; a synchronization correction module for performing master-slave synchronization control and correcting synchronization deviations based on real-time angle feedback of each arm during the collaborative operation of the arms; and an adaptive control module for identifying UAV dynamic parameters changed due to changes in arm configuration during the collaborative operation and attitude control of the arms, and adaptively adjusting the attitude control law based on the dynamic parameters.
[0021] This application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement an intelligent collaborative control method for a deformable and foldable arm structure of a multi-rotor unmanned aerial vehicle.
[0022] This application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements an intelligent collaborative control method for a deformable and foldable arm structure of a multi-rotor unmanned aerial vehicle.
[0023] The beneficial effects achieved by this invention are as follows: By matching the target arm configuration with a pre-established arm configuration library based on flight mission information and / or perceived environmental information, the UAV can autonomously decide on the deformation mode according to environmental changes during flight, solving the problem of lack of dynamic deformation capability and intelligent decision-making capability in existing technologies; By controlling the coordinated operation of each arm to perform master-slave synchronous control and correcting synchronization deviations, high-precision consistency of multi-arm deformation is ensured, avoiding centroid shift and attitude jitter caused by asynchrony; By identifying the changing UAV dynamic parameters during deformation online and adaptively adjusting the attitude control law, attitude disturbances caused by deformation are effectively suppressed, and attitude deviations are controlled within ±3°, which is significantly better than traditional open-loop deformation schemes. Attached Figure Description
[0024] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the present invention. Detailed Implementation
[0025] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention. Example
[0026] Explanation of terms: Target arm configuration: refers to the desired arm shape of the drone during flight, such as X-shaped fully extended, H-shaped fully extended, semi-folded (crossing type), fully folded (storage type), or asymmetrical folding.
[0027] Arm Configuration Library: A pre-built dataset that stores various typical arm configurations and their corresponding characteristic parameters, applicable scenarios, and optimal control parameters to assist in decision-making.
[0028] Master-slave synchronous control: a cooperative control strategy in which one controller (central controller) acts as the "master station" to generate and broadcast unified motion commands (such as target angle trajectory), and multiple controlled objects (each arm drive unit) act as "slave stations" to receive commands and independently execute local closed-loop tracking.
[0029] Synchronization deviation: The difference between the actual angular position of each arm and the target position when multiple arms move in coordination, or the difference in actual position between any two arms.
[0030] Recursive Least Squares (RLS) algorithm: an online parameter estimation algorithm that can recursively update the estimated values of system model parameters using the latest measurement data during system operation, without storing all historical data.
[0031] Parameter Adaptive Active Disturbance Rejection Control (ADRC) Algorithm: An advanced control algorithm whose core is to estimate and compensate for the system's "total disturbance" (including model uncertainty and external disturbance) in real time through an extended state observer (ESO), and to improve control performance by combining adaptively adjustable control parameters such as proportional / derivative gain.
[0032] Asymmetric hysteresis function: A mathematical function with unequal "trigger threshold" and "release threshold" is often used in debouncing logic to make state switching have different sensitivities when triggered in the forward and reverse directions.
[0033] Modular drive unit: refers to an independent functional unit deployed near each folding joint, which integrates drive motor, sensor and local controller, and is responsible for performing the folding / unfolding action of the arm.
[0034] CAN bus: Controller Area Network bus, a serial communication protocol widely used in automotive, industrial control and other fields, characterized by high reliability and real-time performance.
[0035] like Figure 1 As shown, an intelligent collaborative control system and control method for a deformable and foldable arm structure of a multi-rotor unmanned aerial vehicle (UAV) includes: S101: Determine the target arm configuration based on the UAV's flight mission information and / or perceived environmental information; S102: Control each arm of the UAV to move in coordination towards its corresponding target angle position; S103: During the coordinated operation of the robotic arm, the attitude control of the UAV is performed based on the current attitude information of the UAV. S104: During the coordinated operation of the robotic arms, based on the real-time angle feedback of each arm, master-slave synchronization control is executed and synchronization deviations are corrected. S105: During the coordinated operation and attitude control of the arm, the dynamic parameters of the UAV that change due to the change in the arm configuration are identified online, and the attitude control law is adaptively adjusted based on the dynamic parameters. Specifically, step S101 includes: making a matching decision based on the environmental information and / or the flight mission information according to a pre-established arm configuration library, so as to determine the target arm configuration.
[0036] In this embodiment, the above steps work together to form a complete intelligent closed loop of "decision-execution-control-feedback". Step S101 (configuration decision layer) intelligently selects the most suitable target configuration from the arm configuration library based on real-time perception and task requirements, solving the problem of "when to change and what to change to". Step S102 (cooperative execution layer) drives each arm to move collaboratively towards the target position. At the same time, steps S103 (basic attitude control layer) and S105 (adaptive control layer) jointly ensure flight stability during the deformation process. S105 effectively suppresses attitude disturbances caused by configuration changes by identifying dynamic parameters (such as moment of inertia and center of gravity position) online and adaptively adjusting the control law. Step S104 (synchronization control layer) synchronously monitors and fine-tunes the movement of each arm based on real-time angle feedback to ensure the consistency and accuracy of the deformation action and avoid center of gravity shift caused by asynchrony. These four levels are closely coupled, enabling the UAV to actively, smoothly, and precisely change the shape of its arms during flight. Together, they solve the technical problems of existing technologies, such as the inability to dynamically deform during flight, poor synchronization of multi-arm coordination, loss of control of attitude disturbances during deformation, and lack of intelligent decision-making. This achieves a comprehensive technical effect of intelligent decision-making, high-precision coordination, and adaptive stability.
[0037] Taking a specific application scenario as an example, suppose a quadcopter drone is performing a reconnaissance mission between urban buildings, and a narrow passage slightly narrower than its fully extended wingspan appears ahead. At this time, the binocular vision and lidar installed on the drone perceive this environmental information (the width of the passage ahead is less than 1.2 times the current wingspan of the drone). The central controller (corresponding to the configuration decision module in subsequent embodiments) receives this environmental information and accesses the pre-stored arm configuration library. After matching and decision-making (step S101), it determines that it needs to switch to the "half-folding crossing configuration" and calculates the target angle position corresponding to each arm (e.g., the front arm needs to be folded back 45°). Subsequently, the central controller generates an S-shaped acceleration and deceleration target angle trajectory and broadcasts it to each arm (the initial action of step S102). During this transformation process, the attitude controller (S103) continuously receives attitude information from the IMU to maintain basic flight. Simultaneously, the RLS algorithm (the online identification part in S105) uses the real-time angular position and motor speed commands of each arm to update the UAV's rotational inertia matrix online. Based on these updated parameters, the parameter adaptive ADRC algorithm (the adaptive adjustment part in S105) adjusts the attitude controller's gain and observer bandwidth in real time to suppress attitude disturbances caused by the transformation. The central controller also receives angle feedback reported by each arm drive unit in real time. Once it detects that the angular deviation between two arms exceeds 2° (a preset threshold), it sends an acceleration fine-tuning signal to the lagging arm drive unit and a deceleration fine-tuning signal to the leading arm, completing the synchronization deviation correction (step S104). When all arms accurately reach the target angular position, the transformation is complete, the UAV's wingspan decreases by approximately 30%, and it successfully traverses the narrow passage.
[0038] In one embodiment, step S102, which involves controlling each arm to coordinate its movement towards its corresponding target angle position, specifically includes: the central controller generating an S-shaped acceleration / deceleration target angle trajectory and broadcasting it to the drive units of each arm; each drive unit then performing local closed-loop control to track the target angle trajectory. Using an S-shaped acceleration / deceleration trajectory ensures continuous differentiability of acceleration, avoiding impact torque caused by abrupt changes in velocity or acceleration, resulting in a smoother arm deformation process and laying a good foundation for subsequent attitude stabilization control. Furthermore, the target angle trajectory is calculated and generated by the central controller based on the target angles of each arm corresponding to the target arm configuration, taking into account the smoothness and efficiency of the deformation process. As a specific implementation, a sine function or polynomial curve can be used for smooth interpolation calculation of the S-shaped trajectory, ensuring that its time derivative (velocity) and second derivative (acceleration) are continuous throughout the trajectory.
[0039] In one embodiment, the aforementioned local closed-loop control is a position-velocity-current three-loop PID control. The execution of master-slave synchronization control and correction of synchronization deviation includes: the central controller calculates the angular deviation between each arm in real time, and when the angular deviation exceeds a preset threshold, outputs an acceleration fine-tuning signal to the drive unit of the lagging arm and a deceleration fine-tuning signal to the drive unit of the leading arm. Through the position-velocity-current three-loop PID control, each drive unit can perform high-precision, high-response local tracking of the target trajectory. The central controller calculates the synchronization deviation between the arms by comparing the real-time angular feedback from each drive unit. When the synchronization deviation exceeds the threshold, it sends a fine-tuning signal to the designated drive unit to adjust the velocity feedforward or target value of its local position loop, thereby achieving "catching up" with the lagging arm and "waiting" for the leading arm, thus maintaining global synchronization. This active correction mechanism can effectively prevent asynchronous deformation caused by differences in motor performance, uneven load, or accumulated communication delays, ensuring that the centroid offset is controlled within a very small range during deformation and avoiding additional disturbances to the posture.
[0040] In one embodiment, for the aforementioned online identification and adaptive adjustment of the attitude control law (step S105), the online identification employs the Recursive Least Squares (RLS) algorithm; the adaptive adjustment of the attitude control law employs the Parametric Adaptive Active Disturbance Rejection Control (ADRC) algorithm. The RLS algorithm enables real-time online estimation of UAV dynamic parameters (such as the moment of inertia matrix), allowing the system model to follow the dynamic changes in the arm configuration. Based on this real-time updated model, the ADRC algorithm dynamically adjusts the observation bandwidth of its Extended State Observer (ESO) and the proportional and differential gains of the attitude control loop, ensuring that the controller maintains optimal disturbance estimation capability and response speed under different configurations. This combination allows the attitude controller to adapt to the deformation process, compensating in real-time for internal disturbances such as inertial forces and centrifugal forces caused by configuration changes, as well as external wind disturbances, thereby controlling the attitude angle deviation during deformation within ±3°, significantly outperforming traditional schemes using fixed-parameter controllers.
[0041] In one embodiment, for the aforementioned recursive least squares algorithm, the forgetting factor in the algorithm is dynamically adjusted based on the sliding variance of the current information residual and the arm deformation rate. Traditional fixed-forgetting-factor RLS algorithms are prone to parameter identification lag or divergence when the deformation rate changes abruptly or noise increases. By introducing a dynamic adjustment mechanism for the forgetting factor, the forgetting factor can be automatically reduced (e.g., lowered to the lower limit of 0.90) when deformation is drastic or the model changes abruptly (when the variance of the information residual increases and the deformation rate is high), accelerating the forgetting of old data and enabling the algorithm to respond quickly to current changes; in steady state or when noise is high, the forgetting factor can be automatically increased (e.g., increased to the upper limit of 0.999), improving the smoothness and stability of identification. This adjustment optimizes the convergence speed and accuracy of identification, providing a more accurate and timely dynamic model for the parameter adaptive ADRC controller, thereby helping to further reduce attitude control deviation.
[0042] Specifically, the forgetting factor λ(k) is calculated as follows: λ(k) = λ_min + (λ_max - λ_min)·exp(-β·σ²(k) - γ·v_m(k)), where σ²(k) is the moving variance of the innovation residual, v_m(k) is the deformation rate, and λ_max, λ_min, β, and γ are preset coefficients. σ²(k) reflects the fluctuation of the model prediction error, and v_m(k) directly characterizes the drastic change in the arm's motion state. The exponential function-based dynamic adjustment allows λ(k) to respond smoothly and non-linearly to changes in these two key indicators, avoiding parameter abrupt changes. Based on this formula, real-time calculation is performed, updating λ(k) once every control cycle (e.g., 1ms) and substituting it into the standard RLS recursive formula to achieve adaptive forgetting factor. This design significantly improves the robustness and accuracy of parameter identification during deformation.
[0043] In one embodiment, the aforementioned arm configuration library includes at least one of the following: fully deployed configuration, semi-folding crossing configuration, fully folding storage configuration, and asymmetric folding wind-resistant configuration. The environmental information is obtained based on at least one of the following: the width of the forward passage, the current wind speed, and the remaining battery power of the UAV. The pre-established configuration library provides rich alternatives and prior knowledge for intelligent decision-making. For example, the "semi-folding crossing configuration" is suitable for scenarios where the width of the forward passage is less than 1.2 times the current wingspan; the "asymmetric folding wind-resistant configuration" is suitable for scenarios where the current wind speed exceeds a safety threshold, requiring a reduction in the frontal area; and the "fully folding storage configuration" may be triggered by energy-sensing scenarios where the remaining battery power is below 20% and an emergency landing is required. The multi-source environmental information assessment makes deformation decisions more comprehensive and aligned with actual needs, ensuring that the UAV can make the most appropriate deformation response in complex and changing environments, thereby significantly enhancing mission adaptability and flight safety. As a specific implementation, each configuration in the configuration library has been verified through dynamic simulation and associated with preset optimal control parameters such as S-shaped trajectory parameters and ADRC control gain.
[0044] In one embodiment, step S101 (determining the target arm configuration by matching environmental information and / or flight mission information) specifically includes: calculating the deformation demand integral S(k) by integrating trigger indicators from various sensing channels; updating the deformation demand integral S(k) recursively through an integrator with an asymmetric hysteresis function; and outputting a valid deformation request when S(k) is greater than the activation threshold. By employing the integral and hysteresis function mechanism, the impact of instantaneous noise or minor environmental fluctuations in sensing measurements on the decision-making process can be effectively filtered, preventing frequent jittering and switching of deformation commands near the trigger threshold. By calculating the deformation demand integral S(k) and controlling its growth and decay process, deformation actions are only triggered (or revoked) when environmental conditions continuously and stably meet (or release) the deformation demand. This makes deformation decisions more robust and reliable, avoiding frequent start-stop of arm motors due to sensor noise and reducing unnecessary mechanical wear and energy consumption.
[0045] Specifically, the asymmetric hysteresis function H(δ(k)) is defined as follows: when S(k-1) > S_on, H(δ(k)) = -h_off; when S(k-1) ≤ S_on, H(δ(k)) = h_on; where δ(k) is the trigger deviation, h_on > h_off > 0, and S_on is the activation threshold. The asymmetry is reflected in h_on > h_off, meaning that transitioning from an inactive state to an active state (triggering deformation) requires a stronger positive deviation stimulus (higher h_on), while exiting an active state (removing deformation) only requires a smaller negative deviation (smaller h_off). This design makes the system more cautious when "hesitating whether to deform" (preventing false triggers) and more sensitive when "deciding to end deformation" (quickly withdrawing from the deformation state and restoring mobility). It effectively solves the decision jitter problem and improves the stability of the intelligent decision-making layer.
[0046] In one embodiment, in the physical implementation of the above method, at least one arm's folding joint includes at least one of a wear-adaptive compensation structure, an off-center torque balancing structure, and an integrated buffer locking structure. Introducing these mechanical structural optimizations can improve the long-term accuracy and reliability of the deformation motion from a physical perspective. The wear-adaptive compensation structure automatically compensates for increased clearance due to long-term use, maintaining angular feedback accuracy; the off-center torque balancing structure automatically balances asymmetrical loads during deformation, preventing unilateral jamming or unstable movement; and the integrated buffer locking structure absorbs impact energy and reliably locks the arm when it reaches its position, improving positioning accuracy and safety. These structures work together to ensure that the MACM (Multi-Arm Cooperative Deformation Actuator) can operate stably and accurately throughout its entire lifespan, providing a crucial physical foundation for achieving high-precision synchronous deformation.
[0047] In one embodiment, the aforementioned wear-adaptive compensation structure includes a locking pin with a variable cross-section helical groove and a radially micro-moving elastic bushing. The locking pin assembly features a variable cross-section helical groove whose helix angle gradually changes from root to head. This allows the slight axial movement of the locking pin under the retaining thrust after wear to be converted into radial contraction of the bushing through the interaction between the helical groove and the bushing's raised ribs, thus automatically compensating for the increased clearance due to wear. A wave spring on the outer side of the elastic bushing provides preload. This purely mechanical feedback and compensation mechanism, without the need for sensors or electronic control, can maintain the joint radial clearance within a very small range (e.g., ±0.02 mm) over a long period. This ensures that the angular position read by the magnetic encoder accurately reflects the actual mechanical angle of the arm, which is crucial for maintaining high synchronization accuracy.
[0048] In one embodiment, the aforementioned off-center load torque balancing structure includes a pair of symmetrically arranged wedge sliders and a double-headed wedge-shaped lever disposed between them. When the arm experiences off-center loading due to aerodynamic asymmetry or other reasons, such as an increase in load on the right side, the right slider transmits pressure to the double-headed wedge-shaped lever through the inclined plane, forcing it to compress the spring and move to the left. Simultaneously, the lever's leftward movement pushes the left slider outward, increasing the positive pressure on the left side, thereby redistributing the load between the two sliders and achieving instantaneous mechanical balance. At the same time, the needle rollers reduce sliding friction. This structure is a purely mechanical closed-loop system with a rapid response, effectively eliminating the risk of unilateral jamming caused by off-center load torque, ensuring a smooth and stable deformation process, reducing the requirements for the drive motor to overcome additional friction, and improving drive efficiency.
[0049] In one embodiment, the integrated buffer locking structure mentioned above includes a rotating chuck with a variable cross-section U-shaped buffer groove and a cooperating elastic pawl. When the arm approaches its limit position, the teeth of the elastic pawl enter the U-shaped groove and push the rotating chuck to rotate against the force of the helical spring. The teeth climb along the gradually rising slope, converting the impact kinetic energy into rotational potential energy and spring elastic potential energy, thus achieving buffering. At the moment of reaching the limit position, the teeth slide past the highest point and quickly fall into the locking tooth groove under the action of the spring reaction force, completing the mechanical locking. The integrated design integrates the three functions of buffering, positioning, and locking into a single rotating component, resulting in a compact and reliable structure. The nonlinear damping characteristics effectively reduce the impact force at the limit position, and the locking position accuracy is high (up to ±0.1°), providing precise end position assurance for attitude stability.
[0050] In one embodiment, the aforementioned coordinated control can be executed through a central controller, which can be connected to the modular drive units of each arm. These modular drive units may include brushless DC servo motors, high-precision magnetic encoders, current sensors, and microcontrollers. Each modular drive unit is an independent mechatronic node. The brushless DC servo motor provides driving force; the high-precision magnetic encoder (e.g., with a resolution of 0.1°) provides real-time feedback on the absolute angle of the arm; the current sensor monitors the motor current and can be used for torque monitoring and anomaly detection; the microcontroller (MCU) is responsible for executing local three-loop PID control, processing sensor data, and communicating with the central controller. This modular, distributed architecture improves the system's scalability, reliability, and ease of maintenance.
[0051] As a further implementation, the central controller and the microcontrollers of each drive unit can communicate via a CAN bus. The CAN bus is characterized by high reliability and real-time performance, supports multi-master communication and broadcast modes, and is very suitable for scenarios where the central controller broadcasts the target trajectory to all drive units, and each drive unit periodically reports status information. It is key to achieving efficient and reliable collaborative communication.
[0052] It should be noted that in some optional implementations, the modular drive unit mentioned above can also integrate local state detection and fault tolerance functions. For example, the MCU of the drive unit can acquire the angle readings of the magnetic encoder and the phase current of the motor in real time. To combat sensor noise or transient faults, the MCU can run a discrete Kalman filter to smooth the magnetic encoder angle signal and use the statistical characteristics of the innovation sequence (observation residuals) for fault diagnosis. Specifically, a sliding window of length 20 is set, and the variance of the innovation sequence within the window is calculated. If the absolute value of the innovation for three consecutive control cycles exceeds three times its variance, it is determined that the magnetic encoder may have a transient jump fault. At this time, the MCU can stop using the encoder observations to update the Kalman filter and switch to degraded position estimation based on the motor armature current and back EMF model to ensure the continuity and reliability of position feedback during sensor anomalies and avoid generating erroneous synchronization deviation correction commands.
[0053] In some optional implementations, the recursive least squares algorithm for dynamically adjusting the forgetting factor can be refined as follows: In each control cycle (e.g., 1 ms), first calculate the prediction residual of the parameter estimate from the previous cycle, i.e., the innovation e(k). Then, within a sliding window (length N can be 20~50), calculate the mean and variance σ²(k) of e(k). Simultaneously, calculate the deformation rate v_m(k) based on the difference in arm angle between the current and previous cycles. Substitute σ²(k) and v_m(k) into the formula λ(k) = λ_min + (λ_max - λ_min)·exp(-β·σ²(k) - γ·v_m(k)) to calculate the adaptive forgetting factor λ(k) for the current cycle. Here, λ_max can be 0.999, λ_min can be 0.90, and β and γ are adjustment coefficients, for example, 0.5 and 0.1 respectively; their specific values can be determined through ground calibration experiments. Finally, the calculated λ(k) is used to update the gain matrix and covariance matrix of the standard RLS algorithm, and then the updated dynamic parameter estimation vector θ̂(k) is obtained recursively. This process runs in real time, providing a model basis for adaptive control parameters.
[0054] In some optional implementations, the calculation of the integral of deformation demand may specifically include: For each control cycle, acquiring the normalized trigger index d_j(k) of each sensing channel (e.g., space, task, state, energy). Taking the minimum value as the comprehensive trigger index d(k). Calculating the trigger deviation δ(k) = d(k) - d_base, where d_base is typically set to 1.0. Based on the relationship between the integral of deformation demand S(k-1) of the previous cycle and the activation threshold S_on, determining the hysteresis function value H(δ(k)) to be used currently: if S(k-1) is greater than S_on, then H(δ(k)) = -h_off; otherwise, H(δ(k)) = h_on. Here, h_on can be 0.05-0.2, h_off can be 0.02-0.1, and h_on > h_off. Then, the integral is updated using the formula S(k) = max{0, S(k-1) +K_i[δ(k) - H(δ(k))]}, where K_i is the integral coefficient, and its value can be 0.001 to 0.01. When S(k) exceeds the activation threshold S_on (e.g., set to 5080), the deformation request is deemed valid and locked; when S(k) is lower than the release threshold S_off (e.g., set to 0.5S_on), the deformation request is revoked.
[0055] In some optional implementations, the dynamic parameters specifically targeted by the above online identification process may include the UAV's moment of inertia matrix I(t) and center of gravity position vector r_g(t). The identification algorithm utilizes a mathematical model established by multi-rigid-body dynamics equations, taking the real-time angles of each arm and motor commands as inputs, and continuously approximating the parameter values under the current actual configuration through recursive calculations. The update frequency can be consistent with the flight control main loop, such as 1 kHz.
[0056] In some optional implementations, the resolution of the high-precision magnetic encoder in the modular drive unit can be no greater than 0.1°, the stall torque of the brushless DC servo motor can be no less than 2Nm, and the communication cycle between the microcontroller and the central controller can be no greater than 1ms, in order to meet the requirements of high-precision synchronization and fast response.
[0057] In some optional implementations, to further optimize attitude stability during deformation, the maximum velocity and acceleration of the S-shaped trajectory can be dynamically adjusted in each cycle of attitude control based on the current arm angular velocity (deformation rate), remaining deformation angle, and current attitude angular deviation. For example, a joint cost function J_n can be constructed, which includes attitude deviation penalty, time urgency penalty, and actual angular velocity penalty. In each control cycle, several preset combinations of velocity α_c and acceleration a_c are enumerated, and the combination that minimizes J_n is selected as the current optimal deformation parameters ω_opt and a_opt, which are then broadcast to the relevant drive units to update their local trajectories. This dynamic game-theoretic optimization can balance deformation rate and flight stability online based on real-time flight status (such as attitude stability) and environmental urgency, achieving the effect of "deforming as quickly as possible without instability."
[0058] In some optional implementations, in master-slave synchronization control, the synchronization deviation calculated by the central controller can be the maximum value of the real-time angle difference between any two arms, and the preset synchronization deviation threshold can be ±2°. When an arm is determined to be "lagging," the "acceleration fine-tuning signal" sent by the central controller to its drive unit can specifically be to increase the proportional gain of its local speed loop or add a positive speed feedforward; conversely, for "leading" arms, a deceleration signal is sent to reduce the gain or add a negative speed feedforward.
[0059] In some optional implementations, to protect the motor and prevent overload, a virtual variable L_trend, representing the "deformation load trend," can be built into the MCU of each drive unit. This variable can be calculated by weighting the rate of change of the motor's q-axis current (reflecting a sudden load increase), the deviation of the current from the steady-state reference value (reflecting the static load), and the normalized value of the motor winding temperature (reflecting the thermal effect). When L_trend exceeds a preset limit, the maximum acceleration limit of the arm's S-curve is automatically reduced, achieving preventative control of "active deceleration before overcurrent," avoiding triggering emergency braking, and ensuring the continuity of the deformation process.
[0060] This application also provides an intelligent collaborative control device for the arms of a multi-rotor unmanned aerial vehicle (UAV). The device may include a configuration decision module, a collaborative drive module, an attitude control module, a synchronization correction module, and an adaptive control module. The configuration decision module can determine the target arm configuration based on the UAV's flight mission information and / or perceived environmental information; wherein, determining the target arm configuration may include: making a matching decision based on the environmental information and / or the flight mission information according to a pre-established arm configuration library to determine the target arm configuration. The collaborative drive module can control each arm of the UAV to collaboratively operate towards its corresponding target angle position. The attitude control module can perform attitude control on the UAV based on its current attitude information during the collaborative arm operation. The synchronization correction module can perform master-slave synchronization control and correct synchronization deviations based on real-time angle feedback from each arm during the collaborative arm operation. The adaptive control module can identify the UAV's dynamic parameters changed due to arm configuration variations online during the collaborative arm operation and attitude control, and adaptively adjust the attitude control law based on these dynamic parameters.
[0061] Each module can be implemented in software, hardware, or a combination of both. For example, the configuration decision module and synchronization correction module can be implemented by the main control chip and corresponding programs in the central controller; the function of the cooperative drive module can be jointly implemented by the central controller and the modular drive units of each arm; the attitude control module and adaptive control module can be integrated into the core attitude control loop in the flight control software. All modules work together to achieve all or part of the steps of the method described above, achieving the same technical effect.
[0062] This application also provides an electronic device that may include a memory and a processor. The memory (e.g., ROM, RAM, flash memory, etc.) may store a computer program. When the processor (e.g., CPU, DSP, MCU, etc.) executes the program, it implements the intelligent collaborative control method for the arms of a multi-rotor UAV as described above. The electronic device may be, for example, a flight controller (flight controller) for the UAV, a ground control station, or a server carrying the control method.
[0063] This application also provides a computer-readable storage medium (e.g., USB flash drive, portable hard drive, ROM, RAM, magnetic disk, or optical disk). This storage medium can store a computer program. When executed by a processor, this program can implement an intelligent collaborative control method for the deformable folding arm structure of a multi-rotor UAV.
[0064] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
Claims
1. An intelligent collaborative control method for a deformable folding arm structure of a multi-rotor unmanned aerial vehicle (UAV), characterized in that, include: Determine the target arm configuration based on the UAV's flight mission information and / or perceived environmental information; Control each arm of the UAV to move in coordination toward its corresponding target angle position; During the coordinated operation of the robotic arms, the attitude of the UAV is controlled based on its current attitude information. During the coordinated operation of the robotic arms, master-slave synchronization control is performed and synchronization deviations are corrected based on the real-time angle feedback of each robotic arm. During the coordinated operation and attitude control of the robotic arms, the dynamic parameters of the UAV that change due to the change in the robotic arm configuration are identified online, and the attitude control law is adaptively adjusted based on the dynamic parameters. The determination of the target arm configuration includes: making a matching decision based on the environmental information and / or the flight mission information according to a pre-established arm configuration library, so as to determine the target arm configuration.
2. The intelligent collaborative control method for the deformable folding arm structure of a multi-rotor UAV according to claim 1, characterized in that, The control of the coordinated operation of each arm includes: the central controller generating an S-shaped acceleration and deceleration target angle trajectory and broadcasting it to the drive unit of each arm, and each drive unit performing local closed-loop control to track the target angle trajectory.
3. The intelligent collaborative control method for the deformable folding arm structure of a multi-rotor UAV according to claim 2, characterized in that, The local closed-loop control is a position-speed-current three-loop PID control; the execution of master-slave synchronization control and correction of synchronization deviation includes: the central controller calculates the angle deviation between each arm in real time, and when the angle deviation exceeds a preset threshold, it outputs an acceleration fine-tuning signal to the drive unit of the lagging arm and a deceleration fine-tuning signal to the drive unit of the leading arm.
4. The intelligent collaborative control method for the deformable folding arm structure of a multi-rotor UAV according to claim 1, characterized in that, The online identification uses the recursive least squares (RLS) algorithm; the adaptive adjustment of the attitude control law uses the parameter adaptive active disturbance rejection control (ADRC) algorithm.
5. The intelligent collaborative control method for the deformable folding arm structure of a multi-rotor UAV according to claim 4, characterized in that, The forgetting factor in the recursive least squares algorithm is dynamically adjusted based on the sliding variance of the current information residual and the deformation rate of the robotic arm.
6. The intelligent collaborative control method for the deformable folding arm structure of a multi-rotor UAV according to claim 5, characterized in that, The formula for calculating the forgetting factor λ(k) is: λ(k) = λ_min + (λ_max - λ_min)·exp(-β·σ²(k) - γ·v_m(k)), where σ²(k) is the moving variance of the new residual, v_m(k) is the deformation rate, and λ_max, λ_min, β, and γ are preset coefficients.
7. The intelligent collaborative control method for the deformable folding arm structure of a multi-rotor UAV according to claim 1, characterized in that, The arm configuration library includes at least one of the following: fully deployed configuration, semi-folded crossing configuration, fully folded storage configuration, and asymmetric folding wind-resistant configuration; the environmental information is obtained based on at least one of the following: the width of the passage ahead, the current wind speed, and the remaining battery power of the drone.
8. The intelligent collaborative control method for the deformable folding arm structure of a multi-rotor UAV according to claim 1, characterized in that, The matching decision based on the environmental information and / or the flight mission information includes: Based on the triggering indicators of each sensing channel, the integral quantity of deformation demand S(k) is calculated; The integral of the deformation requirement S(k) is updated recursively through an integrator with an asymmetric hysteresis function; When S(k) is greater than the activation threshold, a valid deformation request is output.
9. The intelligent collaborative control method for the deformable folding arm structure of a multi-rotor UAV according to claim 8, characterized in that, The asymmetric hysteresis function H(δ(k)) is defined as follows: when S(k-1) > S_on, H(δ(k)) = -h_off; when S(k-1) ≤ S_on, H(δ(k)) = h_on; where δ(k) is the trigger deviation, h_on > h_off > 0, and S_on is the activation threshold.
10. An intelligent collaborative control system for a deformable folding arm structure of a multi-rotor unmanned aerial vehicle (UAV), used to implement the intelligent collaborative control method for the deformable folding arm structure of a multi-rotor UAV as described in any one of claims 1 to 9, characterized in that, include: Configuration decision module, used for UAV flight mission information and / or perceived environmental information. Determine the target arm configuration; wherein, determining the target arm configuration includes: making a matching decision based on the environmental information and / or the flight mission information according to a pre-established arm configuration library, so as to determine the target arm configuration; The collaborative drive module is used to control each arm of the UAV to move collaboratively towards its corresponding target angle position; An attitude control module is used to control the attitude of the UAV based on the current attitude information of the UAV during the coordinated operation of the robotic arms. The synchronization correction module is used to perform master-slave synchronization control and correct synchronization deviations based on the real-time angle feedback of each arm during the coordinated operation of the arms. The adaptive control module is used to identify the UAV dynamic parameters that change due to the change in the arm configuration during the coordinated operation and attitude control of the arm, and to adaptively adjust the attitude control law based on the dynamic parameters.