Adaptive attitude stabilization control algorithm and device for multi-rotor unmanned aerial vehicle

Abstract

The application provides a multi-rotor unmanned aerial vehicle adaptive attitude stabilization control algorithm and device, relates to the unmanned aerial vehicle control technical field, and comprises a control algorithm and a main control processing unit; the control algorithm comprises a geometric phase driving control process, a fractional order chaotic game control process, a manifold resonance adaptive control process, a topological manifold adaptive control process and a multi-scale disturbance prediction control process; all the processes of the control algorithm are executed through the main control processing unit; the application aims at the problems that the attitude error of the traditional PID control is 0.02-0.05 DEG and it is difficult to adapt to the dynamic environment; the geometric phase driving control process is combined with a plasma aerodynamic fine tuning array and an inertial measurement unit; a high-precision 3*3 rotation matrix is generated by adopting Lie algebra transformation and Newton-Raphson iteration method; the rotation speed of the rotor is optimized; and the attitude error is controlled to be 0.006 DEG, which is better than the effect of the traditional PID control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle control, in particular to a multi-rotor unmanned aerial vehicle adaptive attitude stabilization control algorithm and device. BACKGROUND

[0002] As an important part of modern aviation technology, multi-rotor unmanned aerial vehicles are widely used in aerial photography, inspection, logistics, agriculture and other fields, and their attitude stabilization control directly affects the safety, accuracy and efficiency of task execution. According to industry research, the core requirements of users for unmanned aerial vehicle attitude control systems include high-precision attitude holding (error < 0.01°), complex environment adaptability (such as 5-10 m / s wind speed), cluster coordination consistency (error < 0.02°), fast response (< 2 ms), and low energy consumption operation (< 6 W). Modern unmanned aerial vehicle application scenarios (such as wind farm inspection, urban low-altitude logistics, post-disaster rescue) have higher requirements for attitude stability and cluster coordination capability, especially in dynamic load changes (such as 0.1 kg / s), strong airflow disturbance and vibration interference. However, existing attitude control technologies face the following key technical problems:

[0003] Insufficient attitude control precision and robustness: traditional PID control relies on fixed parameter adjustment and is difficult to adapt to dynamic environmental changes (such as wind speed above 5 m / s or sudden load changes), with attitude errors usually in the range of 0.02°-0.05°, which cannot meet the high-precision task requirements (such as wind turbine blade inspection requiring an error of < 0.01°). Although the inertial navigation system (INS) combines gyroscopes and accelerometers, it lacks effective compensation for high-frequency vibrations (1000-2000 Hz) or thermal flow anomalies (> 0.1 W / m 2 ).

[0004] Limited environmental adaptability: existing technologies are greatly affected by environmental factors such as wind speed, temperature, and humidity. For example, traditional PID control has a response time of 5-10 ms under strong wind (10 m / s), and the attitude error increases to 0.03°. A single sensor (such as a gyroscope with an accuracy of 0.01 rad / s) cannot capture multi-scale disturbances (such as airflow 0.1 Pa, vibration 0.05 Hz), and the false alarm rate is as high as 10% in extreme weather conditions (such as -10°C low temperature or 80% humidity). Existing systems lack adaptive correction mechanisms, limiting the reliability of cross-environment (altitude 0-4000 m) applications.

[0005] Intelligence and low automation: current attitude control systems mostly rely on single control algorithm (such as PID or LQR), lack of multi-modal data fusion and intelligent optimization. For example, the traditional method needs to manually adjust the control parameters, which takes 30-60 minutes per UAV. Cluster cooperative control relies on ground station centralized computing, and communication delay (>100ms) leads to consistency error increase to 0.05°. Existing systems do not integrate predictive control, making it difficult to handle complex disturbances (such as airflow, vibration, and heat flow coupling) in real time.

[0006] Lack of prediction ability: existing technologies mainly focus on current attitude correction, and cannot predict the disturbance trend in the next 0.1-10 seconds (such as airflow intensity change 0.2Pa). Lack of prediction ability leads to feedback delay, increases the risk of attitude drift, and may cause collision or formation disorder in cluster tasks. For example, in wind farm inspection, dynamic load changes may cause attitude deviation accumulation, threatening task safety.

[0007] Low resource utilization efficiency: existing UAV control systems have high power consumption (such as traditional PID controller power consumption accounts for 85% of rated power), and do not combine environmental data to dynamically optimize resource allocation. Sensor sampling frequency is fixed (such as gyroscope 20kHz), causing computational redundancy in low disturbance scenarios. The anti-interference ability of cluster communication module (such as Wi-Fi, power consumption 2W) is limited (<10dB), and the data transmission rate (<500kbps) cannot meet the real-time cooperative demand, limiting the economy of large-scale applications.

[0008] Therefore, a multi-rotor UAV adaptive attitude stabilization control algorithm and device are needed to solve the above problems. SUMMARY

[0009] Technical problems solved

[0010] In view of the deficiencies of the prior art, the present application provides a multi-rotor UAV adaptive attitude stabilization control algorithm and device, which solves the problems mentioned in the above background art.

[0011] Technical scheme

[0012] To achieve the above purpose, the present application realizes the following technical scheme: a multi-rotor UAV adaptive attitude stabilization control algorithm and device, including a control algorithm and a main control processing unit, the control algorithm contains a geometric phase driving control process, a fractional order chaotic game control process, a manifold resonance adaptive control process, a topological manifold adaptive control process, and a multi-scale disturbance prediction control process, all processes of the control algorithm are executed through the main control processing unit.

[0013] The geometric phase driving control process collects airflow disturbance data through a plasma aerodynamic fine-tuning array with a sensitivity of 0.1 pascal, collects angular acceleration data through an inertial measurement unit with an accuracy of 0.001 rad / s2, calculates control input based on geometric phase calculation on special orthogonal groups, adjusts rotor torque, and controls attitude error to within 0.006°;

[0014] The fractional order chaotic game control process collects torque data through a magnetic suspension multi-degree-of-freedom drive system with an accuracy of 0.005 Nm, collects angular velocity data through a high-precision gyroscope with an accuracy of 0.002 rad / s, optimizes multi-agent game strategy based on fractional order chaotic oscillator, distributes rotor control input, and supports swarm collaborative control;

[0015] The manifold resonance adaptive control process collects vibration frequency spectrum data through a dynamic phonon crystal regulation system with a sensitivity of 0.05 Hz, adjusts frequency spectrum based on manifold resonance mode, ranges from 10 Hz to 2000 Hz, amplifies beneficial dynamics, and suppresses harmful vibrations;

[0016] The topological manifold adaptive control process collects heat flow data through a photonic crystal heat flow regulation module with an accuracy of 0.01 W / m2, optimizes control input based on topological manifold homeomorphism mapping, and adapts to dynamic load changes;

[0017] The multi-scale disturbance prediction control process processes data generated by the aforementioned processes through a multi-modal sensor fusion algorithm, predicts disturbance trends based on high-order statistical modeling, and generates feedforward control instructions.

[0018] Preferably, the geometric phase driving control process uses Lie algebra transformation to map the airflow disturbance data collected by the plasma aerodynamic fine-tuning array and the angular acceleration data collected by the inertial measurement unit to a 3x3 three-dimensional rotation matrix, the number of matrix rows corresponds to the number of plasma generator array rows, calculates the geometric phase trajectory through the Newton-Raphson iteration method (iteration number 5 times, convergence error 0.001), generates pulse width modulation control signals (duty cycle resolution 0.1%, frequency 20 kHz), transmits to the rotor motor driver through the serial peripheral interface (rate 10 Mbps), adjusts the rotor speed (accuracy 0.01 rpm, range 500-5000 rpm), response time is 2ms, attitude error is controlled within 0.006°; when the attitude error is greater than 0.01°, the control signal of the geometric phase driving control process is executed preferentially, and when conflicts occur, the final control signal is generated through a weighted fusion algorithm (weights are attitude error 0.6, time urgency 0.4, based on least squares method);

[0019] The geometric phase driving control process is based on the airflow disturbance data (10 kHz x 30 seconds = 300,000 data points) and angular acceleration data (100 kHz x 30 seconds = 3,000,000 data points) in the past 30 seconds, adjusts the phase trajectory structure through a topology optimization algorithm (based on gradient descent method, step size 0.01, iteration number 100), generates a phase state report (data volume 512 bytes) containing attitude deviation (unit °), disturbance intensity (unit Pascal), timestamp (precision 1 ms) and control signal (duty ratio), and transmits it to the fractional order chaotic game control process through the internal integrated circuit bus (rate 400 kHz) for initial condition calibration of the game strategy.

[0020] Preferably, the fractional order chaotic game control process collects torque data through a magnetic levitation multi-degree-of-freedom driving system, with a sampling frequency of 20 kHz and an accuracy of 0.005 Newton-meters, collects angular velocity data through a high-precision gyroscope, with a sampling frequency of 20 kHz and an accuracy of 0.002 radians per second, and receives the phase state report; a chaotic oscillator is modeled based on fractional order Caputo derivative (order 0.8, based on Glenwald-Leontief discretization, step size 0.01 seconds), four time scale windows of 0.5 seconds, 1 second, 2 seconds and 5 seconds are set, multi-agent game strategy is optimized through a Nash equilibrium solver (based on linear programming method, constraint condition 10, target function is torque allocation error minimization), pulse width modulation control signal (duty ratio resolution 0.1%, frequency 20 kHz) is generated, and transmitted to the rotor motor driver (model DRV8301) through the serial peripheral interface (rate 10 megabits per second) to adjust the rotor speed (accuracy 0.01 revolutions per minute, range 500 to 5000 revolutions per minute); when the attitude error is less than 0.01° but the swarm consistency error is greater than 0.02°, the control signal of the fractional order chaotic game control process is preferentially executed, and when there is a conflict, a final control signal is generated through a weighted fusion algorithm (weights are consistency error 0.5, time urgency 0.5, based on least squares method);

[0021] The fractional order chaotic game control process is based on the torque data (20 kHz x 60 seconds = 1,200,000 data points) and angular velocity data (20 kHz x 60 seconds = 1,200,000 data points) in the past 60 seconds, calculates the game weight (initial weight: torque distribution 0.4, angular velocity feedback 0.3, energy optimization 0.2, and cluster cooperation 0.1) through a fractional order integrator (integration step 0.01 seconds), generates a game report (data size 1024 bytes) containing rotor distribution ratio (unit percentage), fault tolerance state (Boolean value), cooperation consistency (unit °), and timestamp (precision 1 ms), and transmits it to other drones through the ultra-wideband communication module to synchronize the cluster control strategy. When the communication is interrupted, the local attitude control is maintained based on the game report in the last 5 seconds (stored in the 128 kilobyte buffer of the main control processing unit) to maintain the consistency error less than 0.02° by adjusting the rotor speed; the game report is transmitted to the manifold resonance adaptive control process through the internal integrated circuit bus (rate 400 kHz) for vibration suppression parameter adjustment.

[0022] Preferably, the manifold resonance adaptive control process collects vibration frequency spectrum data through a dynamic phononic crystal regulation system, with a sampling frequency of 5 kHz and a sensitivity of 0.05 Hz, and receives the game report generated by the fractional order chaotic game control process; based on fast wavelet transform (wavelet basis Daubechies-4, decomposition level 4), the frequency spectrum (range 10 Hz to 2000 Hz) is decomposed, and the geometric shape of the periodic microstructure (adjustment range 0.4 to 0.6 mm) is adjusted through the magnetostrictive actuator (drive voltage 10 volts, deformation 0.1 mm) to change the bandgap frequency (adjustment range 50 Hz), amplify beneficial dynamics (frequency 100 to 500 Hz, gain 1.5 times), and suppress harmful vibrations (frequency 1000 to 2000 Hz, suppression rate 60%); based on the vibration data in the past 10 seconds (5 kHz x 10 seconds = 50,000 data points), the future 0.1 seconds of resonance changes are predicted through the Gaussian process regression algorithm (kernel function square exponential kernel, length scale 0.1 second, noise variance 0.01), and a vibration control report (data size 512 bytes) containing resonance frequency (unit Hz), vibration suppression rate (unit percentage), attitude stability (unit °), and timestamp (precision 1 ms) is generated and transmitted to the topological manifold adaptive control process through the internal integrated circuit bus (rate 400 kHz) for topological mapping optimization; when the vibration suppression rate is less than 50%, the control signal of the manifold resonance adaptive control process is preferentially executed, and when there is a conflict, the final control signal is generated through a weighted fusion algorithm (weight vibration suppression rate 0.7, time urgency 0.3, based on least squares method).

[0023] Preferably, the topological manifold adaptive control process collects heat flow data through the photonic crystal heat flow regulation module, with a sampling frequency of 2 kHz and an accuracy of 0.01 watts per square meter, and receives the vibration control report generated by the manifold resonance adaptive control process; based on the topological manifold homeomorphism mapping algorithm (based on the Hough transform, mapping dimension 3), the control input is calculated through the nonlinear topological optimization algorithm (based on the simulated annealing method, cooling coefficient 0.95, iteration number 50) to generate a pulse width modulation control signal (duty cycle resolution 0.1%, frequency 20 kHz), which is transmitted to the rotor motor driver (model DRV8301) through the serial peripheral interface (rate 10 megabits per second) with a response time of 2 ms;

[0024] The topological manifold adaptive control process generates a topological state report (data size 512 bytes) containing attitude deviation (unit °), heat flow distribution (unit watts per square meter), and timestamp (accuracy 1 ms) based on the past 20 seconds of heat flow data (2 kHz x 20 seconds = 40,000 data points), which is transmitted to the multiscale disturbance prediction control process through the internal integrated circuit bus (rate 400 kHz) for disturbance trend prediction; when the dynamic load change rate is greater than 0.1 kg per second, the control signal of the topological manifold adaptive control process is preferentially executed, and the final control signal is generated through a weighted fusion algorithm in case of conflict.

[0025] Preferably, the multiscale disturbance prediction control process receives the phase state report generated by the geometric phase driving control process, the game report generated by the fractional order chaotic game control process, the vibration control report generated by the manifold resonance adaptive control process, and the topological state report generated by the topological manifold adaptive control process through the multi-modal sensor fusion algorithm to generate comprehensive disturbance data; based on high-order statistical modeling (fourth-order cumulant analysis, window length 0.1 seconds to 10 seconds), the disturbance is decomposed into four time scales of 0.1 seconds, 1 second, 5 seconds, and 10 seconds; based on the past 30 seconds of fusion data, the disturbance trend from 0.1 seconds to 10 seconds in the future is predicted through the time series decomposition algorithm (based on empirical mode decomposition, mode number 8, residual threshold 0.01) to generate a prediction report (data size 1024 bytes) containing disturbance intensity (unit pascal), prediction time window (unit second), and feedforward control instruction (duty cycle), which is transmitted to the rotor motor driver (model DRV8301) through the serial peripheral interface (rate 10 megabits per second) to adjust the rotor speed (accuracy 0.01 revolutions per minute, range 500 to 5000 revolutions per minute); the prediction report is transmitted to the geometric phase driving control process through the internal integrated circuit bus (rate 400 kHz) for phase trajectory optimization; when the disturbance intensity is greater than 0.2 pascal, the control signal of the multiscale disturbance prediction control process is preferentially executed, and the final control signal is generated through a weighted fusion algorithm (weights are disturbance intensity 0.6 and time urgency 0.4 based on the least squares method) in case of conflict.

[0026] Preferably, the fractional order chaotic game control process supports swarm cooperative control, transmits attitude data (unit °), torque data (unit Newton·meter) and angular velocity data (unit radian per second) through the ultra-wideband communication module, adopts BCH(511,493) error correction coding, combines 2.4GHz frequency hopping spread spectrum (hopping rate 1000 hops per second, bandwidth 20 megahertz), and controls the data transmission rate within 1 megabit per second, with a sampling frequency of 1kHz; a master-slave clock synchronization mechanism is adopted, the master node broadcasts IEEE 1588 synchronization frames (frame length 64 bytes, broadcast period 1ms) through the ultra-wideband signal, and the slave node calibrates the local clock through the precision time protocol (PTPv2), with a synchronization accuracy of 500 nanoseconds, ensuring that the swarm attitude consistency error is less than 0.01°; based on the attitude data (1kHz×5 seconds=5,000 data points), torque data (1kHz×5 seconds=5,000 data points) and angular velocity data (1kHz×5 seconds=5,000 data points) of the past 5 seconds, the game strategy (weight update period 0.1 seconds) of the fractional order chaotic game control process is synchronized; when the communication is interrupted, each unmanned aerial vehicle maintains the local attitude control based on the game report of the last 5 seconds (stored in a 128 kilobyte buffer) to adjust the rotor speed to keep the consistency error less than 0.02°;

[0027] The ultra-wideband communication module generates a swarm state report (data size 256 bytes) containing cooperative consistency (unit °), communication delay (unit ms) and timestamp (accuracy 1ms), which is transmitted to the master control processing unit through the internal integrated circuit bus (rate 400kHz) for the topological manifold adaptive control process and the multi-scale disturbance prediction control process.

[0028] The adaptive attitude stability control device of the multi-rotor unmanned aerial vehicle comprises a rotor assembly and a fuselage structure, both sides of the top of the fuselage structure are provided with rotor bases, the top of each rotor base is provided with a rotor, and there are four rotor bases in total, and the rotors at the top of the four rotor bases constitute a complete rotor assembly.

[0029] Each rotor in the rotor assembly is made of carbon fiber composite material, with a diameter of 10-15 cm, mounted on a rotor base fixed to the fuselage structure by a mechanical arm; each rotor surface is coated with a nanoscale dielectric coating for fixing the plasma generator of the plasma aerodynamic fine-tuning array; the plasma aerodynamic fine-tuning array contains four plasma generators (made of tungsten alloy high-voltage electrode and copper alloy ground electrode, size 20 mm x 5 mm x 5 mm, electrode spacing 0.2 mm), distributed in a rectangular array (2x2), 10-15 cm from the center of the rotor, the number of array rows corresponds to the dimension of the three-dimensional rotation matrix; the plasma generator applies a 5-10 kHz multi-band pulsed electric field (voltage 5 kV, current 0.05 mA, pulse width 10 μs) through a pulse power supply to excite the air on the rotor surface to form plasma, change the local airflow speed (change 0.1 m / s), generate airflow disturbance data (sensitivity 0.1 Pa, sampling frequency 10 kHz, resolution 0.015 Pa), transmitted to the main control processing unit through the serial peripheral interface (SPI, rate 10 Mbps);

[0030] The rotor base is embedded with a magnetic suspension multi-degree-of-freedom drive system, which includes a magnetic suspension bearing and a magnetic force drive module; the magnetic suspension bearing (material stainless steel, inner diameter 40 mm, outer diameter 60 mm, thickness 10 mm) is installed at the center of the rotor base, supporting the rotation of the rotor and reducing mechanical friction (friction coefficient less than 0.001); the magnetic force drive module includes six high-temperature superconducting coils (material yttrium barium copper oxide, diameter 5 cm, number of turns 100), evenly distributed in the rotor base (30 cm from the center of the fuselage structure), the coils are energized (current 10 A, frequency 50 Hz) to generate a magnetic field (strength 0.5 Tesla), the change of the magnetic field is measured by a Hall sensor (model SS495A, accuracy 0.2 mT, sampling frequency 20 kHz), and the three-axis torque data (accuracy 0.005 Nm, range ±0.5 Nm) are calculated based on the current-torque conversion formula, transmitted to the main control processing unit through the controller area network bus (CAN, rate 1 Mbps);

[0031] The fuselage structure is an aluminum alloy frame (size, length 50 cm, width 30 cm, height 20 cm, thickness 5 mm), the center inside is fixed with a main control processing unit, the top and bottom are embedded with a dynamic phononic crystal regulation system and a photonic crystal heat flow regulation module; the dynamic phononic crystal regulation system contains six piezoelectric devices (material zirconium titanate lead ceramic, size 10 mm x 10 mm x 1 mm) and four magnetostrictive actuators (material nickel-iron alloy, size 10 mm x 5 mm x 5 mm), which are installed in a periodic microstructure (material polyether ether ketone, periodic length 0.5 mm, 3D printing, size 50 mm x 50 mm x 2 mm), and the periodic microstructure is arranged at the top and bottom of the fuselage structure (5 cm away from the main control processing unit); the piezoelectric device detects the microstructure deformation to generate vibration spectrum data (range 0 to 5 volts, resolution 0.01 volts, sampling frequency 5 kHz, sensitivity 0.05 hertz), which is transmitted to the main control processing unit through an internal integrated circuit bus (I2C, rate 400 kHz); the magnetostrictive actuator adjusts the microstructure geometry by driving voltage to achieve vibration suppression; the photonic crystal heat flow regulation module contains a photonic crystal structure (material photosensitive resin, periodic length 0.1 mm, 3D printing, size 50 mm x 50 mm x 1 mm) and an infrared spectrum analyzer (model FLIR A65, resolution 0.01 microns, size 30 mm x 30 mm x 20 mm), the photonic crystal structure is arranged on the surface of the fuselage structure (5 cm away from the main control processing unit), and the infrared spectrum analyzer measures the thermal radiation intensity to generate heat flow data (range 0 to 100 watts per square meter, accuracy 0.01 watts per square meter, sampling frequency 2 kHz), which is transmitted to the main control processing unit through the I2C bus;

[0032] The surface of the fuselage structure is mounted with an inertial measurement unit (model MPU-9250, size 20 mm x 20 mm x 5 mm, 3 cm away from the main control processing unit), which generates angular acceleration data (sampling frequency 100 kHz, accuracy 0.001 rad / s2), which is transmitted to the main control processing unit through the SPI bus;

[0033] The surface of the fuselage structure is mounted with a high-precision gyroscope (model ADIS16470, size 20 mm x 20 mm x 5 mm, 3 cm away from the main control processing unit), which generates angular velocity data (sampling frequency 20 kHz, accuracy 0.002 rad / s), which is transmitted to the main control processing unit through the SPI bus;

[0034] The master processing unit is an aluminum alloy shell package (size 100mm*100mm*50mm) fixed in the center of the fuselage structure, connected with the sensors in the rotor assembly and the fuselage structure through SPI bus (speed 10 Mbps), I2C bus (speed 400kHz) and CAN bus (speed 1 Mbps); the master processing unit integrates an ultra-wideband communication module for transmitting attitude data, torque data and status reports, adopts BCH (511, 493) error correction coding and 2.4GHz frequency hopping spread spectrum; the master processing unit executes the above-mentioned control algorithm through an integrated circuit chip (model Xilinx Zynq-7000, clock frequency 2GHz) and a digital signal processing chip (model TIC6748, processing capacity 1000 million floating point operations per second), processes airflow disturbance data (10kHz), vibration spectrum data (5kHz), torque data (20kHz), heat flow data (2kHz), angular acceleration data (100kHz) and angular velocity data (20kHz); a 32-bit high-precision timer (frequency 100MHz, precision 10ns) is used to synchronize all sensor data, and the synchronization error is compensated by Kalman filtering (state vector dimension 6, process noise covariance 0.001), and the error is less than 10 microseconds; a pulse width modulation control signal (duty cycle resolution 0.1%, frequency 20kHz) and a status report are generated, the calculation delay is 3ms, the power consumption is 5W, and the pulse width modulation control signal and the status report are transmitted to the rotor motor driver (model DRV8301) through the CAN bus and to other unmanned aerial vehicles and ground stations through the ultra-wideband communication module.

[0035] Advantages

[0036] The present application provides a multi-rotor unmanned aerial vehicle adaptive attitude stabilization control algorithm and device.

[0037] Advantages

[0038] 1. The present application aims at the problem that the traditional PID control attitude error is 0.02°-0.05° and is difficult to adapt to dynamic environment, the present application generates a high-precision 3*3 rotation matrix by using Lie algebra transformation and Newton-Raphson iteration method through a geometric phase driving control process, combines a plasma aerodynamic fine-tuning array and an inertial measurement unit, optimizes the rotor speed, and controls the attitude error to 0.006°, which is better than the traditional PID control effect. The topological manifold adaptive control process effectively compensates high-frequency vibration and heat flow anomaly through a photonic crystal heat flow regulation module, ensures that the attitude drift is reduced under dynamic load change, significantly improves the robustness, and meets the high-precision task demand of wind turbine blade inspection.

[0039] 2、The present application is aimed at the problem of existing technology that the response time is prolonged (5-10 ms) and the false positive rate is as high as 10% under strong wind, low temperature or high humidity, the present application optimizes environmental adaptability through multi-modal sensor fusion and adaptive algorithm. The dynamic phonon crystal regulation system adjusts the vibration spectrum in real time through fast wavelet transform, suppresses harmful vibration, combines with the photonic crystal heat flow module to correct heat flow anomaly, and reduces the environmental interference error to ±0.005°.

[0040] 3、The present application is aimed at the problem of traditional method that relies on a single algorithm, needs manual adjustment of parameters and cluster communication delay, the present application integrates multi-modal sensor fusion algorithm and intelligent control process. The geometric phase driving control, fractional order chaotic game control, manifold resonance adaptive control, topological manifold adaptive control and multi-scale disturbance prediction control process are automatically executed through the main control processing unit, process airflow disturbance, vibration, torque and other multi-source data, and the calculation delay is only 3ms. The fractional order chaotic game control realizes cluster cooperation through the ultra-wideband communication module (1Mbps, BCH (511, 493) error correction, 2.4GHz frequency hopping).

[0041] 4、The present application is aimed at the problem that existing technology cannot predict 0.1-10 second disturbance trend (such as air flow 0.2Pa), resulting in feedback delay, the present application through multi-scale disturbance prediction control process, fusion phase state, game, vibration and topological state report, uses high-order statistical modeling (fourth-order cumulant, window 0.1-10s) and empirical mode decomposition (mode number 8, residual 0.01), predicts future 0.1-10 second disturbance trend, and improves accuracy. The prediction report (1024 bytes) generates feedforward control instructions, adjusts the rotor speed through the SPI interface (10Mbps), the response time is <2ms, the attitude error is <0.006°, effectively reduces the risk of collision or out-of-sequence in cluster tasks, and is especially suitable for dynamic scenes such as wind farm inspection.

[0042] 5、The present application is aimed at the problem of traditional control system that the power consumption is high, and the sensor sampling frequency is fixed, resulting in calculation redundancy, the present application optimizes efficiency through dynamic resource allocation and low power consumption design. The power consumption of the main control processing unit is only 5W, which is lower than that of the traditional PID controller. The sensor sampling frequency is dynamically adjusted according to the disturbance intensity, reducing 30% of the calculation redundancy. In the low disturbance scene, the sampling and calculation load is automatically reduced, the overall energy consumption is controlled below 60% of the rated power, significantly improving the economy and the feasibility of large-scale promotion. BRIEF DESCRIPTION OF DRAWINGS

[0043] Fig. 1 is the control flowchart of the present application;

[0044] Fig. 2 is the control framework diagram of the present application;

[0045] Fig. 3The unmanned aerial vehicle structure diagram of the present application.

[0046] Legend:

[0047] 1, rotor assembly; 2, fuselage structure; 3, rotor base. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application. Specific embodiment one:

[0050] As shown in the figure, Figs. 1-3 The adaptive attitude stabilization control algorithm and device of the multi-rotor unmanned aerial vehicle include a control algorithm and a main control processing unit. The control algorithm includes a geometric phase driving control process, a fractional order chaotic game control process, a manifold resonance adaptive control process, a topological manifold adaptive control process, and a multi-scale disturbance prediction control process. All processes of the control algorithm are executed through the main control processing unit.

[0051] The geometric phase driving control process collects airflow disturbance data through a plasma aerodynamic fine-tuning array, with a sensitivity of 0.1 pascal. It collects angular acceleration data through an inertial measurement unit, with an accuracy of 0.001 rad / s2. It calculates the control input based on the geometric phase on the special orthogonal group, adjusts the rotor moment, and controls the attitude error to within 0.006°.

[0052] The fractional order chaotic game control process collects moment data through a magnetic levitation multi-degree-of-freedom drive system, with an accuracy of 0.005 N·m. It collects angular velocity data through a high-precision gyroscope, with an accuracy of 0.002 rad / s. It optimizes the multi-agent game strategy based on a fractional order chaotic oscillator, distributes the rotor control input, and supports swarm collaborative control.

[0053] The manifold resonance adaptive control process collects vibration frequency spectrum data through a dynamic phononic crystal regulation system, with a sensitivity of 0.05 Hz. It adjusts the frequency spectrum based on the manifold resonance mode, with a range of 10 Hz to 2000 Hz, amplifies beneficial dynamics, and suppresses harmful vibrations.

[0054] The topological manifold adaptive control process collects heat flow data through a photonic crystal heat flow regulation module, with an accuracy of 0.01 W / m2. It optimizes the control input based on the topological manifold homeomorphism mapping, and adapts to dynamic load changes.

[0055] The multi-scale disturbance prediction control process processes the data generated by the foregoing process through a multi-modal sensor fusion algorithm, predicts the disturbance trend based on high-order statistical modeling, and generates feedforward control instructions.

[0056] The geometric phase driving control process uses Lie algebra transformation to map the airflow disturbance data collected by the plasma aerodynamic fine-tuning array and the angular acceleration data collected by the inertial measurement unit to a 3x3 three-dimensional rotation matrix. The number of matrix rows corresponds to the number of plasma generator array rows. The geometric phase trajectory is calculated by the Newton-Raphson iteration method (iteration number 5 times, convergence error 0.001). The pulse width modulation control signal (duty cycle resolution 0.1%, frequency 20 kHz) is generated, transmitted to the rotor motor driver through the serial peripheral interface (rate 10 megabits per second), and the rotor speed is adjusted (accuracy 0.01 revolutions per minute, range 500 to 5000 revolutions per minute). The response time is 2ms, and the attitude error is controlled within 0.006°. When the attitude error is greater than 0.01°, the control signal of the geometric phase driving control process is executed preferentially, and when there is a conflict, the final control signal is generated through a weighted fusion algorithm (weighting attitude error 0.6, time urgency 0.4, based on least squares method);

[0057] The geometric phase driving control process is based on airflow disturbance data (10kHzx30s=300,000 data points) and angular acceleration data (100kHzx30s=3,000,000 data points) in the past 30 seconds. The phase trajectory structure is adjusted through a topology optimization algorithm (based on gradient descent method, step size 0.01, iteration number 100) to generate a phase state report (data size 512 bytes) containing attitude deviation (unit °), disturbance intensity (unit Pascal), timestamp (accuracy 1ms) and control signal (duty cycle). It is transmitted to the fractional-order chaotic game control process through the internal integrated circuit bus (rate 400kHz) to calibrate the initial conditions of the game strategy.

[0058] The fractional chaotic game control process collects torque data through a magnetic levitation multi-degree-of-freedom drive system, with a sampling frequency of 20 kHz and an accuracy of 0.005 Newton-meters. Angular velocity data is collected through a high-precision gyroscope, with a sampling frequency of 20 kHz and an accuracy of 0.002 radians per second, and phase state reports are received. A chaotic oscillator is modeled based on fractional Caputo derivatives (order 0.8, based on the Glenwald-Leontief discretization with a step size of 0.01 seconds), with four time scale windows set to 0.5 seconds, 1 second, 2 seconds, and 5 seconds. A multi-agent game strategy is optimized through a Nash equilibrium solver (based on the linear programming method, with 10 constraint conditions and a target function of minimizing torque allocation error), generating a pulse width modulation control signal (duty cycle resolution 0.1%, frequency 20 kHz) that is transmitted to a rotor motor driver (model DRV8301) through a serial peripheral interface (rate 10 megabits per second) to adjust rotor speed (accuracy 0.01 revolutions per minute, range 500 to 5000 revolutions per minute). When the attitude error is less than 0.01° but the swarm consistency error is greater than 0.02°, the control signal of the fractional chaotic game control process is executed first, and when conflicts occur, a final control signal is generated through a weighted fusion algorithm (weights: consistency error 0.5, time urgency 0.5, based on the least squares method).

[0059] The fractional chaotic game control process is based on torque data (20 kHz x 60 seconds = 1,200,000 data points) and angular velocity data (20 kHz x 60 seconds = 1,200,000 data points) over the past 60 seconds, and calculates game weights (initial weights: torque allocation 0.4, angular velocity feedback 0.3, energy optimization 0.2, swarm collaboration 0.1) through a fractional integrator (integration step size 0.01 seconds) to generate a game report (data size 1024 bytes) containing rotor allocation proportions (unit percentage), fault tolerance states (Boolean value), collaboration consistency (unit °), and timestamps (accuracy 1 ms), which is transmitted to other drones through an ultra-wideband communication module to synchronize swarm control strategies. When communication is interrupted, local attitude control is maintained based on the latest 5-second game report (stored in a 128 kilobyte buffer in the main control processing unit) to adjust rotor speed to maintain a consistency error of less than 0.02°. The game report is transmitted to the manifold resonance adaptive control process through an internal integrated circuit bus (rate 400 kHz) for vibration suppression parameter adjustment.

[0060] The manifold resonance adaptive control process collects vibration frequency spectrum data through the dynamic phononic crystal regulation system, with a sampling frequency of 5 kHz and a sensitivity of 0.05 Hz, and receives the game report generated by the fractional order chaos game control process; based on the fast wavelet transform (wavelet basis is Daubechies-4, decomposition level is 4), the frequency spectrum (range 10 Hz to 2000 Hz) is decomposed, the geometric shape of the periodic microstructure is adjusted (adjustment range 0.4 to 0.6 mm) through the magnetostrictive actuator (driving voltage 10 volts, deformation 0.1 mm), the band gap frequency is changed (adjustment range 50 Hz), the beneficial dynamics are amplified (frequency 100 to 500 Hz, gain 1.5 times), and the harmful vibration is suppressed (frequency 1000 to 2000 Hz, suppression rate 60%); based on the vibration data of the past 10 seconds (5 kHz x 10 seconds = 50,000 data points), the future 0.1 second resonance change is predicted through the Gaussian process regression algorithm (kernel function is square exponential kernel, length scale 0.1 second, noise variance 0.01), and the vibration control report (data size 512 bytes) containing the resonance frequency (unit Hz), vibration suppression rate (unit percentage), attitude stability (unit °) and time stamp (accuracy 1 ms) is generated, which is transmitted to the topological manifold adaptive control process through the internal integrated circuit bus (rate 400 kHz) for topological mapping optimization; when the vibration suppression rate is less than 50%, the control signal of the manifold resonance adaptive control process is preferentially executed, and when there is a conflict, the final control signal is generated through a weighted fusion algorithm (weights are vibration suppression rate 0.7, time urgency 0.3, based on least squares method).

[0061] The topological manifold adaptive control process collects heat flow data through the photonic crystal heat flow regulation module, with a sampling frequency of 2 kHz and an accuracy of 0.01 watts per square meter, and receives the vibration control report generated by the manifold resonance adaptive control process; based on the topological manifold homeomorphism mapping algorithm (based on the Hough transform, mapping dimension 3), the control input is calculated through the nonlinear topological optimization algorithm (based on the simulated annealing method, cooling coefficient 0.95, iteration number 50), and the pulse width modulation control signal (duty cycle resolution 0.1%, frequency 20 kHz) is generated, which is transmitted to the rotor motor driver (model DRV8301) through the serial peripheral interface (rate 10 megabits per second) with a response time of 2 ms;

[0062] The topological manifold adaptive control process is based on the heat flow data of the past 20 seconds (2 kHz x 20 seconds = 40,000 data points) to generate a topological state report (data size 512 bytes) containing attitude deviation (unit °), heat flow distribution (unit watt per square meter), and timestamp (precision 1 ms), which is transmitted to the multi-scale disturbance prediction control process through the internal integrated circuit bus (rate 400 kHz) for disturbance trend prediction; when the dynamic load change rate is greater than 0.1 kg per second, the control signal of the topological manifold adaptive control process is preferentially executed, and when there is a conflict, the final control signal is generated through a weighted fusion algorithm.

[0063] The multi-scale disturbance prediction control process receives the phase state report generated by the geometric phase driving control process, the game report generated by the fractional order chaotic game control process, the vibration control report generated by the manifold resonance adaptive control process, and the topological state report generated by the topological manifold adaptive control process through the multi-modal sensor fusion algorithm to generate comprehensive disturbance data; based on high-order statistical modeling (fourth-order cumulant analysis, window length 0.1 seconds to 10 seconds), the 0.1 second, 1 second, 5 second, and 10 second time scale disturbances are decomposed; based on the fusion data of the past 30 seconds, the disturbance trend of the future 0.1 seconds to 10 seconds is predicted through a time series decomposition algorithm (based on empirical mode decomposition, mode number 8, residual threshold 0.01) to generate a prediction report (data size 1024 bytes) containing disturbance intensity (unit Pascal), prediction time window (unit second), and feedforward control instruction (duty cycle), which is transmitted to the rotor motor driver (model DRV8301) through the serial peripheral interface (rate 10 megabits per second) to adjust the rotor speed (precision 0.01 revolutions per minute, range 500 to 5000 revolutions per minute); the prediction report is transmitted to the geometric phase driving control process through the internal integrated circuit bus (rate 400 kHz) for phase trajectory optimization; when the disturbance intensity is greater than 0.2 Pascal, the control signal of the multi-scale disturbance prediction control process is preferentially executed, and when there is a conflict, the final control signal is generated through a weighted fusion algorithm (weights are disturbance intensity 0.6, time urgency 0.4, based on least squares method).

[0064] The fractional-order chaotic game control process supports cluster cooperative control. The attitude data (unit °), torque data (unit Newton-meter), and angular velocity data (unit radian per second) are transmitted through the ultra-wideband communication module, BCH (511, 493) error correction coding is used, 2.4 GHz frequency hopping spread spectrum (hopping rate 1000 hops per second, bandwidth 20 megahertz) is combined, the data transmission rate is controlled within 1 megabit per second, and the sampling frequency is 1 kHz; a master-slave clock synchronization mechanism is used, the master node broadcasts IEEE 1588 synchronization frames (frame length 64 bytes, broadcast period 1 ms) through the ultra-wideband signal, the slave node calibrates the local clock through the precision time protocol (PTPv2), the synchronization accuracy reaches 500 nanoseconds, and the cluster attitude consistency error is less than 0.01°; based on the attitude data (1 kHz x 5 seconds = 5,000 data points), torque data (1 kHz x 5 seconds = 5,000 data points), and angular velocity data (1 kHz x 5 seconds = 5,000 data points) in the past 5 seconds, the game strategy (weight update period 0.1 seconds) of the fractional-order chaotic game control process is synchronized; when the communication is interrupted, each unmanned aerial vehicle maintains the local attitude control based on the game report in the last 5 seconds (stored in a 128 kilobyte buffer) to adjust the rotor speed to keep the consistency error less than 0.02°.

[0065] The ultra-wideband communication module generates a cluster state report (data size 256 bytes) containing cooperative consistency (unit °), communication delay (unit ms), and timestamp (accuracy 1 ms), which is transmitted to the master control processing unit through the internal integrated circuit bus (rate 400 kHz) for use by the topological manifold adaptive control process and the multi-scale disturbance prediction control process.

[0066] The geometric phase driving control process generates a 3x3 rotation matrix based on Lie algebra transformation through airflow disturbance data collected by the plasma aerodynamic fine-tuning array and angular acceleration data collected by the inertial measurement unit, calibrates the phase trajectory through the Newton-Raphson iteration method, generates a pulse width modulation control signal, the response time is 2 ms, and the fine-tuning accuracy is 0.005°; the phase state report is transmitted to the fractional-order chaotic game control process through the internal integrated circuit bus.

[0067] The fractional-order chaotic game control process models the chaotic oscillator based on the fractional-order Caputo derivative through torque data collected by the magnetic suspension multi-degree-of-freedom driving system and angular velocity data (20 kHz, 0.002 radian per second) collected by the high-precision gyroscope, optimizes the game strategy through the Nash equilibrium solver, generates a game report (1024 bytes), and transmits it to other unmanned aerial vehicles through the ultra-wideband communication module to synchronize the cluster control strategy.

[0068] The application discloses a multi-rotor unmanned aerial vehicle adaptive attitude stabilization control device.

[0069] Each rotor in the rotor assembly 1 is made of carbon fiber composite material, has a diameter of 10-15 cm, is installed on the rotor base 3, and the rotor base 3 is fixed to the fuselage structure 2 through a mechanical arm; the surface of each rotor is coated with a nanoscale dielectric coating (0.1 microns thick, polyimide material), and a plasma generator for fixing a plasma aerodynamic fine-tuning array; the plasma aerodynamic fine-tuning array comprises four plasma generators (made of tungsten alloy high-voltage electrodes and copper alloy grounding electrodes, 20 mm*5 mm*5 mm in size, and 0.2 mm in electrode spacing), is distributed in a rectangular array (2*2), is located 10-15 cm away from the center of the rotor, and the number of array rows corresponds to the dimension of a three-dimensional rotation matrix; the plasma generator applies a 5-10 kHz multi-band pulse electric field (5 kV in voltage, 0.05 mA in current and 10 microseconds in pulse width) through a pulse power supply, excites air on the surface of the rotor to form plasma, changes the local airflow speed (by 0.1 m / s), generates airflow disturbance data (0.1 Pascal in sensitivity, 10 kHz in sampling frequency and 0.015 Pascal in resolution), and transmits the airflow disturbance data to the main control processing unit through a serial peripheral interface (SPI, 10 Mbps in rate).

[0070] The rotor base 3 is internally provided with a magnetic suspension multi-degree-of-freedom driving system, and the magnetic suspension multi-degree-of-freedom driving system comprises a magnetic suspension bearing and a magnetic force driving module; the magnetic suspension bearing (made of stainless steel, 40 mm in inner diameter, 60 mm in outer diameter and 10 mm in thickness) is installed at the center of the rotor base 3, supports the rotation of the rotor, and reduces mechanical friction (with a friction coefficient less than 0.001); the magnetic force driving module comprises six high-temperature superconducting coils (made of yttrium barium copper oxide, 5 cm in diameter and 100 in number of turns), is uniformly distributed in the rotor base 3 (30 cm away from the center of the fuselage structure 2), and generates a magnetic field (0.5 Tesla in intensity) through coil energization (10 A in current and 50 Hz in frequency); the change of the magnetic field is measured through a Hall sensor (model SS495A, 0.2 mT in accuracy and 20 kHz in sampling frequency), three-axis torque data (0.005 Nm in accuracy and in the range of ±0.5 Nm) are calculated based on a current-torque conversion formula, and the three-axis torque data are transmitted to the main control processing unit through a controller area network bus (CAN, 1 Mbps in rate).

[0071] The fuselage structure 2 is an aluminum alloy frame (size, length 50 cm, width 30 cm, height 20 cm, thickness 5 mm), with a central internal fixed main control processing unit, a top and bottom embedded dynamic phononic crystal regulation system and photonic crystal heat flow regulation module; the dynamic phononic crystal regulation system includes six piezoelectric devices (material zirconium titanate lead ceramic, size 10 mm x 10 mm x 1 mm) and four magnetostrictive actuators (material nickel-iron alloy, size 10 mm x 5 mm x 5 mm), installed in a periodic microstructure (material polyether ether ketone, periodic length 0.5 mm, 3D printing, size 50 mm x 50 mm x 2 mm), which is arranged on the top and bottom (5 cm from the main control processing unit) of the fuselage structure (2); the piezoelectric device detects the microstructure deformation and generates vibration spectrum data (range 0 to 5 volts, resolution 0.01 volts, sampling frequency 5 kHz, sensitivity 0.05 hertz), which is transmitted to the main control processing unit through the internal integrated circuit bus (I2C, rate 400 kHz); the magnetostrictive actuator adjusts the microstructure geometry by driving voltage to achieve vibration suppression; the photonic crystal heat flow regulation module includes a photonic crystal structure (material photosensitive resin, periodic length 0.1 mm, 3D printing, size 50 mm x 50 mm x 1 mm) and an infrared spectrum analyzer (model FLIR A65, resolution 0.01 microns, size 30 mm x 30 mm x 20 mm), the photonic crystal structure is arranged on the surface of the fuselage structure 2 (5 cm from the main control processing unit), and the infrared spectrum analyzer measures the thermal radiation intensity to generate heat flow data (range 0 to 100 watts per square meter, accuracy 0.01 watts per square meter, sampling frequency 2 kHz), which is transmitted to the main control processing unit through the I2C bus;

[0072] An inertial measurement unit (model MPU-9250, size 20 mm x 20 mm x 5 mm, 3 cm from the main control processing unit) is installed on the surface of the fuselage structure 2, generating angular acceleration data (sampling frequency 100 kHz, accuracy 0.001 rad / s2), which is transmitted to the main control processing unit through the SPI bus;

[0073] A high-precision gyroscope (model ADIS16470, size 20 mm x 20 mm x 5 mm, 3 cm from the main control processing unit) is installed on the surface of the fuselage structure 2, generating angular velocity data (sampling frequency 20 kHz, accuracy 0.002 rad / s), which is transmitted to the main control processing unit through the SPI bus;

[0074] The main control processing unit is an aluminum alloy shell package (size 100 mm x 100 mm x 50 mm) fixed in the center of the fuselage structure 2, connected with the sensors in the rotor assembly 1 and the fuselage structure 2 through SPI bus (speed 10 Mbps), I2C bus (speed 400 kHz) and CAN bus (speed 1 Mbps); the main control processing unit integrates an ultra-wideband communication module for transmitting attitude data, torque data and status reports, using BCH (511, 493) error correction coding and 2.4 GHz frequency hopping spread spectrum; the main control processing unit executes the above-mentioned control algorithm through an integrated circuit chip (model Xilinx Zynq-7000, clock frequency 2 GHz) and a digital signal processing chip (model TIC6748, processing capacity 1000 million floating point operations per second), processing airflow disturbance data (10 kHz), vibration spectrum data (5 kHz), torque data (20 kHz), heat flow data (2 kHz), angular acceleration data (100 kHz), angular velocity data (20 kHz); a 32-bit high-precision timer (frequency 100 MHz, precision 10 ns) is used to synchronize all sensor data, and the synchronization error is compensated by Kalman filtering (state vector dimension 6, process noise covariance 0.001), with an error less than 10 microseconds; a pulse width modulation control signal (duty cycle resolution 0.1%, frequency 20 kHz) and a status report are generated, with a calculation delay of 3 ms and a power consumption of 5 watts, transmitted to the rotor motor driver (model DRV8301) through the CAN bus, and transmitted to other unmanned aerial vehicles and ground stations through the ultra-wideband communication module.

[0075] It should be noted that the five control processes in this scheme are the core components of the algorithm, and the hardware device is the execution carrier of the algorithm, and the two together constitute a complete adaptive attitude stabilization control system. The control algorithm is executed by the main control processing unit, and relies on the sensors and actuators of the hardware device to achieve high-precision attitude control; each control process collects multi-modal data through sensors, generates control signals to drive actuators, and cooperatively achieves a control effect of attitude error 0.006° and response time 2 ms. Specific embodiment two:

[0077] As shown in Figs. 1-3 , the following is a supplement to the content in embodiment one:

[0078] (1) Hardware parameters and structure description supplement:

[0079] Rotor assembly details supplement:

[0080] Rotor material performance parameters: the rotor is made of carbon fiber composite material, with a tensile strength of 3500 MPa and a density of 1.8 g / cm 3, weight is 50 grams per piece, with high strength and low weight characteristics, suitable for high dynamic load (maximum load 5 kg) and high speed rotation (5000 revolutions per minute). The rotor surface is polished, with a surface roughness Ra of 0.8 microns, reducing air resistance and improving aerodynamic efficiency.

[0081] Plasma generator pulse power parameters: pulse power (model HV1000) outputs square wave pulses, modulation mode is pulse width modulation (pulse width modulation control), duty cycle range 10% to 50%, adjustable frequency 5kHz to 10kHz, waveform rise time less than 1 microsecond, ensuring high responsiveness of plasma excitation. Power consumption 2 watts, efficiency 85%, passive cooling through heat sink (aluminum, size 30mm x 20mm), operating temperature range -20°C to 60°C.

[0082] Magnetic levitation multi-degree-of-freedom drive system cooling method: high temperature superconducting coil (yttrium barium copper oxide) uses liquid nitrogen cooling system, cooling temperature 77K (-196°C), cooling device contains micro liquid nitrogen tank (capacity 200ml, stainless steel, size 50mm x 50mm x 100mm), installed at the bottom of the fuselage structure (2), power consumption 10 watts, endurance time 4 hours. The coil works through a thermal isolation layer (polyimide, 0.5mm thick) to isolate the rotor base, preventing heat conduction from affecting rotor performance.

[0083] Sensor and actuator interface specifications:

[0084] Dynamic phononic crystal regulation system channel number: six piezoelectric devices correspond to six independent acquisition channels, each generating single-axis vibration data (range 0 to 5 volts), digitized by a 16-bit analog-to-digital converter (resolution 0.01 volts), channel data is aggregated to the I2C bus through a multiplexer (model CD74HC4067), fusion method is weighted average (weight based on vibration amplitude, range 0.1 to 0.9). The I2C interface between the piezoelectric device and the main control processing unit supports a maximum rate of 400kHz, with a single transmission delay of less than 0.1ms.

[0085] Infrared spectrum analyzer wavelength range: infrared spectrum analyzer (FLIRA65) measures wavelength range 3 to 5 microns, suitable for detecting unmanned aerial vehicle surface thermal radiation (temperature range -40°C to 150°C). The analyzer ensures a heat flow data accuracy of 0.01 watts per square meter through internal calibration (based on blackbody radiation model), with a data output format of 16-bit grayscale values, which are filtered by mean filtering (window size 10 frames) before being transmitted to the main control processing unit.

[0086] Master processing unit bus interface quantity: the master processing unit supports 8-way SPI interface (10 megabits per second, connected to plasma generator, inertial measurement unit, gyroscope), 4-way I2C interface (400 kHz, connected to piezoelectric device, infrared spectrum analyzer), 2-way CAN interface (1 megabit per second, connected to Hall sensor, motor driver). Each interface is managed by interrupt priority (based on ARM Cortex-M7 core), ensuring high-frequency sensor (such as inertial measurement unit, 100 kHz) priority processing, and interface compatibility supporting 3.3 volt and 5 volt level.

[0087] Airframe structure mechanics parameters:

[0088] Aluminum alloy frame material: airframe structure 2 uses 6061-T6 aluminum alloy, elastic modulus 70 GPa, yield strength 275 MPa, density 2.7 g / cm 3 , frame weight 1.2 kilograms, corrosion-resistant coating (thickness 0.02 microns) ensures oxidation resistance. The frame is optimized through finite element analysis, with a maximum deformation of less than 0.1 mm (under a load of 10 kg).

[0089] Rotor base and support arm connection: the rotor base 3 is fixed to the support arm (material carbon fiber, length 120 mm, cross-section diameter 8 mm) by M4 stainless steel bolts (torque 10 newton·meters), 4 bolts are used for each base, the maximum bearing torque of the connection point is 50 newton·meters, meeting the dynamic load of high-speed rotation of the rotor (5000 revolutions per minute). The support arm is connected to the top of the airframe structure through a T-shaped slot, with an installation error of less than 0.05 mm, ensuring the accuracy of torque transmission.

[0090] (2) Control algorithm logic supplement:

[0091] Specific definition and function of each control process:

[0092] Geometric phase driving control process:

[0093] Definition: based on the special orthogonal group (SO(3)) differential geometry framework, the geometric phase trajectory is calculated using airflow disturbance and angular acceleration data, and high-precision attitude control signals are generated.

[0094] Independent function: external airflow disturbance (sensitivity 0.1 pascal) and body angular acceleration (accuracy 0.001 rad / s2) are collected through the plasma aerodynamic fine-tuning array and the inertial measurement unit, mapped to a three-dimensional rotation matrix, calibrated rotor torque, and quickly responded to attitude deviation (error <0.006°), suitable for strong wind interference (wind speed 5 m / s) scenarios.

[0095] Connection with other processes: Generate phase state report (512 bytes), containing attitude deviation and disturbance intensity, transmit to fractional order chaotic game control process, provide initial attitude calibration data for game strategy; receive prediction report from multi-scale disturbance prediction control process, optimize phase trajectory.

[0096] Fractional order chaotic game control process:

[0097] Definition: Model chaotic oscillator based on fractional order Caputo derivative, combine multi-agent game theory, optimize rotor moment distribution, support swarm cooperative control.

[0098] Independent function: Use torque data (precision 0.005 Newton-meter) of magnetic suspension multi-degree-of-freedom drive system and angular velocity data (precision 0.002 rad / s) of high-precision gyroscope to distribute rotor speed through Nash equilibrium solver, balance single-machine attitude (error <0.01°) and swarm consistency (error <0.02°), suitable for multi-unmanned aerial vehicle formation mission.

[0099] Connection with other processes: Receive phase state report from geometric phase process, calibrate game initial conditions; generate game report (1024 bytes), transmit to manifold resonance adaptive control process, adjust vibration suppression parameters; synchronize swarm strategy through ultra-wideband communication module.

[0100] Manifold resonance adaptive control process:

[0101] Definition: Based on manifold resonance principle, adjust system vibration frequency spectrum through dynamic phononic crystal regulator, suppress harmful vibration, enhance structural stability.

[0102] Independent function: Collect vibration frequency spectrum data (sensitivity 0.05 Hz), decompose 10 Hz to 2000 Hz frequency band, amplify beneficial vibration (100-500 Hz, gain 1.5 times), suppress high-frequency harmful vibration (1000-2000 Hz, suppression rate 60%), prolong machine body life (fatigue life improved by 30%).

[0103] Connection with other processes: Receive game report from fractional order chaotic game process, optimize vibration suppression weight; generate vibration control report (512 bytes), transmit to topological manifold adaptive control process, assist topological mapping.

[0104] Topological manifold adaptive control process:

[0105] Definition: Based on topological manifold homeomorphism mapping, optimize heat flow distribution and control input, adapt to dynamic load change (change rate >0.1 kg / s).

[0106] Independent function: Collect heat flow data (precision 0.01 watt per square meter) through photonic crystal heat flow regulation module, map to topological space, adjust rotor speed, maintain attitude stability, suitable for load mutation (such as hanging objects) scenarios.

[0107] Connection with other processes: Receive vibration control report of manifold resonance process, optimize topological mapping parameters; Generate topological state report (512 bytes), transmit to multi-scale disturbance prediction control process, provide heat flow distribution data.

[0108] Multi-scale disturbance prediction control process:

[0109] Definition: Based on multi-modal sensor fusion and high-order statistical modeling, predict multi-time scale disturbance (0.1 seconds to 10 seconds), generate feedforward control instructions.

[0110] Independent function: Integrate process reports (phase state, game, vibration control, topological state), predict airflow, vibration, heat flow and other disturbance trends, generate feedforward signals, reduce feedback control delay (response time <2ms), improve system robustness.

[0111] Connection with other processes: Receive all process reports, generate comprehensive prediction report (1024 bytes), feedback to geometric phase driving control process, optimize phase trajectory; Interact with main control processing unit through I2C bus.

[0112] Combined effect:

[0113] Overall function: Five control processes are connected in series through data flow (phase state report → game report → vibration control report → topological state report → prediction report), forming a closed-loop control system, achieving adaptive attitude stabilization (error <0.006°) and cluster coordination (consistency error <0.01°) of multi-rotor unmanned aerial vehicles in complex environments (wind speed 5 meters per second, load change 0.1 kilograms per second, vibration frequency 1000 hertz). Each process has a clear division of labor: geometric phase process quickly corrects attitude, fractional order game process optimizes cluster allocation, manifold resonance process suppresses vibration, topological manifold process adapts to load, multi-scale prediction process feeds forward disturbance, after combination, the system response speed (2ms), control accuracy (0.006°) and fault tolerance ability (communication interruption within 5 seconds to maintain control) are improved.

[0114] Technical advantages: Compared with traditional PID control, the combined process reduces 70% attitude drift (test wind speed 5 meters per second), improves 50% cluster consistency (10 unmanned aerial vehicle formations), and reduces 30% energy consumption (power consumption 5 watts) through multi-modal data fusion and fractional order modeling.

[0115] Ultimate arbitration mechanism for multi-process conflict:

[0116] Global arbitration logic: when multiple processes trigger control signals simultaneously (e.g. attitude error > 0.01°, vibration suppression rate < 50%, cluster consistency error > 0.02°), the master processing unit makes decisions through a global priority arbitration mechanism:

[0117] Disturbance type urgency ranking: air flow disturbance (weight 0.7, threshold 0.2 Pascal), vibration (weight 0.2, threshold 0.1 Hz), heat flow (weight 0.1, threshold 0.1 Watt per square meter), cluster coordination error (weight 0.05, threshold 0.02°).

[0118] Decision-making process:

[0119] Detect the triggering conditions of each process (attitude error, vibration suppression rate, cluster consistency error, etc.).

[0120] Calculate priority score (score = weight × error / threshold, e.g. air flow disturbance score = 0.7 × 0.3 Pascal / 0.2 Pascal = 1.05).

[0121] Select the control signal of the process with the highest score, if the scores are equal (e.g. air flow disturbance and vibration trigger simultaneously), generate the final signal through a weighted fusion algorithm (weighting urgency 0.6, time urgency 0.4).

[0122] The arbitration result is recorded in the status report (256 bytes), including the triggering process, priority score and timestamp, stored in the master processing unit buffer (128 kilobytes).

[0123] Implementation details: arbitration logic is executed by the digital signal processing chip (TIC6748) of the master processing unit, with a calculation period of 1ms and a delay of less than 0.1ms, ensuring real-time performance.

[0124] Data preprocessing and fusion algorithm details:

[0125] Multi-modal sensor fusion algorithm: Extended Kalman Filter (EKF) is used to fuse air flow disturbance (10kHz), angular acceleration (100kHz), torque (20kHz), angular velocity (20kHz), vibration (5kHz), heat flow (2kHz) data, state vector dimension 12 (including three-axis attitude, velocity, acceleration), measurement noise covariance 0.001, process noise covariance 0.0001. The fusion algorithm is executed every 1ms, outputting a comprehensive state vector (accuracy 0.005°), stored in a 128 kilobyte buffer.

[0126] Time synchronization method: 32-bit high-precision timer (100 megahertz) triggers sensor sampling through hardware interrupt, calibration steps include:

[0127] The master processing unit broadcasts synchronization pulses (period 1 ms, precision 10 nanoseconds).

[0128] Each sensor (plasma generator, inertial measurement unit, etc.) records a local timestamp (64 bits, precision 1 nanosecond).

[0129] The time offset is corrected by linear interpolation (maximum offset 10 microseconds).

[0130] Abnormal data filtering: sliding window filtering (window size 50 ms, 50 samples), median filtering threshold set to 3 times the standard deviation, abnormal values (such as air flow disturbance > 1 Pascal) are removed, filtered data is transmitted through the SPI / I 2C bus, delay less than 0.05 ms.

[0131] High-order statistical modeling application: fourth-order cumulant analysis is used to decompose multi-scale disturbances (0.1 seconds, 1 second, 5 seconds, 10 seconds), window overlap rate 50%, frequency resolution 0.01 Hz, applied to air flow disturbance (detecting periodic turbulence), vibration (identifying harmonic peaks), and heat flow (predicting thermal gradient changes), improving prediction accuracy by 30% (compared to second-order statistics).

[0132] Algorithm convergence and robustness proof:

[0133] Newton-Raphson iteration initial value: initial value based on past 1 second phase trajectory mean (calculated by sliding window, window size 1000 samples), if no historical data, use inertial measurement unit angular acceleration (0.001 rad / s^2) as initial value. Convergence is verified by Jacobian matrix analysis, condition number <10, ensuring error <0.001 within 5 iterations.

[0134] Fractional order chaotic oscillator stability: order 0.8, chaotic attractor Lyapunov exponent range 0.1 to 0.5 (verified by numerical simulation), noise immunity supports input data with signal-to-noise ratio > 20 dB, stability is corrected by fractional order integrator (step size 0.01 seconds), error <0.01 Newton·m.

[0135] (3) Cluster collaborative control supplement:

[0136] Communication protocol and anti-interference details:

[0137] BCH (511, 493) error correction coding: code rate 0.964 (493 bits of information / 511 bits of total length), can correct 9-bit errors, redundancy 3.6%, encoding delay 0.05 ms, decoding complexity O(nlogn), suitable for 1 megabit per second transmission. Frequency hopping spread spectrum uses a pseudo-random frequency hopping pattern, with anti-interference ability improved by 20 dB (test environment contains Wi-Fi interference).

[0138] IEEE1588 synchronization implementation: master node broadcasts synchronization frame (64 bytes) every 1 ms, slave node calculates offset and delay through PTPv2 protocol, calibration steps:

[0139] Master node sends Sync message, records sending time T1.

[0140] Slave node receives Sync message, records receiving time T2.

[0141] Master node sends Follow_Up message containing T1.

[0142] Slave node sends Delay_Req message, records sending time T3, master node records receiving time T4.

[0143] Slave node receives Delay_Resp message, calculates offset = (T2-T1+T4-T3) / 2, adjusts local clock.

[0144] Synchronization accuracy 500 nanoseconds, test conditions: communication distance 100 meters, unobstructed environment, environmental interference <10 decibels.

[0145] Cluster fault tolerance mechanism:

[0146] Communication interruption strategy: when communication is interrupted (delay >100 ms), switch to local control, maintain attitude based on the last 5 seconds of game reports (stored in a ring buffer with a capacity of 128 kilobytes and an update period of 0.1 seconds). The game report structure includes rotor allocation ratio (32-bit floating point), fault tolerance state (1-bit Boolean), consistency error (16-bit floating point), and timestamp (64-bit integer). Switching judgment is detected through a heartbeat signal (period 10 ms), and switching delay <1 ms.

[0147] Consistency error quantification: consistency error is calculated by root mean square error, formula is the square root of the sum of squares of the attitude deviation of each UAV, threshold 0.01°. The influence of cluster size (2 to 10) on consistency is verified by simulation, and the consistency error of 10 UAVs is <0.015° (communication delay <5 ms).

[0148] (4) Technical effect quantification supplement:

[0149] Control accuracy test conditions:

[0150] Attitude error 0.006°: test environment: wind speed 0 to 5 meters per second (simulated by wind tunnel, wind speed accuracy 0.1 meters per second), load range 0 to 5 kilograms (dynamic change rate 0.1 kilograms per second), test duration 30 minutes, attitude data verified by laser tracker (accuracy 0.001°).

[0151] Rotor speed accuracy 0.01 revolutions per minute: tested by optical encoder (resolution 0.001 revolutions per minute), speed range 500-5000 revolutions per minute, load rate of change 0.1 kg / s, ambient temperature -10-40°C.

[0152] Vibration suppression and heat flux regulation performance boundaries:

[0153] Vibration suppression rate 60%: test load is motor vibration (frequency 1000 Hz, amplitude 0.1 mm) and external impact (impact force 10 N), measured by accelerometer (model ADXL345, accuracy 0.01 g), suppression rate 60% in the 1000-2000 Hz frequency band, load mass 5 kg.

[0154] Heat flux accuracy 0.01 watts per square meter: test ambient temperature -20-60°C, heat radiation source (infrared lamp, power 100 watts) distance 0.5 meters, infrared spectrum analyzer ensures accuracy by internal calibration (blackbody radiation model), heat flux rate of change <0.1 watts per square meter per second.

[0155] (5) Other not fully disclosed content:

[0156] Hardware reliability and maintenance:

[0157] Sensor fault detection: master control unit runs self-test program every 10 seconds, detects whether sensor output (air flow, torque, etc.) exceeds range (e.g. air flow > 2 Pa), checks redundancy (double sensor comparison, deviation > 0.5% marks fault), fault rate <0.5%. Fault sensor data is replaced with the last valid value, reports fault status (fault type, sensor number, 256 bytes).

[0158] Battery management: uses lithium polymer battery (capacity 5000 mAh, voltage 11.1 V), supports 4 hours of continuous use, charging time 1.5 hours (100W fast charging). Monitors voltage, current, temperature (accuracy 0.1°C) through BMS chip (model BQ40Z50), over-temperature protection threshold 60°C, low power mode triggered by low battery (<10%), preferentially turns off unnecessary sensors (e.g. infrared spectrum analyzer).

[0159] Maintenance cycle: rotor every 1000 hours check wear (maximum wear depth 0.1 mm), magnetic suspension coil every 2000 hours check coolant volume (replace if less than 50%), piezoelectric device every 500 hours calibrate sensitivity, maintenance tools include standard screwdrivers and laser alignment instrument.

[0160] Algorithm fault tolerance and boundary conditions:

[0161] Sensor data missing handling: If a sensor (e.g. inertial measurement unit) has no data for 10 ms, switch to redundant estimation algorithm (autoregressive model based on historical data, window 50 ms), estimation error < 0.1°, switch delay 0.1 ms. Missing reports are logged to status buffer (256 bytes).

[0162] Extreme load scenario: When the load exceeds 7 kg (change rate 0.2 kg / s), the topological manifold adaptive control flow is executed first, limiting the rotor speed < 4000 rpm, the attitude error is relaxed to 0.02°, and the abnormal state (512 bytes) is recorded. The test scenario includes a suspended object (mass 5 kg, oscillation frequency 10 Hz).

[0163] Data storage and logging:

[0164] Data logging mechanism: The main control processing unit has a built-in 64 MB flash memory (model W25Q512) that stores 30 minutes of full data (including air flow, torque, vibration, heat flow, etc., about 50 MB), using a circular coverage strategy, and prioritizing the retention of the last 5 minutes of abnormal data (e.g. attitude error > 0.05°). Data can be exported via USB3.0 (speed 5 Gbps) in < 30 seconds.

[0165] Log format: 1000 logs per second (64 bytes each, including timestamp, sensor ID, data value, status flag) support offline analysis (CSV format compatible with MATLAB) for fault diagnosis and performance evaluation.

[0166] Environmental adaptability:

[0167] Water and dust resistance: The body structure 2 achieves IP65 protection level, the rotor base 3 and sensor interface use silicone sealing rings (thickness 0.5 mm), supporting light rain (rainfall 5 mm / hour) and dust environment (particle diameter < 50 microns).

[0168] High altitude adaptation: Supports altitudes from 0 to 4000 meters, air pressure range 50 to 101 kPa, dynamically corrects rotor power through air pressure sensor (model BMP280, accuracy 0.01 kPa), power adjustment range ±10%.

[0169] It should be noted that in the structure of the present application, in addition to the rotor assembly 1, the body structure 2 and the rotor base 3 shown are limited, the installation positions of other components can be designed or selected according to actual conditions. Specific embodiment three:

[0171] As Figs. 1-3 shown below are the mathematical formulas of the core algorithm described in embodiment one and the corresponding explanations and descriptions:

[0172] Geometric phase driving control flow:

[0173] SO(3) based geometric phase calculation:

[0174]

[0175] R t+1 = R t · exp(Δt· skew(ω)), ω = ω IMU + K p · Δp

[0176] where:

[0177] R ∈ SO(3): 3 × 3 rotation matrix, representing the UAV attitude; ω = [ω x , ω y , ω z ] T : angular velocity vector (unit: rad / s), composed of inertial measurement unit (IMU) angular acceleration ω IMU (accuracy 0.001 rad / s 2 ) and air flow disturbance correction amount Δp (sensitivity 0.1 Pa) collected by the plasma aerodynamic fine-tuning array. K p : proportional gain (tuned to 0.1, ensuring a 2 ms response time). Δt: time step (0.0001 s, corresponding to a 10 kHz sampling frequency). exp: matrix exponential operation on Lie groups, used for attitude integration.

[0178] Problem solved:

[0179] The formula solves the problem of fast attitude correction under strong air flow disturbance (such as wind speed 5 m / s). By mapping air flow disturbance and angular acceleration to the special orthogonal group SO(3), the geometric phase trajectory is calculated, generating accurate rotor torque control signals to control the attitude error within 0.006°.

[0180] Specific implementation means:

[0181] Data acquisition: plasma aerodynamic fine-tuning array (10 kHz, 0.1 Pa) collects air flow disturbance data, IMU (100 kHz, 0.001 rad / s 2 ) collects angular acceleration.

[0182] Calculation flow: generate a 3 × 3 rotation matrix through Lie algebra transformation, use the Newton-Raphson iteration method (5 iterations, error 0.001) to solve R t+1 .

[0183] Control output: Generate pulse width modulation (PWM control) signals (20 kHz, duty cycle resolution 0.1%) and transmit to rotor motor drivers via SPI interface (10 Mbps) to adjust rotation speed (500-5000 RPM, precision 0.01 RPM).

[0184] Priority arbitration: When the attitude error is greater than 0.01°, this process signal is executed first, and the final signal is generated using a weighted fusion algorithm (weights: attitude error 0.6, time urgency 0.4, based on least squares) in case of conflict.

[0185] Data feedback: Generate a phase state report (512 bytes, containing attitude deviation, disturbance intensity, timestamp) and transmit to the fractional-order chaotic game control process via the I2C bus (400 kHz) for initial calibration of game strategy.

[0186] Actual effect: In a 5m / s wind speed, the attitude response time is 2ms, and the error is controlled within 0.006°, which is significantly better than traditional PID control (error about 0.02°).

[0187] Fractional-order chaotic game control process:

[0188] Fractional-order Caputo derivative and Nash equilibrium optimization:

[0189] α = 0.8

[0190]

[0191] Where:

[0192] D α : Fractional-order Caputo derivative (order α = 0.8) used to model the chaotic oscillator, capturing the nonlinear dynamics of the rotor torque. x(t): State vector, including magnetic suspension drive system torque (precision 0.005 N·m) and gyroscope angular velocity (precision 0.002 rad / s). Γ: Gamma function used for fractional-order integration. J: Nash equilibrium cost function, minimizing rotor torque allocation error T i (unit: N·m). k j : Weight of the jth rotor control input u j (PWM control signal). N, M: Number of rotors (usually 4) and number of control inputs.

[0193] Problem solved: This formula solves the single-machine attitude stabilization (error < 0.01°) and multi-UAV swarm coordination control (consistency error < 0.02°) problems. The fractional-order chaotic oscillator captures complex dynamics, and the Nash equilibrium optimizes rotor torque allocation, suitable for multi-agent game scenarios.

[0194] Implementation means:

[0195] Data acquisition: torque is collected by magnetic suspension drive system (20 kHz, 0.005 N·m), angular velocity is collected by gyroscope (20 kHz, 0.002 rad / s).

[0196] Calculation process: D is calculated using the Glivenko-Contnikov discretization (step size 0.01 s) α , J is solved by linear programming, and multi-time scale (0.5 s, 1 s, 2 s, 5 s) game strategy is optimized.

[0197] Control output: pulse width modulation control signal (20 kHz, 0.1% resolution) is generated and transmitted to rotor motor driver (DRV8301) through SPI interface (10 Mbps).

[0198] Cluster synchronization: game report (1024 bytes, including rotor allocation ratio, fault tolerance state, consistency error) is transmitted through ultra-wideband communication module (1 Mbps, BCH (511, 493) error correction), cluster strategy is synchronized, and local control is maintained based on the last 5 seconds of data (128 kB buffer) when communication is interrupted.

[0199] Priority arbitration: when the attitude error is <0.01° but the consistency error is >0.02°, this process signal is executed first, and the weighted fusion algorithm (weights: consistency error 0.5, time urgency 0.5) is used in case of conflict.

[0200] Actual effect: in a formation of 10 drones, the consistency error is <0.015°, the energy consumption is reduced by 30% (5 W), and complex formation tasks are supported.

[0201] Manifold resonance adaptive control process:

[0202] Fast wavelet transform decomposition:

[0203] ψ j,k (t)=2 j / 2 ψ(2 j t-k)

[0204] u(t)=G·X(f target ),f target ∈[100,500]Hz

[0205] Where:

[0206] X(f): vibration spectrum, collected by dynamic phononic crystal system (5 kHz, sensitivity 0.05 Hz). ψ j,k : Daubechies-4 wavelet basis, j, k: scale and translation index. c j,kWavelet coefficients (decomposition level 4). G: Gain (set 1.5 for 100-500 Hz, 0.4 for 1000-2000 Hz). u(t): Control signal, driving magnetostrictive actuator to adjust microstructure.

[0207] Problem solved: Solve the problem of structural fatigue and attitude instability caused by high-frequency vibration (1000-2000 Hz) of the fuselage. Decompose the vibration spectrum by wavelet transform, amplify the beneficial dynamics (100-500 Hz), suppress harmful vibrations (suppression rate 60%), and prolong the service life of the fuselage.

[0208] Specific implementation means:

[0209] Data acquisition: Dynamic phononic crystal system (5 kHz, 0.05 Hz) collects vibration data.

[0210] Calculation process: Fast wavelet transform decomposes the frequency spectrum (10-2000 Hz), and predicts the resonance change within 0.1s by Gaussian process regression (kernel function: square exponential kernel, length scale 0.1s).

[0211] Control output: Adjust the magnetostrictive actuator (voltage 10V, deformation 0.1mm) to change the microstructure geometry, and generate a vibration control report (512 bytes, containing resonance frequency, suppression rate).

[0212] Data transmission: The report is transmitted to the topological manifold control process through the I2C bus (400 kHz).

[0213] Priority arbitration: When the vibration suppression rate is <50%, this process signal is executed preferentially, and the weighted fusion algorithm is used in conflict (weight: suppression rate 0.7, time urgency 0.3).

[0214] Actual effect: 1000-2000 Hz vibration suppression rate reaches 60%, fuselage fatigue life increases by 30%, attitude stability improves by 0.01°

[0215] Topological manifold adaptive control process:

[0216] Topological manifold homeomorphism:

[0217] h: M→N,

[0218] H i = k h ·Q i ,Q i : Heat flow data (W / m 2 )

[0219] Where:

[0220] h: Homophonic mapping function from heat flow manifold M to control manifold N, based on Hough transform (dimension 3). u(t): Control input (pulse width modulation control signal). H i : Target heat flow distribution. Q i : Photonic crystal heat flow data (2kHz, precision 0.01W / m 2 ). k h : Mapping gain (tuned to 0.95).

[0221] Problem solved:

[0222] Solves the problem of heat flow unevenness and attitude deviation caused by dynamic load changes (e.g. 0.1kg / s). Adjusts control input through topology optimization to adapt to load mutations and maintain attitude stability.

[0223] Implementation means:

[0224] Data acquisition: Photonic crystal heat flow module (2kHz, 0.01W / m 2 ) collects heat flow data.

[0225] Calculation process: Simulated annealing method (cooling coefficient 0.95, 50 iterations) optimizes h, generates pulse width modulation control signal (20kHz, 0.1% resolution).

[0226] Control output: Transmitted to rotor driver through SPI interface (10Mbps) to adjust speed.

[0227] Data transmission: Generate topology state report (512 bytes, including attitude deviation, heat flow distribution), transmitted to multi-scale disturbance prediction process through I2C bus.

[0228] Priority arbitration: When the load change rate is >0.1kg / s, this process signal is executed first, and the weighted fusion algorithm is used in conflict.

[0229] Actual effect: When the load changes 0.1kg / s, the attitude error is <0.01°, and the heat flow control precision is 0.01W / m 2 .

[0230] Multi-scale disturbance prediction control process:

[0231] Empirical mode decomposition and high-order statistics:

[0232] c i (t): Modal function

[0233]

[0234] Where:

[0235] x(t): Integrated disturbance data (airflow, torque, vibration, heat flow).c i (t): i-th modal function (modal number 8). r(t): Residual (threshold 0.01). Predicted disturbance (time window 0.1s - 10s). φ i (t): Modal weight, based on 4th order cumulant analysis. EKF: Extended Kalman Filter (state dimension 12, noise covariance 0.001).

[0236] Problem solved: Solve the prediction problem of multi-time scale disturbances (airflow, vibration, heat flow), generate feedforward control instructions, reduce feedback delay, and improve system robustness.

[0237] Specific implementation means:

[0238] Data fusion: Integrate phase state, game, vibration, topology report, and use EKF fusion (state vector dimension 12, noise covariance 0.001).

[0239] Computational flow: Empirical mode decomposition (modal number 8) decomposes 0.1s - 10s disturbance to predict future 0.1s - 10s trends.

[0240] Control output: Generate prediction report (1024 bytes, including disturbance intensity, time window, feedforward instructions), transmitted to rotor driver through SPI interface.

[0241] Data feedback: Report transmitted to geometric phase flow through I2C bus to optimize phase trajectory.

[0242] Priority arbitration: When disturbance intensity > 0.2 Pa, this flow signal is executed first, and in case of conflict, a weighted fusion algorithm is used (weights: disturbance intensity 0.6, time urgency 0.4).

[0243] Actual effect: Prediction accuracy improved by 30%, response time < 2ms, attitude error < 0.006°. Specific embodiment four:

[0245] As Figs. 1-3 shown below is the complete implementation of the content described in the above embodiments:

[0246] This embodiment is based on the technical content in the above examples, and describes the implementation steps of the adaptive attitude stabilization control algorithm and device of the multi-rotor unmanned aerial vehicle in detail, including device installation and preparation, control algorithm execution process, attitude stabilization mechanism heat conduction and vibration regulation principle explanation, and specific use cases. All control processes (geometric phase driving control, fractional order chaotic game control, manifold resonance adaptive control, topological manifold adaptive control, multi-scale disturbance prediction control) are executed in the main control processing unit, suitable for high-precision attitude control and cluster collaborative tasks in complex environments.

[0247] Device installation and preparation:

[0248] Before the unmanned aerial vehicle is assembled, the fuselage structure 2 is selected as the installation basis. The fuselage structure 2 is an aluminum alloy frame (size 50 cm x 30 cm x 20 cm, thickness 5 mm, 6061-T6 aluminum alloy, elastic modulus 70 GPa, yield strength 275 MPa, weight 1.2 kg), the surface is polished (roughness Ra 0.8 μm), and there is no crack or looseness, coated with a 0.02 mm corrosion-resistant coating to resist oxidation, reaching IP65 protection level, suitable for altitudes of 0-4000 m (air pressure 50-101 kPa).

[0249] Rotor assembly 1 manufacturing and installation:

[0250] Rotor manufacturing: carbon fiber composite material rotor (diameter 10-15 cm, tensile strength 3500 MPa, density 1.8 g / cm 3 , single piece weight 50 g) is selected, the surface is polished (roughness Ra 0.8 μm) to reduce air resistance, the maximum load is 5 kg, and the speed range is 500-5000 RPM.

[0251] Plasma aerodynamic fine-tuning array: each rotor surface is coated with a 0.1 μm polyimide nanodielectric coating, and four plasma generators (tungsten alloy high-voltage electrodes and copper alloy ground electrodes, size 20 mm x 5 mm x 5 mm, electrode spacing 0.2 mm, arranged in a 2 x 2 rectangular array, 10-15 cm from the rotor center) are fixed. Air is excited to form plasma by a pulse power supply (model HV1000, 5-10 kHz, square wave pulse, voltage 5 kV, current 0.05 mA, pulse width 10 μs, power consumption 2 W, efficiency 85%, heat sink 30 mm x 20 mm), changes the air flow speed (change amount 0.1 m / s), generates air flow disturbance data (sensitivity 0.1 Pa, sampling frequency 10 kHz, resolution 0.015 Pa), and transmits to the main control processing unit through the SPI interface (10 Mbps).

[0252] Mounting method: the rotor is fixed to the rotor base 3 (carbon fiber support arm, length 120 mm, cross-sectional diameter 8 mm, maximum bearing torque 50 N·m) by M4 stainless steel bolts (torque 10 N·m, 4 per base), connected to the top of the fuselage structure 2 through a T-slot, with an installation error of <0.05 mm.

[0253] Rotor base 3 configuration:

[0254] Magnetic levitation multi-degree-of-freedom drive system: contains magnetic levitation bearings (stainless steel, inner diameter 40 mm, outer diameter 60 mm, thickness 10 mm, friction coefficient <0.001) and magnetic force drive modules (6 yttrium barium copper oxide high-temperature superconducting coils, diameter 5 cm, 100 turns, current 10 A, frequency 50 Hz, magnetic field strength 0.5 T). The magnetic field is measured by a Hall sensor (SS495A, accuracy 0.2 mT, 20 kHz), and the three-axis torque is calculated (accuracy 0.005 N·m, range ±0.5 N·m), transmitted through a CAN bus (1 Mbps).

[0255] Cooling system: the coil is cooled by liquid nitrogen (tank capacity 200 ml, stainless steel, size 50 mm x 50 mm x 100 mm, cooling temperature 77 K, power consumption 10 W, endurance 4 h), and a 0.5 mm polyimide thermal isolation layer is used to prevent heat conduction.

[0256] Fuselage structure 2 sensor and module installation:

[0257] Inertial measurement unit (MPU-9250, 20 mm x 20 mm x 5 mm, 3 cm from the main control processing unit, sampling frequency 100 kHz, accuracy 0.001 rad / s 2 ), generating angular acceleration data, transmitted through an SPI bus (10 Mbps).

[0258] High-precision gyroscope (ADIS16470, 20 mm x 20 mm x 5 mm, 3 cm from the main control processing unit, 20 kHz, accuracy 0.002 rad / s), generating angular velocity data, transmitted through an SPI bus.

[0259] Dynamic phononic crystal regulation system: contains 6 piezoelectric devices (lead zirconate titanate ceramic, 10 mm x 10 mm x 1 mm, range 0-5 V, resolution 0.01 V, 5 kHz, sensitivity 0.05 Hz) and 4 magnetostrictive actuators (nickel-iron alloy, 10 mm x 5 mm x 5 mm, drive voltage 10 V, deformation 0.1 mm), installed in a polyether ether ketone periodic microstructure (period length 0.5 mm, 3D printed, 50 mm x 50 mm x 2 mm), arranged on the top and bottom of the fuselage (5 cm from the main control processing unit). Vibration spectrum data is transmitted through an I2C bus (400 kHz).

[0260] Photonic crystal heat flow regulation module: contains photonic crystal structure (photosensitive resin, period length 0.1 mm, 3D printing, 50 mm x 50 mm x 1 mm) and infrared spectrum analyzer (FLIRA65, wavelength 3-5 pm, resolution 0.01 pm, 30 mm x 30 mm x 20 mm) arranged on the surface of the fuselage, generating heat flow data (range 0-100 W / m 2 , accuracy 0.01 W / m 2 , 2 kHz) transmitted through the I2C bus.

[0261] Master control processing unit: aluminum alloy shell (100 mm x 100 mm x 50 mm) fixed in the center of the fuselage, integrated Xilinx Zynq-7000 (2 GHz) and TIC6748 (1000 MFLOPS). Supports 8-way SPI (10 Mbps), 4-way I2C (400 kHz), 2-way CAN (1 Mbps) interface, uses 32-bit high-precision timer (100 MHz, accuracy 10 ns) to synchronize data, power consumption 5 W. Built-in ultra-wideband communication module (BCH (511, 493) error correction, 2.4 GHz frequency hopping, 1 Mbps) to transmit attitude, torque, and status reports.

[0262] Battery system: lithium polymer battery (5000 mAh, 11.1 V, endurance 4 h, 100 W fast charging, 1.5 h full charge), monitored by BMS chip (BQ40Z50, accuracy 0.1 °C), over-temperature protection threshold 60 °C, low power (<10%) triggers low-power mode, preferentially shutting down the infrared spectrum analyzer.

[0263] Environmental adaptability: sensor interface uses silicone seal ring (thickness 0.5 mm) to achieve IP65 protection level, dynamically corrects rotor power (±10%) through air pressure sensor (BMP280, accuracy 0.01 kPa), adapts to altitudes 0-4000 m, temperatures -20 °C to 60 °C, light rain (5 mm / h), and dust (particles <50 pm) environments.

[0264] Control algorithm implementation method:

[0265] All control processes are executed in the data processing module of the master control processing unit, with the following specific steps:

[0266] Geometric phase drive control process:

[0267] Data acquisition: plasma aerodynamic fine-tuning array (10 kHz, sensitivity 0.1 Pa) collects airflow disturbance data, inertial measurement unit (100 kHz, accuracy 0.001 rad / s 2 ) collects angular acceleration data.

[0268] Data processing: 3x3 rotation matrix based on Lie algebra transformation, geometric phase trajectory calculated by Newton-Raphson iteration (5 iterations, error 0.001), trajectory structure optimized by gradient descent method (step size 0.01, 100 iterations) based on past 30 seconds of data (airflow 300,000 points, angular acceleration 3,000,000 points).

[0269] Control output: Pulse width modulation (PWM control) signal generated (20kHz, duty cycle resolution 0.1%), transmitted to rotor motor driver (DRV8301) through SPI interface (10Mbps), adjusting speed (500-5000RPM, accuracy 0.01RPM), response time 2ms, attitude error controlled within 0.006°.

[0270] Data transmission: Phase state report generated (512 bytes, containing attitude deviation [°], disturbance intensity [Pa], timestamp [1ms accuracy], control signal duty cycle), transmitted to fractional order chaotic game control process through I2C bus (400kHz) for game strategy initial calibration.

[0271] Priority arbitration: When attitude error > 0.01°, this process signal is executed first, and when conflicts occur, the final control signal is generated by weighted fusion algorithm (weights: attitude error 0.6, time urgency 0.4, based on least squares method).

[0272] Effect: In 5m / s wind speed, attitude error <0.006°, response time 2ms, better than traditional PID control (error about 0.02°).

[0273] Fractional order chaotic game control process:

[0274] Data acquisition: Magnetic levitation multi-degree-of-freedom drive system (20kHz, accuracy 0.005N·m) collects torque data, high-precision gyroscope (20kHz, accuracy 0.002rad / s) collects angular velocity data, receives phase state report.

[0275] Data processing: Chaotic oscillator modeled based on fractional Caputo derivative (order 0.8, Grunwald-Letnikov discretization, step size 0.01s), set 0.5s, 1s, 2s, 5s four time scale windows, multi-agent game strategy optimized by Nash equilibrium solver (linear programming, 10 constraints, target is to minimize torque allocation error), initial weights are torque allocation 0.4, angular velocity feedback 0.3, energy optimization 0.2, swarm cooperation 0.1.

[0276] Control output: Generate pulse width modulation control signal (20 kHz, duty cycle resolution 0.1%), transmit to rotor driver through SPI interface, adjust rotation speed (precision 0.01 RPM).

[0277] Cluster synchronization: Transmit game report (1024 bytes, containing rotor allocation ratio [%], fault tolerance state [Boolean], consistency error [°], timestamp [1 ms precision]) through ultra-wideband communication module (1 Mbps, BCH (511, 493) error correction, 2.4 GHz frequency hopping 1000 hops / s, bandwidth 20 MHz) to synchronize cluster strategy. When communication is interrupted (delay > 100 ms), maintain local control based on the last 5 seconds of game reports (128 kB ring buffer) with a consistency error of <0.02°.

[0278] Data transmission: Game report transmitted to manifold resonance adaptive control process through I2C bus for vibration suppression parameter adjustment.

[0279] Priority arbitration: When the attitude error is <0.01° but the consistency error is >0.02°, this process signal is executed first, and in case of conflict, the final signal is generated by a weighted fusion algorithm (weights: consistency error 0.5, time urgency 0.5).

[0280] Effect: 10 drones in formation with consistency error <0.015°, energy consumption reduced by 30% (power consumption 5W), supporting complex formation tasks.

[0281] Manifold resonance adaptive control process:

[0282] Data acquisition: Dynamic phononic crystal regulation system (5 kHz, sensitivity 0.05 Hz) collects vibration spectrum data (10-2000 Hz) and receives game reports.

[0283] Data processing: Decompose the frequency spectrum through fast wavelet transform (Daubechies-4, decomposition level 4), predict future 0.1s resonance changes based on Gaussian process regression (square exponential kernel, length scale 0.1s, noise variance 0.01) based on the last 10 seconds of data (50,000 points).

[0284] Control output: Adjust the periodic microstructure geometry (0.4-0.6 mm) through magnetostrictive actuators (drive voltage 10V, deformation 0.1mm) to change the bandgap frequency (adjustment range 50Hz), amplify beneficial dynamics (100-500Hz, gain 1.5 times), and suppress harmful vibrations (1000-2000Hz, suppression rate 60%).

[0285] Data transmission: Generate vibration control report (512 bytes, containing resonant frequency [Hz], vibration suppression rate [%], attitude stability [°], timestamp [1 ms precision]) and transmit to topological manifold adaptive control flow through I2C bus (400 kHz).

[0286] Priority arbitration: When the vibration suppression rate is <50%, this flow signal is executed preferentially, and the final signal is generated by a weighted fusion algorithm (weights: vibration suppression rate 0.7, time urgency 0.3) in case of conflict.

[0287] Effect: 1000-2000 Hz vibration suppression rate reaches 60%, body fatigue life increases by 30%, attitude stability improves by 0.01°.

[0288] Topological manifold adaptive control flow:

[0289] Data acquisition: Photonic crystal heat flow regulation module (2 kHz, accuracy 0.01 W / m 2 ) collects heat flow data and receives vibration control report.

[0290] Data processing: Based on topological manifold homeomorphism (Hough transform, mapping dimension 3), control input is calculated by nonlinear topological optimization algorithm (simulated annealing method, cooling coefficient 0.95, 50 iterations) based on past 20 seconds of data (40,000 points).

[0291] Control output: Generate pulse width modulation control signal (20 kHz, duty cycle resolution 0.1%) and transmit to rotor driver (DRV8301) through SPI interface, response time 2 ms.

[0292] Data transmission: Generate topological state report (512 bytes, containing attitude deviation [°], heat flow distribution [W / m 2 ], timestamp [1 ms precision]) and transmit to multiscale disturbance prediction control flow through I2C bus.

[0293] Priority arbitration: When the dynamic load change rate is >0.1 kg / s, this flow signal is executed preferentially, and the final signal is generated by a weighted fusion algorithm in case of conflict.

[0294] Effect: When the load changes by 0.1 kg / s, the attitude error is <0.01°, and the heat flow regulation accuracy is 0.01 W / m 2 .

[0295] Multiscale disturbance prediction control flow:

[0296] Data fusion: receive phase state report of geometric phase driven control process, game report of fractional order chaotic game control process, vibration control report of manifold resonance adaptive control process, topological state report of topological manifold adaptive control process, use extended Kalman filter (state vector dimension 12, measurement noise covariance 0.001, process noise covariance 0.0001) to fuse data.

[0297] Data processing: based on high-order statistical modeling (fourth-order cumulant analysis, window 0.1s-10s, overlap rate 50%), predict future 0.1s-10s disturbance trend through time series decomposition algorithm (empirical mode decomposition, mode number 8, residual threshold 0.01), based on past 30 seconds fused data.

[0298] Control output: generate prediction report (1024 bytes, containing disturbance intensity [Pa], prediction time window [s], feedforward control instruction duty cycle), transmit to rotor driver through SPI interface (10Mbps), adjust rotation speed (precision 0.01RPM).

[0299] Data feedback: prediction report is transmitted to geometric phase driven control process through I2C bus for phase trajectory optimization.

[0300] Priority arbitration: when disturbance intensity >0.2Pa, preferentially execute this process signal, in conflict, generate final signal through weighted fusion algorithm (weights: disturbance intensity 0.6, time urgency 0.4).

[0301] Effect: disturbance prediction accuracy improved by 30%, response time <2ms, attitude error <0.006°.

[0302] Cluster cooperative control process:

[0303] Data transmission: transmit attitude data (°), torque data (N·m), angular velocity data (rad / s) through ultra-wideband communication module (1Mbps, BCH (511,493) error correction coding, code rate 0.964, 2.4GHz frequency hopping spread spectrum, 1000 hops / s, bandwidth 20MHz), sampling frequency 1kHz.

[0304] Time synchronization: use master-slave clock synchronization mechanism, master node broadcasts IEEE1588 synchronization frame (64 bytes, period 1ms) through ultra-wideband signal, slave node calibrates local clock through PTPv2 protocol, synchronization accuracy 500ns, ensure cluster attitude consistency error <0.01°.

[0305] Fault-tolerant mechanism: switch to local control when communication is interrupted (delay > 100 ms), adjust rotor speed based on the last 5 seconds of game reports (128 kB ring buffer, update period 0.1 s) to maintain consistency error < 0.02°. Switching is detected by a heartbeat signal (period 10 ms) with a delay < 1 ms.

[0306] Data transmission: generate cluster status reports (256 bytes, including cooperative consistency [°], communication delay [ms], timestamp [1 ms precision]) and transmit them to the main control unit via the I2C bus (400 kHz) for topology manifold and multi-scale perturbation prediction processes.

[0307] Effect: 10 UAV formation consistency error < 0.015°, anti-interference ability improved by 20 dB (test includes Wi-Fi interference).

[0308] Attitude stabilization mechanism: heat conduction and vibration regulation principles:

[0309] Heat conduction principle: in normal flight, rotor high-speed rotation (500-5000 RPM) generates airflow disturbance and heat, the body quickly dissipates heat through an aluminum alloy frame (thermal conductivity 237 W / (m·K)), and the heat flow distribution is uniform (variation rate < 0.01 W / m 2 ). In abnormal situations (such as sudden load changes of 0.1 kg / s), local thermal resistance increases (such as air layer thermal conductivity 0.026 W / (m·K)), resulting in a cliff-like drop in heat flux density (> 0.1 W / m 2 , temperature difference increases by about 10 times). The photonic crystal heat flow regulation module (2 kHz, accuracy 0.01 W / m 2 ) detects heat flow abnormalities and triggers the topology manifold adaptive control process, optimizing rotor speed through homeomorphism mapping to restore attitude stability (error < 0.01°).

[0310] Vibration regulation principle: high-frequency vibration (1000-2000 Hz, amplitude > 0.1 mm) caused by motor or external impact (such as 10N) may cause attitude deviation. The dynamic phononic crystal regulation system (5 kHz, 0.05 Hz) forms a forbidden frequency band through periodic microstructure (period 0.5 mm), and the magnetostrictive actuator (10V, deformation 0.1 mm) adjusts the geometry to suppress high-frequency vibration (1000-2000 Hz, suppression rate 60%) and amplify beneficial dynamics (100-500 Hz, gain 1.5 times), ensuring attitude error < 0.006° and increasing body fatigue life by 30%.

[0311] Use case: UAV formation patrol in complex environments:

[0312] Scenario description: Wind turbine blade inspection in coastal wind farm (wind speed 5-10 m / s, temperature -10°C to 40°C, altitude 0-1000 m, humidity 80% RH, with salt spray interference), 10 drones form a formation, perform 30-minute route inspection task, each drone carries 2 kg of high-definition camera, dynamic load change rate 0.1 kg / s (due to wind or equipment jitter). Task requires attitude error <0.01°, cluster consistency error <0.02°, supports communication interruption within 5 seconds stable flight.

[0313] Implementation steps:

[0314] Device preparation:

[0315] Assemble 10 drones in the above manner, ensure that the rotors, sensors, and main control processing unit are installed, and the battery is fully charged (5000 mAh, endurance 4 h).

[0316] Calibrate rotor power through air pressure sensor (BMP280), adapt to altitude 0-1000 m (air pressure 80-101 kPa).

[0317] Test plasma aerodynamic fine-tuning array (10 kHz, 0.1 Pa) and inertial measurement unit (100 kHz, 0.001 rad / s 2 ), ensure that the airflow disturbance and angular acceleration data are normal.

[0318] Task initialization:

[0319] Main control processing unit loads control algorithm, initializes weights (geometric phase: attitude error 0.6; fractional order game: consistency 0.5; manifold resonance: suppression rate 0.7; topological manifold: load change 0.6; multi-scale prediction: disturbance intensity 0.6).

[0320] Establish 10 drone network through ultra-wideband communication module (1 Mbps, BCH error correction), main node broadcasts IEEE1588 synchronization frame (500 ns precision).

[0321] Set route (cover 1000 m x 500 m area, height 50-200 m), each drone spacing 10 m, inspection speed 5 m / s.

[0322] Control flow execution:

[0323] Geometric phase driven control: real-time collection of 5-10 m / s wind speed disturbance (0.3 Pa peak), generation of pulse width modulation control signal, adjustment of rotation speed, attitude error control at 0.006°, response time 2 ms.

[0324] Fractional-order chaotic game control: Optimize formation assignment based on torque (0.005 N·m) and angular velocity (0.002 rad / s), synchronization game report (1024 bytes), consistency error <0.015°. When 1 UAV communication interruption for 3 seconds, switch to local control, based on 5 seconds of buffered data (128 kB) to maintain stability.

[0325] Manifold resonance adaptive control: Detect 1000 Hz vibration (amplitude 0.15 mm) caused by blade inspection, adjust microstructure bandgap, suppression rate 60%, attitude stability improved by 0.01°.

[0326] Topological manifold adaptive control: Monitor thermal flow anomalies (0.12 W / m 2 Down) caused by load changes (0.1 kg / s), optimize rotation speed, attitude error <0.01°.

[0327] Multi-scale disturbance prediction control: Fusion of airflow, vibration, and thermal flow data, predict 0.1s-10s disturbance trend (accuracy improved by 30%), generate feedforward instructions, response time <2ms.

[0328] Data recording and analysis:

[0329] The main control processing unit stores 30 minutes of full data (64 MB flash memory, 50 MB), including airflow (10 kHz), torque (20 kHz), vibration (5 kHz), and thermal flow (2 kHz) data. 1000 logs are generated per second (64 bytes per log, including timestamp, sensor ID, and data value).

[0330] Export data through USB3.0 (5 Gbps), generate inspection report (CSV format, compatible with MATLAB), analyze attitude deviation, vibration suppression rate, and cluster consistency.

[0331] Maintenance and fault tolerance:

[0332] Self-check sensors every 10 seconds, replace with the latest valid value if 1 piezoelectric device deviation >0.5%, failure rate <0.5%.

[0333] Inspect rotor wear (<0.1 mm) after inspection, replace every 1000 hours, coil coolant is replenished every 2000 hours.

[0334] Implementation effect:

[0335] Attitude accuracy: attitude error 0.005° (verified by laser tracker, accuracy 0.001°), better than traditional PID (0.02°).

[0336] Cluster consistency: 10 UAVs have a consistency error of 0.012°, and within 5 seconds of communication interruption, it maintains 0.018°.

[0337] Vibration and heat flow control: 60% vibration suppression rate at 1000-2000 Hz (verified by accelerometer ADXL345), heat flow accuracy 0.01 W / m 2 .

[0338] Environmental adaptability: adapt to 5-10 m / s wind speed, 80% humidity, salt spray environment, endurance 30 minutes, power consumption 5W.

[0339] Advantages: compared with traditional PID control, attitude drift is reduced by 70%, energy consumption is reduced by 30%, and cluster consistency is improved by 50%. Better than single sensor control (error 0.03°) and infrared thermal imaging navigation (affected by fog), supports remote monitoring and dynamic management. Specific embodiment five:

[0341] As Figs. 1-3 shown below are specific experimental data:

[0342]

[0343]

[0344] The above table is the experimental data of attitude stability and cluster coordination performance. The experimental data is derived from the following tests and simulations:

[0345] Laboratory simulation experiment:

[0346] Environment: wind tunnel simulation (wind speed 0-10 m / s, accuracy 0.1 m / s, temperature -10℃ to 40℃, humidity 80% RH), load simulator (0-5kg, change rate 0.1kg / s).

[0347] Equipment: 10 unmanned aerial vehicle prototypes (assembled according to the embodiment, including rotor assembly, sensor, main control processing unit), laser tracker (accuracy 0.001°) to measure attitude, accelerometer (ADXL345, accuracy 0.01g) to measure vibration, optical encoder (resolution 0.001RPM) to measure speed, infrared spectrum analyzer (FLIRA65, accuracy 0.01W / m 2 ) to measure heat flow.

[0348] Test conditions: 30 minutes of inspection task, simulate wind farm environment (5-10 m / s wind speed, 2kg load, 1000Hz vibration, Wi-Fi interference 10dB).

[0349] Data collection: The master processing unit stores 30 minutes of full data (64 MB flash, 50 MB), including airflow (10 kHz), angular acceleration (100 kHz), torque (20 kHz), angular velocity (20 kHz), vibration (5 kHz), heat flow (2 kHz), generating 1000 logs per second (64 bytes per log), exported through USB3.0 (5 Gbps), MATLAB analysis and verification.

[0350] Field test:

[0351] Location: Coastal wind farm (Eastern China, May 2025, wind speed 5-10 m / s, temperature 5-30 °C, humidity 70-80% RH, salt spray environment).

[0352] Task: 10 drones patrol wind turbine blades in formation (route 1000 m x 500 m, height 50-200 m, spacing 10 m, speed 5 m / s, load 2 kg high-definition camera).

[0353] Measurement equipment: Ground station receives cluster status report (256 bytes, 1 kHz), laser tracker verifies attitude error, infrared thermal imager calibrates heat flow data, accelerometer verifies vibration suppression effect.

[0354] Fault-tolerant test: simulate 1 drone communication interruption for 5 seconds, verify local control performance; introduce Wi-Fi interference (10 dB) to test communication robustness.

[0355] Under 5-10 m / s wind speed and 0.1 kg / s load variation, the attitude error is stabilized at 0.005-0.006°, which is better than traditional PID control (0.02°), thanks to the high precision of geometric phase driving control (0.006°) and the feedforward optimization of multi-scale disturbance prediction (30% improvement in precision).

[0356] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without more limitations, an element defined by an "comprising" statement is not excluded from a process, method, article, or apparatus that includes the element— other than where a contrary implication is explicitly stated or required by context.

[0357] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. An adaptive attitude stabilization control algorithm for multi-rotor unmanned aerial vehicles, comprising a control algorithm and a main control processing unit, characterized in that: The control algorithm includes geometric phase-driven control process, fractional-order chaotic game control process, manifold resonance adaptive control process, topological manifold adaptive control process, and multi-scale disturbance prediction control process. All processes of the control algorithm are executed by the main control processing unit. The geometric phase drive control process acquires airflow disturbance data through a plasma aerodynamic fine-tuning array, acquires angular acceleration data through an inertial measurement unit, calculates the control input based on the geometric phase on a special orthogonal group, and adjusts the rotor torque. The fractional-order chaotic game control process collects torque data through a magnetic levitation multi-degree-of-freedom drive system, collects angular velocity data through a high-precision gyroscope, optimizes the multi-agent game strategy based on a fractional-order chaotic oscillator, allocates rotor control input, and supports cluster collaborative control. The manifold resonance adaptive control process acquires vibration spectrum data through a dynamic phonon crystal modulation system, adjusts the frequency spectrum based on the manifold resonance mode, amplifies beneficial dynamics, and suppresses harmful vibrations. The topological manifold adaptive control process acquires heat flow data through a photonic crystal heat flow control module with an accuracy of 0.01 watts per square meter, and optimizes the control input based on topological manifold homeomorphism mapping to adapt to dynamic load changes. The multi-scale disturbance prediction control process processes the data generated by the aforementioned process through a multi-modal sensor fusion algorithm, predicts the disturbance trend based on high-order statistical modeling, and generates feedforward control commands. The control algorithm consists of five control processes: "geometric phase-driven control process, fractional-order chaotic game control process, manifold resonance adaptive control process, topological manifold adaptive control process, and multi-scale disturbance prediction control process". These processes are connected in series through data flow to form a closed-loop control system. The geometric phase process quickly corrects the attitude, the fractional-order game process optimizes cluster allocation, the manifold resonance process suppresses vibration, the topological manifold process adapts to the load, and the multi-scale prediction process feeds forward disturbances.

2. The adaptive attitude stabilization control algorithm for multi-rotor UAVs according to claim 1, characterized in that: The geometric phase drive control process uses Lie algebra transformation to map the airflow disturbance data collected by the plasma aerodynamic fine-tuning array and the angular acceleration data collected by the inertial measurement unit to a 3×3 three-dimensional rotation matrix. The number of rows in the matrix corresponds to the number of rows in the plasma generator array. The geometric phase trajectory is calculated using the Newton-Raphson iterative method to generate a pulse width modulation control signal, which is transmitted to the rotor motor driver through a serial peripheral interface to adjust the rotor speed. The response time is 2ms, and the attitude error is controlled within 0.006°. When the attitude error is greater than 0.01°, the control signal of the geometric phase drive control process is executed first. In case of conflict, a weighted fusion algorithm is used to generate the final control signal. The geometric phase drive control process is based on the airflow disturbance data and angular acceleration data of the past 30 seconds. It adjusts the phase trajectory structure through a topology optimization algorithm and generates a phase state report containing attitude deviation, disturbance intensity, timestamp and control signal. The report is transmitted to the fractional-order chaotic game control process through the internal integrated circuit bus for the initial condition calibration of the game strategy.

3. The adaptive attitude stabilization control algorithm for multi-rotor UAVs according to claim 2, characterized in that: The fractional-order chaotic game control process receives phase state reports; it models a chaotic oscillator based on the fractional-order Caputo derivative, sets four time scale windows of 0.5 seconds, 1 second, 2 seconds, and 5 seconds, optimizes the multi-agent game strategy through the Nash equilibrium solver, generates pulse width modulation control signals, and transmits them to the rotor motor driver through the serial peripheral interface to adjust the rotor speed. When the attitude error is less than 0.01° but the cluster consistency error is greater than 0.02°, the control signal of the fractional chaotic game control process is executed first. When there is a conflict, the final control signal is generated by the weighted fusion algorithm. The fractional-order chaotic game control process is based on torque and angular velocity data from the past 60 seconds. It calculates game weights using a fractional-order integrator and generates a game report containing rotor allocation ratios, fault tolerance states, cooperative consistency, and timestamps. This report is transmitted to other UAVs via an ultra-wideband communication module to synchronize the cluster control strategy. In the event of a communication interruption, local attitude control is maintained based on the game report from the most recent 5 seconds, and the rotor speed is adjusted to maintain a consistency error of less than 0.02°. The game report is transmitted to the manifold resonance adaptive control process via an internal integrated circuit bus for adjusting vibration suppression parameters.

4. The adaptive attitude stabilization control algorithm for multi-rotor UAVs according to claim 3, characterized in that: The manifold resonance adaptive control process acquires vibration spectrum data through a dynamic phonon crystal control system and receives a game report generated by a fractional-order chaotic game control process. Based on fast wavelet transform to decompose the frequency spectrum, it adjusts the geometry of the periodic microstructure through a magnetostrictive actuator, changes the bandgap frequency, amplifies beneficial dynamics, and suppresses harmful vibrations. Based on the vibration data of the past 10 seconds, it predicts the resonance change in the next 0.1 seconds through a Gaussian process regression algorithm, generating a vibration control report containing the resonant frequency, vibration suppression rate, attitude stability, and timestamp. This report is transmitted to the topology manifold adaptive control process via an internal integrated circuit bus for topology mapping optimization. When the vibration suppression rate is below 50%, the control signal of the manifold resonance adaptive control process is executed first. In case of conflict, a weighted fusion algorithm is used to generate the final control signal.

5. The adaptive attitude stabilization control algorithm for multi-rotor UAVs according to claim 4, characterized in that: The topology manifold adaptive control process acquires heat flow data through a photonic crystal heat flow control module with a sampling frequency of 2kHz and an accuracy of 0.01 watts per square meter. It also receives vibration control reports generated by the manifold resonance adaptive control process. Based on the topology manifold homeomorphism mapping algorithm, the control input is calculated through a nonlinear topology optimization algorithm to generate a pulse width modulation control signal, which is transmitted to the rotor motor driver through a serial peripheral interface with a response time of 2ms. The topology manifold adaptive control process generates a topology status report containing attitude deviation, heat flow distribution, and timestamps based on the heat flow data of the past 20 seconds. This report is transmitted to the multi-scale disturbance prediction control process via the internal integrated circuit bus for disturbance trend prediction. When the dynamic load change rate is greater than 0.1 kg / s, the control signal of the topology manifold adaptive control process is executed first. In case of conflict, the final control signal is generated through a weighted fusion algorithm.

6. The adaptive attitude stabilization control algorithm for multi-rotor UAVs according to claim 5, characterized in that: The multi-scale disturbance prediction and control process uses a multi-modal sensor fusion algorithm to receive phase state reports generated by the geometric phase-driven control process, game reports generated by the fractional-order chaotic game control process, vibration control reports generated by the manifold resonance adaptive control process, and topological state reports generated by the topological manifold adaptive control process, generating comprehensive disturbance data; and decomposes disturbances at four time scales of 0.1 seconds, 1 second, 5 seconds, and 10 seconds based on high-order statistical modeling. Based on the fused data from the past 30 seconds, a time series decomposition algorithm is used to predict the disturbance trend for the next 0.1 to 10 seconds, generating a prediction report that includes the disturbance intensity, prediction time window, and feedforward control commands. This report is transmitted to the rotor motor driver via a serial peripheral interface to adjust the rotor speed. The prediction report is also transmitted to the geometric phase drive control process via an internal integrated circuit bus for phase trajectory optimization. When the disturbance intensity is greater than 0.2 Pascals, the control signals of the multi-scale disturbance prediction control process are executed first. In case of conflict, a weighted fusion algorithm is used to generate the final control signal.

7. The adaptive attitude stabilization control algorithm for multi-rotor UAVs according to claim 3, characterized in that: The fractional-order chaotic game control process supports cluster collaborative control. It transmits attitude, torque, and angular velocity data via an ultra-wideband communication module, employing error correction coding and frequency hopping spread spectrum in the 2.4GHz band, keeping the data transmission rate below 1 megabit per second. A master-slave clock synchronization mechanism is used, with the master node broadcasting an IEEE 1588 synchronization frame via an ultra-wideband signal, and slave nodes calibrating their local clocks using a precision clock protocol, ensuring a cluster attitude consistency error of less than 0.01°. Based on the attitude, torque, and angular velocity data from the past 5 seconds, the game strategy of the fractional-order chaotic game control process is synchronized. In the event of a communication interruption, each UAV maintains local attitude control based on the game report from the most recent 5 seconds, adjusting rotor speed to maintain a consistency error of less than 0.02°. The ultra-wideband communication module generates a cluster status report that includes coordination consistency, communication delay, and timestamps. This report is transmitted to the main control processing unit via the internal integrated circuit bus for use by the topology manifold adaptive control process and the multi-scale disturbance prediction control process.

8. An adaptive attitude stabilization control device for a multi-rotor unmanned aerial vehicle (UAV), used to implement the adaptive attitude stabilization control algorithm for a multi-rotor UAV as described in any one of claims 1-7, characterized in that: The system includes a rotor assembly (1) and a fuselage structure (2). Rotor bases (3) are installed on both sides of the top of the fuselage structure (2). Rotors are installed on the top of each rotor base (3). There are four rotor bases (3). The rotors on the top of the four rotor bases (3) constitute a complete rotor assembly. The main control processing unit is implanted inside the fuselage structure (2). Each rotor in the rotor assembly (1) is made of carbon fiber composite material and is installed on the rotor base (3). The rotor base (3) is fixed to the fuselage structure (2) by a mechanical arm. Each rotor surface is coated with a nano-level dielectric coating to fix the plasma generator of the plasma aerodynamic fine-tuning array. The plasma aerodynamic fine-tuning array contains four plasma generators distributed in a rectangular array. The plasma generators apply a 5kHz to 10kHz multi-band pulse electric field through a pulse power supply to excite the air on the rotor surface to form plasma, change the local airflow speed, generate airflow disturbance data, and transmit it to the main control processing unit through a serial peripheral interface. The rotor base (3) has a built-in magnetic levitation multi-degree-of-freedom drive system, which includes a magnetic levitation bearing and a magnetic drive module. The magnetic levitation bearing is installed at the center of the rotor base (3) to support rotor rotation and reduce mechanical friction. The magnetic drive module includes six high-temperature superconducting coils, which are evenly distributed in the rotor base (3). The coils generate a magnetic field when energized. The change in magnetic field is measured by a Hall sensor. The three-axis torque data is calculated based on the current-torque conversion formula and transmitted to the main control processing unit through the controller local area network bus. The fuselage structure (2) is an aluminum alloy frame with a main control processing unit fixed in the center, and a dynamic phonon crystal control system and a photonic crystal heat flow control module embedded in the top and bottom. The dynamic phonon crystal control system comprises six piezoelectric devices and four magnetostrictive actuators, which are installed within periodic microstructures arranged at the top and bottom of the fuselage structure (2). The piezoelectric device detects the deformation of the microstructure, generates vibration spectrum data, and transmits it to the main control processing unit through the internal integrated circuit bus; the magnetostrictive actuator adjusts the geometry of the microstructure through the driving voltage to achieve vibration suppression; the photonic crystal heat flow control module includes a photonic crystal structure and an infrared spectrometer. The photonic crystal structure is arranged on the surface of the fuselage structure (2), and the infrared spectrometer measures the thermal radiation intensity, generates heat flow data, and transmits it to the main control processing unit through the I2C bus; An inertial measurement unit is installed on the surface of the fuselage structure (2) to generate angular acceleration data, which is then transmitted to the main control processing unit via the SPI bus. A high-precision gyroscope is mounted on the surface of the fuselage structure (2) to generate angular velocity data, which is then transmitted to the main control processing unit via the SPI bus. The main control processing unit is encapsulated in an aluminum alloy shell and fixed in the center of the fuselage structure (2). It is connected to the rotor assembly (1) and the sensors in the fuselage structure (2) via SPI bus, I2C bus and CAN bus. The main control processing unit integrates an ultra-wideband communication module for transmitting attitude data, torque data and status reports. It uses error correction coding and 2.4GHz frequency band frequency hopping spread spectrum. The main control processing unit executes control algorithms through integrated circuit chips and digital signal processing chips to process airflow disturbance data, vibration spectrum data, torque data, heat flow data, angular acceleration data and angular velocity data. It uses a 32-bit high-precision timer to synchronize all sensor data. The synchronization error is compensated by Kalman filtering, and the error is less than 10 microseconds. It generates pulse width modulation control signals and status reports with a calculation delay of 3ms and a power consumption of 5 watts. The signals are transmitted to the rotor motor driver via CAN bus and to the surrounding UAVs and ground station via the ultra-wideband communication module.

Images