Flight attitude control method based on inertial navigation technology

Abstract

The present invention provides a flight attitude control method based on inertial navigation technology, which relates to the field of aircraft control technology and includes an attitude control system. The attitude control system includes a quantum inertial navigation fusion module, a distributed multi-source collaborative inertial navigation module, a real-time error compensation module based on deep learning, an energy collection and self-powered inertial system module, a distributed attitude data management module based on blockchain, a multimodal fusion and virtual twin collaborative control module, and a terahertz communication and inertial navigation fusion anti-interference module. The present invention uses a quantum annealing algorithm combined with a compressed sensing algorithm to optimize quantum sensor data, and then combines the D-S evidence theory and the particle swarm optimization algorithm in the distributed multi-source collaborative inertial navigation module to fuse multi-source data, which can deeply mine the deep connection and complementary information between different sensor data, break the simple data level superposition limitation, and greatly improve the accuracy of the attitude information after fusion.

Description

Technical Field

[0001] The present invention relates to the field of aircraft control technology, in particular to a flight attitude control method based on inertial navigation technology. Background Art

[0002] With the rapid development of modern aerospace, aircraft are playing an increasingly critical role in many fields such as civil, scientific research, and military. As the core "brain" of an aircraft, the performance of the flight attitude control system directly determines whether the aircraft can perform its mission stably and safely. Once a deviation occurs in the flight attitude control, it will not only cause the mission to fail, but may also cause serious safety accidents and cause inestimable losses. Therefore, the development of a flight attitude control system with high precision, high reliability, and strong adaptability has become a key problem that needs to be overcome in the aerospace field.

[0003] At present, flight attitude control technology mainly includes the following types: Traditional inertial navigation system: mainly relies on accelerometers to sense the acceleration of the aircraft in all directions, and uses gyroscopes to measure the angular velocity of the aircraft. Based on the principles of Newtonian mechanics, the attitude angle and position information of the aircraft are obtained by integrating the acceleration and angular velocity. In conventional flight scenarios, such as the stable cruising phase of civil airliners, it can provide basic navigation and attitude solution data for the aircraft to meet general flight control needs.

[0004] Satellite navigation assistance system: With the help of global satellite navigation systems, such as the US GPS and China's BeiDou, aircraft can receive signals transmitted by satellites and accurately calculate their own position coordinates and running speed. Combining satellite navigation data with the inertial navigation system can complement each other, suppress the accumulation of inertial navigation errors to a certain extent, and improve the overall navigation accuracy. In an open flight environment, such as flying over the sea, the satellite navigation signal is stable, which can provide high-precision positioning information for the aircraft, assisting inertial navigation to achieve more accurate attitude control.

[0005] Simple sensor fusion technology: convert the air pressure value measured by the pressure sensor into altitude information, use the magnetometer to detect the geomagnetic field to obtain the aircraft's heading information, and then fuse it with other sensor data. In this way, we try to obtain the aircraft's status information from multiple dimensions to make attitude perception more comprehensive. For example, when flying at low altitude, the altitude data of the pressure sensor can assist inertial navigation to make altitude corrections.

[0006] Application of classic control algorithms: represented by PID control algorithm, through the three links of proportion (P), integration (I), and differentiation (D), according to the deviation between the current attitude of the aircraft and the target attitude, the corresponding control quantity is calculated, and then the aircraft's control surfaces, engine thrust and other actuators are adjusted to achieve the control of the aircraft's attitude. When the aircraft's flight state and external environment are relatively stable, such as when a drone is flying in a straight line at a constant speed in an open field, PID control can better maintain the stability of the aircraft's attitude and achieve basic control functions.

[0007] Although existing technologies have played a certain role in aircraft flight attitude control, there are still the following key issues that need to be solved: 1. Insufficient depth and accuracy of multi-source data fusion: Currently, the fusion of various sensor data mostly remains at the simple data level superposition, failing to fully explore the deep connection and complementary information behind different sensor data, resulting in limited accuracy of fused attitude information, making it difficult to meet the demand for high-precision attitude perception in complex flight scenarios.

[0008] 2. Poor adaptability to complex environments: Whether it is sensors or control algorithms, their performance will be significantly reduced when facing complex environments such as extreme temperatures, strong electromagnetic interference, and drastic changes in high-altitude airflow. For example, the zero-bias stability of traditional inertial navigation sensors deteriorates in high-temperature environments, satellite navigation signals are easily interrupted under strong electromagnetic interference, and classic control algorithms cannot effectively adjust control strategies when airflow changes drastically.

[0009] 3. It is difficult to balance real-time performance and reliability: When the aircraft is flying at high speed or performing complex tasks, the attitude control system needs to have extremely high real-time response capabilities and reliability. However, existing attitude control technologies often find it difficult to simultaneously ensure real-time performance and reliability when processing large amounts of data and responding to emergencies. For example, traditional data processing methods have delays when processing multi-source big data, resulting in untimely attitude adjustments; and some strategies to improve real-time performance may sacrifice system reliability and increase flight risks.

[0010] 4. Challenges of energy supply and management: The attitude control system of the aircraft requires a continuous and stable energy supply, especially in long-term flight missions. The existing energy collection and management technology is inefficient and cannot meet the high-efficiency energy demand of the attitude control system. At the same time, there are also deficiencies in energy storage and distribution, making it difficult to ensure stable power support for attitude control equipment under various flight conditions.

[0011] Therefore, a flight attitude control method based on inertial navigation technology is needed to solve the above problems. Summary of the invention

[0012] Technical issues solved

[0013] In view of the deficiencies in the prior art, the present invention provides a flight attitude control method based on inertial navigation technology, which solves the problems in the above background technology.

[0014] Technical Solution To achieve the above objectives, the present invention is implemented through the following technical solutions: a flight attitude control method based on inertial navigation technology, including an attitude control system, wherein the attitude control system includes a quantum inertial navigation fusion module, a distributed multi-source collaborative inertial navigation module, a real-time error compensation module based on deep learning, an energy collection and self-powered inertial system module, a distributed attitude data management module based on blockchain, a multimodal fusion and virtual twin collaborative control module, and a terahertz communication and inertial navigation fusion anti-interference module. The attitude control system controls the attitude of the aircraft by sequentially executing the following steps through inertial navigation technology: Sp1. Startup and initialization: Turn on the attitude control system, initialize the quantum inertial navigation fusion module, distributed multi-source collaborative inertial navigation module, real-time error compensation module based on deep learning, energy collection and self-powered inertial system module, distributed attitude data management module based on blockchain, multimodal fusion and virtual twin collaborative control module and terahertz communication and inertial navigation fusion anti-interference module to ensure that each module is in a ready state; Sp2, quantum inertial navigation fusion: using atomic interferometer gyroscopes, combined with accelerometers and gyroscopes in traditional inertial navigation sensors, to collect acceleration data and angular velocity data of the aircraft in three axes, and obtain high-precision raw data. Traditional inertial navigation systems perform preliminary data processing such as filtering, denoising and zero bias correction on these raw data, and perform short-term attitude solution through strapdown algorithms to obtain attitude angles (pitch angle, yaw angle, roll angle) as the initial value for subsequent precise solution. At the same time, the high-precision raw data collected by quantum sensors is processed by combining quantum annealing algorithms with compressed sensing algorithms to optimize data transmission and storage. The processed data is transmitted to the distributed multi-source collaborative inertial navigation module, the real-time error compensation module based on deep learning, and the multi-modal fusion and virtual twin collaborative control module. Sp3, Distributed multi-source collaborative inertial navigation: deploy multiple inertial navigation units at both ends of the aircraft's wings, the head and tail of the fuselage, etc., to fully perceive the aircraft's motion state. Fuse visual (optical cameras obtain surrounding visual image information, infrared cameras obtain thermal infrared images at night or in low visibility environments for identifying landmarks and obstacles) and air pressure (combine the altitude-air pressure model to obtain aircraft altitude information) multi-type sensor data, use the data fusion algorithm that combines D-S evidence theory with the particle swarm optimization algorithm to process multi-source sensor data, dynamically adjust the sensor data weights, and achieve full-scale and accurate perception of the aircraft's attitude. Transmit the fused high-precision attitude data to the real-time error compensation module based on deep learning, the multi-modal fusion and virtual twin collaborative control module, and the distributed attitude data management module based on blockchain. Sp4. Real-time error compensation based on deep learning: Establish a neural network model, input the original inertial navigation data from the quantum inertial navigation fusion module (measurement data output by the accelerometer and gyroscope, and the speed and position data after preliminary processing), environmental parameters (temperature, air pressure, magnetic field strength of the aircraft environment, and geographic information of the flight area), and fusion data of the distributed multi-source collaborative inertial navigation module, and train through the echo state network algorithm combined with the deep belief network algorithm to automatically extract data features, learn error laws, and predict and compensate errors in real time during flight. Feedback the attitude data after error compensation to the multimodal fusion and virtual twin collaborative control module, and transmit the compensation related information to the distributed attitude data management module based on blockchain for recording. Sp5. Energy collection and self-powered inertial system: Energy collection devices are integrated on the wing surface, fuselage skin and engine air inlet of the aircraft. The piezoelectric material energy collector on the wing surface generates electricity by using wing vibration. The temperature difference power generation material of the fuselage skin generates electricity based on the temperature difference between the inside and outside of the fuselage. The engine air inlet converts the kinetic energy of airflow into electricity through the principle of electromagnetic induction. The collected electrical energy is processed by rectification and voltage stabilization circuits and stored in energy storage elements, and then powered by the voltage stabilization circuit to the inertial navigation system. The firefly algorithm is combined with the adaptive dynamic programming algorithm to optimize the energy conversion and storage strategy of the energy collection device and improve the energy conversion efficiency. Continuously provide stable power supply to each power-consuming module, especially each inertial navigation-related module, to ensure its normal operation. Sp6, distributed attitude data management based on blockchain: The inertial navigation unit, visual sensor and air pressure sensor on the aircraft are used as blockchain network nodes, and the attitude data from the distributed multi-source collaborative inertial navigation module and the compensation information of the real-time error compensation module based on deep learning are encrypted and recorded in the blockchain with a timestamp. Distributed ledger technology is used to achieve data security and efficient transmission. The homomorphic encryption algorithm is combined with the zero-knowledge proof algorithm to achieve efficient data verification and sharing while ensuring data security. Each module can obtain authenticated and reliable data from the blockchain according to needs. Sp7, multimodal fusion and virtual twin collaborative control: Inertial navigation data (acceleration, angular velocity, position and speed) and laser radar point cloud data (obtaining three-dimensional spatial information of the surrounding environment), millimeter-wave radar distance and speed data (detecting the distance and relative speed of target objects), visual image data (images taken by optical cameras and infrared cameras) and other multimodal data are decomposed into multiple intrinsic modal functions through the variational mode decomposition algorithm to extract the characteristic information of each mode. The multimodal data features are input into the virtual model, and the dynamic equations and flight mechanics principles in the model are used to build a real-time virtual twin model to simulate the real-time flight state of the aircraft, and compare it with the preset normal flight attitude range. When it exceeds the range, it is predicted to be abnormal attitude. After analyzing the cause of the abnormality, according to the pre-set control strategy library, combined with fuzzy logic reasoning, a feedback optimization strategy is generated to adjust the aircraft control parameters and realize attitude adjustment. The attitude adjustment command is transmitted to the actuator of the aircraft, and the attitude adjustment related information is fed back to other modules, such as updating the attitude data record in the blockchain. Sp8, anti-interference of terahertz communication and inertial navigation fusion: Taking advantage of the high bandwidth and strong directionality of terahertz communication, a high-speed data transmission link is established to encrypt and transmit attitude data and verify it in real time. A low-density parity check code algorithm combined with a fractional-order Fourier transform algorithm is used to improve the anti-interference ability and data transmission rate of the terahertz communication link. The accuracy and timeliness of data interaction between modules are guaranteed, especially the reliable transmission of key attitude data to external nodes such as the ground control center, while receiving external commands and passing them to the corresponding modules. Sp9, Circular monitoring and continuous adjustment: Each module runs continuously, repeating the above steps, and real-time monitoring of aircraft attitude, sensor data, energy status, communication quality and other information. According to the actual flight conditions and system feedback, dynamically optimize the parameters and operation strategies of each module to ensure that the aircraft attitude remains stable and the system operates efficiently and reliably.

[0015] Preferably, in the quantum inertial navigation fusion module, the specific steps of combining the quantum annealing algorithm with the compressed sensing algorithm are as follows: Sp101. Use quantum annealing algorithm to initially optimize the raw data collected by quantum sensor. The raw data include acceleration and angular velocity raw data measured by accelerometer and gyroscope. Use quantum annealing algorithm to initially optimize these raw data. By setting different parameter combinations of quantum bits to represent data processing, simulate the annealing process of quantum system from high temperature to low temperature. During the annealing process, the state of quantum bits will change. Through continuous iteration, find the optimal parameter combination that minimizes data processing error and reduces data noise interference. The optimized data includes removing measurement noise and correcting zero bias. Sp102. Use the compressed sensing algorithm to sparsely represent and reconstruct the data processed by the quantum annealing algorithm. According to the sparse characteristics of the data, by designing a suitable measurement matrix, the high-dimensional data is projected into a low-dimensional space to reduce the amount of data and optimize data transmission and storage efficiency.

[0016] Preferably, in the distributed multi-source collaborative inertial navigation module, the steps of the data fusion algorithm using the DS evidence theory and the particle swarm optimization algorithm are as follows: Sp201. Use DS evidence theory to preliminarily fuse the data collected by the inertial navigation unit, optical camera, infrared camera, and air pressure sensor, calculate the basic probability distribution function of each data source, and evaluate the reliability of the data. Use DS evidence theory to first determine the basic probability distribution function of each data source based on the characteristics and historical data of each sensor. For the inertial navigation unit, assign probability based on its measurement accuracy and stability, evaluate the reliability of the data, and then synthesize the evidence of the same target aircraft attitude from different sensors to obtain the preliminary results after fusion. Sp202. The particle swarm optimization algorithm is used to dynamically adjust the data weights after fusion of the DS evidence theory. The position of the particles represents the weight distribution scheme. Each particle represents a possible weight combination. Through information sharing and iterative search among particles, the degree of match between the fused data and the actual flight status is used as the fitness function to find the optimal weight combination and improve the accuracy of the fused data.

[0017] Preferably, in the real-time error compensation module based on deep learning, the training steps of the echo state network algorithm combined with the deep belief network algorithm are as follows: Sp301, using the echo state network algorithm to build the initial network model, determine the network's reserve pool size and connection weight parameters, take the inertial navigation raw data (acceleration, angular velocity, speed, position) and environmental parameters (temperature, air pressure, magnetic field strength, geographic information) as input, and use its sparse connection characteristics and dynamic reserve pool to perform preliminary feature extraction on the input data. The extracted features include data change trends, periodic features, and correlation features with environmental parameters; Sp302. Use the features extracted by the echo state network as the input of the deep belief network. Through layer-by-layer greedy training and reverse fine-tuning, further mine the data features, learn the error rules of the habitual navigation system, and realize real-time error compensation.

[0018] Preferably, in the energy collection and self-powered inertial system module, the optimization steps of combining the firefly algorithm with the adaptive dynamic programming algorithm are as follows: Sp401. Use the firefly algorithm to optimize the working parameters of the energy collection device, where the energy collection device includes a piezoelectric vibration energy collector, a temperature difference power generation module, and an electromagnetic induction airflow energy collector. The optimized working parameters mainly include the resonant frequency of the piezoelectric vibration energy collector, the material ratio of the temperature difference power generation module, and the number of coil turns of the electromagnetic induction airflow energy collector. Simulate the luminescence and movement behavior of fireflies, take energy conversion efficiency as the objective function, find the optimal parameter combination for energy collection, and improve energy conversion efficiency. Sp402. Adopt an adaptive dynamic programming algorithm to dynamically adjust the charging and discharging strategy of the energy storage element according to the energy collection and consumption situation, so as to adapt to the energy requirements under different flight conditions such as take-off, cruising, landing and maneuvering flight. During the take-off phase, the energy demand is large, which is mainly used to start the engine and increase the altitude of the aircraft. The energy storage element needs to discharge quickly to meet the high power demand. During the cruising phase, the energy demand is relatively stable. The energy collection device charges the energy storage element while meeting the current power consumption. During the landing phase, the energy demand is reduced, but the power supply of key equipment needs to be guaranteed. During maneuvering flight, the energy demand fluctuates greatly, and the charging and discharging strategy is adjusted according to the intensity of the maneuvering action.

[0019] Preferably, in the distributed posture data management module based on blockchain, the implementation steps of the homomorphic encryption algorithm combined with the zero-knowledge proof algorithm are as follows: Sp501, use homomorphic encryption algorithm to encrypt posture data, calculate and process data in ciphertext state, and ensure the security of data during transmission and storage; Sp502, using zero-knowledge proof algorithm, without leaking the specific content of the data, prove the authenticity and integrity of the data to other nodes (i.e. other sensor nodes on the aircraft and ground control center nodes), to achieve efficient data verification and sharing.

[0020] Preferably, in the multimodal fusion and virtual twin collaborative control module, the processing steps of the variational mode decomposition algorithm combined with the fuzzy neural network algorithm are as follows: Sp601. The variational mode decomposition algorithm is used to decompose multimodal data such as inertial navigation data, lidar point cloud data, millimeter-wave radar data and visual image data. For inertial navigation data, it is decomposed into high-frequency dynamic change mode and low-frequency trend mode according to the frequency characteristics of the signal. The lidar point cloud data is decomposed into near-field target mode, mid-field environment mode and far-field background mode according to the spatial distribution characteristics. The millimeter-wave radar data is decomposed into stationary target mode and moving target mode according to the target characteristics. The visual image data is decomposed into edge mode, texture mode and target mode according to texture and shape characteristics. The feature information of each mode is extracted, including the rapid attitude change characteristics in the high-frequency dynamic change mode of inertial navigation and the target distance and shape characteristics in the near-field target mode of lidar.

[0021] Sp602. Input the extracted feature information into the fuzzy neural network. The fuzzy logic fuzzifies the feature information by setting fuzzy rules and converts it into fuzzy language variables. The neural network uses its powerful learning ability to further process and fuse the fuzzified features. According to the degree of correlation between the input features and the posture abnormality, it determines whether there is a posture abnormality and makes a prediction, thereby improving the virtual twin model's prediction accuracy for aircraft posture abnormalities.

[0022] Preferably, in the terahertz communication and inertial navigation fusion anti-interference module, the application steps of the low-density parity check code algorithm combined with the fractional-order Fourier transform algorithm are as follows: Sp701, using a low-density parity check code algorithm to encode the transmitted posture data, increase the redundant information of the data, and improve the error correction capability of the data during transmission. During encoding, the original posture data is grouped according to the set check matrix, and redundant check bits are added to each group of data, so that the receiving end can detect and correct the error bits that occur during the transmission process according to the check bits, increase the redundant information of the data, and improve the error correction capability of the data during transmission; Sp702. Use the fractional Fourier transform algorithm to transform the terahertz communication signal. Select the fractional order according to the channel characteristics such as multipath fading and noise interference, transform the signal from the time domain to the fractional Fourier domain, adjust the time-frequency distribution of the signal, and enhance the signal's anti-interference ability and transmission rate.

[0023] Preferably, the data interaction and collaborative control mechanism between the modules of the system is as follows: Sp801, by building a dedicated data transmission channel on the aircraft's avionics system bus, using a high-speed serial communication protocol, and establishing a unified data bus, each module transmits and interacts data through the data bus to ensure data consistency and timeliness; Sp802, formulate a data interaction protocol, clarify the responsibilities and processes of each module in the process of data sending, receiving and processing, and ensure the stable operation of the system. The data interaction protocol includes data format definition (using binary-coded fixed-length frame format), transmission rate setting (dynamically adjusted according to data volume and real-time requirements) and data verification rules (using CRC verification or hash verification). During data transmission, the sending module is responsible for packaging the data according to the protocol format, adding verification bits and sending it. The receiving module is responsible for receiving data and verifying data integrity. If the verification passes, data parsing is performed. The processing module performs corresponding processing based on the parsed data. The processing steps include attitude solution and error compensation to ensure the stable operation of the system. Sp803. Task priorities are divided according to the urgency and importance of each module's tasks. When the aircraft encounters an emergency, the inertial navigation and attitude control module tasks have the highest priority, and system computing resources and data transmission bandwidth are allocated first. During normal flight, data recording and non-critical sensor data processing modules have a lower priority. Through the task scheduler, system resources such as CPU computing time and memory resources are dynamically allocated according to priority to achieve efficient collaborative control.

[0024] Preferably, the data interaction and collaborative control mechanism between the modules of the system is as follows: Sp901. By building a dedicated data transmission channel on the aircraft's avionics system bus, using a high-speed serial communication protocol, and establishing a unified data bus, each module transmits and interacts data through the data bus to ensure data consistency and timeliness; Sp902, formulate a data interaction protocol, clarify the responsibilities and processes of each module in the process of data sending, receiving and processing, and ensure the stable operation of the system. The data interaction protocol includes data format definition, transmission rate setting and data verification rules. In the process of data sending, the sending module is responsible for packaging the data according to the protocol format, adding the check bit and sending it. The receiving module is responsible for receiving the data and verifying the data integrity. If the verification passes, the data is parsed. The processing module performs corresponding processing according to the parsed data. The processing steps include attitude solution and error compensation to ensure the stable operation of the system; Sp903. Assign task priorities according to the urgency and importance of each module's tasks. When the aircraft encounters an emergency, the inertial navigation and attitude control module tasks have the highest priority, and system computing resources and data transmission bandwidth are allocated first. During normal flight, data recording and non-critical sensor data processing modules have lower priority. Through the task scheduler, CPU computing time and memory resources and other system resources are dynamically allocated according to priority to achieve efficient collaborative control.

[0025] Beneficial Effects

[0026] The present invention provides a flight attitude control method based on inertial navigation technology, which has the following beneficial effects: 1. The present invention uses a quantum annealing algorithm combined with a compressed sensing algorithm to optimize quantum sensor data, and then combines the DS evidence theory and the particle swarm optimization algorithm in the distributed multi-source collaborative inertial navigation module to fuse multi-source data. It can deeply explore the deep connections and complementary information between different sensor data, break the limitations of simple data layer superposition, and greatly improve the accuracy of the fused attitude information, meet the stringent requirements for high-precision attitude perception in complex flight scenarios, and provide a solid data foundation for the precise control of aircraft in complex environments.

[0027] 2. The quantum inertial navigation fusion module of the present invention utilizes the quantum characteristics of the atomic interferometer gyroscope to reduce interference in complex environments and maintain high-precision measurement; the terahertz communication and inertial navigation fusion anti-interference module adopts a low-density parity check code algorithm combined with a fractional-order Fourier transform algorithm to enhance the communication anti-interference capability; the energy collection and self-powered inertial system module adjusts the energy strategy according to different environments and flight conditions through an adaptive dynamic programming algorithm. These measures enable the entire system, whether it is the sensor or the control algorithm, to maintain stable performance in complex environments such as extreme temperatures, strong electromagnetic interference, and drastic changes in high-altitude airflow, thereby ensuring the safe flight of the aircraft.

[0028] 3. The real-time error compensation module based on deep learning of the present invention utilizes the echo state network and the deep belief network to learn and predict errors in real time and compensate for them, and quickly processes large amounts of data; the distributed attitude data management module based on blockchain uses a homomorphic encryption algorithm combined with a zero-knowledge proof algorithm to achieve efficient data verification and sharing while ensuring data security, thereby ensuring the reliability of data transmission and processing, so that when the aircraft is flying at high speed or performing complex tasks, it can have extremely high real-time response capabilities and reliability at the same time, and adjust the aircraft attitude in a timely and accurate manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a system structure diagram of the present invention; Figure 2 It is a model simulation diagram of the present invention; Figure 3 is a system flow chart of the present invention; Figure 4 This is a flow chart of the quantum inertial navigation fusion of the present invention. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Specific embodiment one: like Figure 1-Figure 4 As shown, the flight attitude control method based on inertial navigation technology includes an attitude control system, which includes a quantum inertial navigation fusion module, a distributed multi-source collaborative inertial navigation module, a real-time error compensation module based on deep learning, an energy collection and self-powered inertial system module, a distributed attitude data management module based on blockchain, a multimodal fusion and virtual twin collaborative control module, and a terahertz communication and inertial navigation fusion anti-interference module. The attitude control system executes each step through the above modules: The quantum inertial navigation fusion module performs the quantum inertial navigation fusion steps: using atomic interferometer gyroscopes, combined with traditional inertial navigation sensors, the traditional inertial navigation sensors here include accelerometers and gyroscopes, and using a calculation method that combines quantum annealing algorithms with compressed sensing algorithms to process the high-precision raw data collected by quantum sensors and optimize data transmission and storage. The traditional inertial navigation system is responsible for preliminary data processing and short-term attitude solution. The accelerometer is used to collect the acceleration data of the aircraft in three axes, and the gyroscope is used to collect the angular velocity data of the aircraft. These data are high-precision raw data. By collecting these raw data, basic motion information can be provided for the aircraft attitude solution, thereby accurately calculating the attitude changes of the aircraft. The traditional inertial navigation system refers to a system based on the principles of classical mechanics. , a system that infers the position, velocity and attitude of an aircraft by integrating acceleration and angular velocity. The system is responsible for preliminary data processing, which is to filter and de-noise the raw data collected by the accelerometer and gyroscope and perform zero-bias correction to remove noise and errors in the data. The preliminary data is mainly the original measurement values ​​of the above acceleration and angular velocity. The short-term attitude solution uses the processed acceleration and angular velocity data to quickly calculate the attitude angle (pitch angle, yaw angle, roll angle) of the aircraft in a short time through the strapdown algorithm, providing initial values ​​for subsequent precise solutions. The high-precision raw data collected by the quantum sensor is processed by combining the quantum annealing algorithm with the compressed sensing algorithm to optimize data transmission and storage. The traditional inertial navigation system is responsible for preliminary data processing and short-term attitude solution; The distributed multi-source cooperative inertial navigation module executes the distributed multi-source cooperative inertial navigation steps: deploy multiple inertial navigation units at both ends of the aircraft's wings, the head and tail of the fuselage, etc. The two ends of the wings can sense the aircraft's roll motion, the head of the fuselage can effectively sense the aircraft's pitch and yaw motion, and the tail assists in verifying and supplementing relevant motion information. These parts can fully and more accurately sense the aircraft's motion state, and integrate visual and air pressure multi-type sensor data. Among them, the visual sensor includes an optical camera and an infrared camera. The optical camera is used to obtain visual image information around the aircraft. The infrared camera can work at night or in low visibility environments to obtain thermal infrared images for identifying landmarks and obstacles; the air pressure sensor is used to measure atmospheric pressure, and combined with the altitude-air pressure model, the altitude information of the aircraft is obtained. These data are collected to assist the inertial navigation unit, compensate for the errors accumulated over time by the inertial navigation system, improve the accuracy and reliability of the aircraft's attitude solution, and realize the full-scale and accurate perception of the aircraft's attitude. The data fusion algorithm that combines the DS evidence theory with the particle swarm optimization algorithm is used to process the multi-source sensor data, dynamically adjust the sensor data weights, and realize the full-scale and accurate perception of the aircraft's attitude. The real-time error compensation module based on deep learning performs the real-time error compensation steps: establish a neural network model, input the original inertial navigation data and environmental parameters, train through the echo state network algorithm combined with the deep belief network algorithm, automatically extract data features, learn error laws, and predict and compensate errors in real time during flight. The input original data here refers to the measurement data output by the accelerometer and gyroscope in the inertial navigation system, as well as the speed and position data after preliminary processing. The environmental parameters include the temperature, air pressure, magnetic field strength of the aircraft's environment and the geographical information of the flight area. These environmental factors will affect the measurement accuracy of the inertial navigation system. The echo state network algorithm is combined with the deep belief network algorithm for training, automatically extracting data features, learning error laws, and predicting and compensating errors in real time during flight. The energy collection and self-powered inertial system module performs energy collection and self-powered steps: integrating energy collection devices on the wing surface, fuselage skin and engine air inlet of the aircraft to convert mechanical energy into electrical energy to power the inertial navigation system, and equipping energy storage elements. The firefly algorithm is combined with the adaptive dynamic programming algorithm to optimize the energy conversion and storage strategy of the energy collection device and improve the energy conversion efficiency. When the energy collection device is installed on the wing surface, an energy collector made of piezoelectric material is used. When the aircraft is flying, the wing vibrates, and the piezoelectric material deforms under the action of mechanical stress, thereby generating electric charge and realizing the conversion of mechanical energy into electrical energy. The fuselage skin uses temperature difference power generation materials to generate electrical energy based on the temperature difference between the inside and outside of the fuselage. The engine air inlet can use the principle of electromagnetic induction to convert the kinetic energy of the airflow into electrical energy. After that, the collected electrical energy is stored in the energy storage element through rectification and voltage stabilization circuit processing, and then the inertial navigation system is powered by the voltage stabilization circuit to ensure its stable operation. The distributed attitude data management module based on blockchain performs attitude data management steps: the inertial navigation unit, visual sensor and air pressure sensor on the aircraft are used as data acquisition devices of the blockchain network nodes, and the attitude data is encrypted and recorded in the blockchain with a timestamp. Distributed ledger technology is used to achieve data security and efficient transmission. The homomorphic encryption algorithm is combined with the zero-knowledge proof algorithm to achieve efficient data verification and sharing while ensuring data security. The multimodal fusion and virtual twin collaborative control module executes the multimodal fusion and virtual twin collaborative control steps: fusing the inertial navigation data with the multimodal data of the lidar and millimeter-wave radar, building a real-time virtual twin model, predicting posture anomalies and feeding back optimization strategies. The multimodal data specifically includes inertial navigation data (acceleration, angular velocity, position and velocity), lidar point cloud data (used to obtain three-dimensional spatial information of the surrounding environment), millimeter-wave radar distance and velocity data (detecting the distance and relative speed of the target object), and visual image data (images taken by optical cameras and infrared cameras). These multimodal data are decomposed by the variational mode decomposition algorithm to decompose the complex signal into multiple intrinsic mode functions. In actual application, the lidar point cloud data is decomposed into structures of different scales according to the spatial distribution characteristics. Construct modality: Visual image data is decomposed into different image modes according to texture and shape features, and feature information of each mode is extracted, including the structural features of the lidar point cloud, the texture and shape features of the visual image. When constructing a real-time virtual twin model, multimodal data features are input into the virtual model, and the dynamic equations and flight mechanics principles in the model are used to simulate the real-time flight state of the aircraft, and compared with the preset normal flight attitude range. When it exceeds the range, it is predicted to be an abnormal attitude. After analyzing the cause of the abnormality, a feedback optimization strategy is generated based on the pre-set control strategy library combined with fuzzy logic reasoning, and the control parameters of the aircraft are adjusted to achieve attitude adjustment. The variational mode decomposition algorithm is combined with the fuzzy neural network algorithm to extract and fuse features of multimodal data to improve the prediction accuracy of the virtual twin model. The terahertz communication and inertial navigation fusion anti-interference module executes the communication and inertial navigation fusion anti-interference steps: utilizing the high bandwidth and strong directionality of terahertz communication to establish a high-speed data transmission link, encrypt the transmission attitude data and verify it in real time, and use a low-density parity check code algorithm combined with a fractional-order Fourier transform algorithm to improve the anti-interference capability and data transmission rate of the terahertz communication link.

[0032] The following supplements are made based on the above content and the contents of the claims: The quantum sensor uses an atomic interferometer gyroscope, which is based on cold atom interferometry technology and uses lasers to manipulate ultra-cold atomic clouds to produce interference fringes to accurately measure angular velocity.

[0033] The specific parameters of the deep model are as follows: Echo State Network (ESN): The number of neurons in the reserve pool is set to

[500] , and the connection weights of the reserve pool are randomly initialized with a uniform distribution of [-0.3, 0.3]. The input weights and output weights are trained and optimized through a supervised learning algorithm. The sparsity of the reserve pool is set to [0.2], that is, 20% of the connection weights are non-zero values. This sparse structure helps improve the efficiency of network training and reduce the risk of overfitting. During the training process, the input data enters the reserve pool after being transformed by the weight matrix. The dynamic interaction between neurons in the reserve pool produces complex time series features, which are mapped to the final error prediction value by the output weight.

[0034] Deep Belief Network (DBN): It is composed of three layers of restricted Boltzmann machines (RBM). The number of neurons in the visible layer of the first layer RBM is consistent with the number of ESN output features, which is

[100] . The number of neurons in the hidden layer is set to

[150] . The input data is extracted and reduced in dimension through an unsupervised learning algorithm. The number of neurons in the hidden layer of the second layer RBM is

[120] , which further abstracts and fuses the features extracted in the first layer. The number of neurons in the hidden layer of the third layer RBM is

[80] , which is used to generate higher-level feature representations. The last layer is connected to a Softmax classifier, which performs supervised learning based on the error labels of the training data to achieve accurate prediction of inertial navigation errors.

[0035] Blockchain related technical details: The blockchain adopts a consortium chain architecture, and the consensus mechanism uses the Practical Byzantine Fault Tolerance (PBFT) algorithm. In the consortium chain, the participating nodes mainly include key sensor nodes on the aircraft and ground control center nodes. The PBFT algorithm can tolerate up to [(n-1) / 3] malicious nodes (n is the total number of nodes) while ensuring data consistency, greatly improving the reliability and security of the system. In actual applications, when a node generates a new posture data block, it will be broadcast to other nodes, and each node will verify and reach consensus through the PBFT algorithm to ensure that the data block is correctly added to the blockchain. Specific embodiment 2: like Figure 1-Figure 4 As shown, the following experimental data are provided according to the content in Example 1: Table 1 in the following table is a data comparison of the existing traditional inertial navigation and the quantum inertial navigation fusion module control in this technical solution; Compare Projects Traditional inertial navigation Quantum Inertial Navigation Fusion Module Flight duration 10 hours 10 hours Flight altitude 10,000 meters 10,000 meters Pitch angle error 5° 0.1° Yaw angle error 3° 0.08° Roll angle error 4° 0.12° Error reduction ratio - 98% Error growth trend Rapid increase in the late Linear slow rise

[0037] The following table 2 compares the data when using the compensation module and not using the compensation module: Test conditions Position error growth without compensation module (per hour) Position error growth using compensation module (per hour) Example test (8000m altitude, 0.8 Mach speed, cloudy weather, 6 hours flight) 5000m altitude, subsonic speed, clear sky 80 meters 8 meters Uncompensated: 480 meters, 48 ​​meters after compensation 8000m altitude, transonic speed, light rain 120m 12 meters Uncompensated: 720 meters Compensated: 72 meters 10,000m altitude, subsonic speed, cloudy 90 meters 9 meters Before compensation: 540 meters After compensation: 54 meters 5000m altitude, transonic speed, cloudy sky 100 meters 10 meters Before compensation: 600 meters After compensation: 60 meters 12,000m altitude, subsonic speed, light snow 110m 11 meters Before compensation: 660 meters After compensation: 66 meters The following Table 3 shows the test data of multimodal fusion and virtual twin collaborative control module: Test environment Abnormal posture prediction accuracy False Positive Rate Example test (mountainous area, heavy rain weather) Mountainous area, sunny day 96% 2% Predict potential anomalies 6 seconds in advance and avoid them after adjustments Canyon, light rain 94% 3.5% Accurately predict attitude abnormalities 4 seconds in advance, and the aircraft responds promptly Plain, strong wind 95% 3% The abnormality is predicted 5 seconds in advance and the aircraft responds promptly City sky, smog 93% 4% Predict 3 seconds in advance and maintain stable flight through strategic adjustment Sea, rainstorm 95% 3% 5-second early warning to successfully avoid abnormal posture risks Specific embodiment three:

[0038] like Figure 1-Figure 4 As shown, the key algorithm mentioned in Example 1 is analyzed in detail below, including its core mathematical formula and explanation: 1. Quantum inertial navigation fusion module 1.1. Quantum sensor (atomic interferometer gyroscope):

[0039] is the phase difference of the atomic interference fringes, is the atomic mass, is the effective arm length of the atom interferometer, is the reduced Planck constant, is the angular velocity to be measured. This formula reflects the linear relationship between the phase difference of the interference fringes and the angular velocity. The angular velocity can be calculated by measuring the phase difference.

[0040] Derivation source: Based on the principle of atomic matter wave interference, when atoms propagate in different paths, the angular velocity causes the path optical path to change and produce a phase difference. From the perspective of quantum mechanics, the phase change of atomic matter waves is related to the state of motion. This formula is obtained through a series of theoretical derivations (involving knowledge such as wave functions and phase relationships in quantum mechanics).

[0041] Parameter value range and basis: Taking the common rubidium atom as an example, kg, which is determined by the physical properties of atoms. , depends on the interferometer design and experimental conditions. J s is a physical constant.

[0042] Application logic and steps: (1) Prepare a cold atomic cloud to a suitable quantum state; (2) Use lasers to manipulate atoms to propagate along different paths to form interference fringes; (3) Measure the phase difference of the interference fringes ; (4) Calculate the angular velocity according to the formula .

[0043] Problems and means to be solved: Traditional gyroscopes have limited accuracy and stability. Atomic interferometer gyroscopes use quantum properties to achieve high-precision angular velocity measurement and provide accurate data for inertial navigation.

[0044] 1.2. Combination of quantum annealing algorithm and compressed sensing: Compressed Sensing:

[0045] is a low-dimensional measurement vector, is the measurement matrix, is the original high-dimensional sparse signal, and the formula represents the measurement matrix For the original signal Perform linear projection to obtain the measured value .

[0046] Derivation source: Based on the signal sparse representation theory. When the signal is sparse in a certain transform domain, the original signal can be restored through a small number of linear measurements. This formula is a mathematical expression of the linear measurement process.

[0047] Parameter value range and basis: Measurement Matrix : It needs to satisfy the constrained isometry property (RIP), such as Gaussian random matrix, where the element values ​​are usually ( is the number of matrix rows) to ensure the accuracy of signal reconstruction.

[0048] Signal : It is required that the number of non-zero elements in a certain basis is much smaller than the signal dimension, that is, it satisfies sparsity.

[0049] Application logic and steps: (1) Optimizing the measurement matrix using the quantum annealing algorithm , find the optimal structure; (2) collect a small amount of measurement values ;(3)By solving optimization problems (such as st Restore the original signal .

[0050] Solution and means: Reduce the amount of data collected and the processing complexity of quantum sensors, and use the quantum annealing algorithm to optimize the measurement matrix to improve reconstruction accuracy.

[0051] 2. Real-time error compensation module based on deep learning 2.1. Echo State Network (ESN):

[0052] It's time The reservoir state vector, It's time The input vector is is the input weight matrix, is the recursive weight matrix, is the activation function. This formula describes the update process of the reservoir state with the input and the state of the previous moment.

[0053] Derivation source: Based on the idea of ​​recursive neural network, the concept of reserve pool is introduced to simplify training. The new state is obtained by weighted summation of input and previous state and then through nonlinear activation function.

[0054] Parameter value range and basis: The number of neurons in the reserve pool is generally determined based on the complexity of the problem and the amount of data, such as dozens to hundreds. For example, when processing simple inertial navigation error prediction, it can be set to 50-100.

[0055] and : Elements are usually randomly initialized in [-1,1]. To ensure network stability, The spectrum radius is scaled and generally controlled between 0.5 and 1.

[0056] Application logic and steps: (1) Initialization And the output weight matrix ; (2) Input training data in sequence; (3) Update the status of the reserve pool ;(4)Calculate the output ; (5) Calculate the error and update (e.g. solving with linear regression); (6) then using the trained ESN to predict and compensate for the error in real-time data.

[0057] Problem solving and means: Learn the dynamic characteristics of habitual navigation errors, predict and compensate errors in real time, and improve navigation accuracy.

[0058] 2.2. Deep Belief Network (DBN): Restricted Boltzmann Machine (RBM) energy function:

[0059] is the visible layer vector, h is the hidden layer vector, The visible layer neurons The bias, is the hidden layer neuron The bias, is the connection weight of the visible layer and the hidden layer neurons, and the energy function describes the joint energy of the visible layer and the hidden layer states.

[0060] Derivation source: Based on the Boltzmann distribution in statistical physics, the concepts of visible layer and hidden layer are introduced and applied to machine learning.

[0061] Parameter value range and basis: The number of neurons in the visible layer and hidden layer is determined according to the dimension of the input data and the complexity of the features. For example, if the input is 10-dimensional data, the first visible layer can be set to 10 neurons, and the hidden layer can be set to 15-20 neurons. Initialized randomly in [-1, 1] and adjusted during training.

[0062] Application logic and steps: (1) Train RBM layer by layer, with the output of the previous hidden layer as the input of the next visible layer, and use the contrastive divergence algorithm to train each RBM; (2) Update the bias and weights, and stack the trained RBMs to form a DBN; (3) Add a Softmax classifier on top of the DBN, and use a supervised learning algorithm to fine-tune the DBN to improve the error prediction accuracy.

[0063] Problem solving and means: Automatically learn the complex characteristics of habitual navigation errors to improve error prediction accuracy. 3. Multimodal fusion and virtual twin collaborative control module 3.1 Variational Mode Decomposition (VMD) algorithm: The variational mode decomposition algorithm is used to decompose the signal into multiple intrinsic mode functions (IMFs). Its core is to solve a variational problem: Variational Problem:

[0064] in: It is decomposed The intrinsic mode functions, is the center frequency of each eigenmode function, is the original signal, is the Dirac function, Represents a convolution operation.

[0065] VMD solves the above variational problem iteratively and decomposes the original signal into multiple intrinsic mode functions with different center frequencies. Each intrinsic mode function has a certain physical meaning and reflects the characteristics of the signal at different frequency components.

[0066] Derivation source: The VMD algorithm is based on the idea of ​​modal decomposition in signal processing. By introducing variational problems and constraints, the effectiveness and stability of the decomposition results are guaranteed.

[0067] Parameter value range and basis: Mode number : It needs to be determined according to the characteristics of the signal and the purpose of analysis. Usually, the appropriate value.

[0068] Penalty Factor and Lagrange multipliers Used to control the solution process of variational problems. Its value needs to be adjusted according to the specific situation and can generally be determined through experiments.

[0069] 3.2 Fuzzy Neural Network: Fuzzy neural network combines the advantages of fuzzy logic and neural network. Its core formula includes fuzzification, rule reasoning and defuzzification: Fuzzification:

[0070] in: is the input variable Belongs to fuzzy sets The membership degree of is a fuzzy set The center of is a fuzzy set Width.

[0071] Rule reasoning:

[0072] in: It is The excitation intensity of the rule, It is The conclusion parameter of the rule.

[0073] Defuzzification:

[0074] Parameter value range and basis: in is the output of the fuzzy neural network.

[0075] Explanation and Description: Fuzzification maps input variables to the membership of fuzzy sets, rule reasoning calculates the excitation strength and conclusion of each rule based on fuzzy rules and membership, and defuzzification converts the results of fuzzy reasoning into clear output values.

[0076] Derivation source: The theoretical basis of fuzzy neural network is the combination of fuzzy logic and neural network. Through the process of fuzzification, rule reasoning and defuzzification, it can realize the modeling and control of complex nonlinear systems.

[0077] Application logic and steps: (1) Determine the number of fuzzy sets and the type of membership function; (2) Initialize the center of the fuzzy set ,width And conclusion parameters ; (3) Fuzzify the input data and calculate the membership degree; (4) Reason according to the fuzzy rules and calculate the excitation intensity and conclusion of each rule; (5) Defuzzify the data to obtain a clear output value; (6) Use the training data to adjust the parameters of the fuzzy neural network to improve the accuracy of the output.

[0078] In multimodal fusion, data from different sensors have different characteristics and uncertainties. Fuzzy neural networks can handle these uncertainties and fuse multimodal data through fuzzy rules and the learning ability of neural networks to achieve accurate prediction and coordinated control of aircraft attitude anomalies.

[0079] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprising a reference structure" do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0080] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A flight attitude control method based on inertial navigation technology, including an attitude control system, characterized in that: The attitude control system includes a quantum inertial navigation fusion module, a distributed multi-source collaborative inertial navigation module, a real-time error compensation module based on deep learning, an energy collection and self-powered inertial system module, a distributed attitude data management module based on blockchain, a multimodal fusion and virtual twin collaborative control module, and a terahertz communication and inertial navigation fusion anti-interference module. The attitude control system controls the aircraft attitude by sequentially executing the following steps through inertial navigation technology: Sp1, Startup and initialization: Start the attitude control system; Sp2, Quantum Inertial Navigation Fusion: Using atomic interferometer gyroscopes, combined with accelerometers and gyroscopes in traditional inertial navigation sensors, to collect acceleration data and angular velocity data of the aircraft in three axes, and obtain high-precision raw data; Sp3, Distributed multi-source collaborative inertial navigation: Deploy inertial navigation units at both ends of the aircraft's wings, the head and tail of the fuselage to fully perceive the aircraft's motion status; Sp4. Real-time error compensation based on deep learning: Establish a neural network model, input the initial inertial navigation data processed by the quantum inertial navigation fusion step, environmental parameters, and fused data processed by the distributed multi-source collaborative inertial navigation step, and train it through the echo state network algorithm combined with the deep belief network algorithm; Sp5, Energy harvesting and self-powered inertial processing: Energy harvesting devices are integrated on the wing surface, fuselage skin and engine air inlet of the aircraft, and the firefly algorithm is combined with the adaptive dynamic programming algorithm; Sp6, attitude data management: the inertial navigation unit, visual sensor and air pressure sensor on the aircraft are used as blockchain network nodes, and the attitude data from the distributed multi-source collaborative inertial navigation step and the compensation information from the real-time error compensation step based on deep learning are encrypted and recorded in the blockchain with a timestamp; Sp7, multimodal fusion and virtual twin collaborative control: decompose the inertial navigation data and the surrounding three-dimensional spatial information obtained by the lidar, the distance and speed data of the millimeter-wave radar, and the multimodal data of the visual image data into intrinsic mode functions through the variational mode decomposition algorithm, and extract the characteristic information of each mode; Sp8, terahertz communication and inertial navigation fusion anti-interference: Utilize the high bandwidth and strong directionality of terahertz communication to establish a high-speed data transmission link, encrypt the transmission attitude data and verify it in real time; Sp9, cyclic monitoring and continuous adjustment: each module runs continuously and repeats the above steps.

2. The flight attitude control method based on inertial navigation technology according to claim 1, characterized in that: The quantum inertial navigation fusion step is performed by a quantum inertial navigation fusion module. When processing the initial data, a short-term attitude solution is performed by a strapdown algorithm to obtain the attitude angle data of the pitch angle, yaw angle and roll angle as the initial value for the subsequent precise solution. At the same time, the high-precision raw data collected by the quantum sensor is processed by combining the quantum annealing algorithm with the compressed sensing algorithm to optimize data transmission and storage. The processed data is transmitted to the distributed multi-source collaborative inertial navigation module, the real-time error compensation module based on deep learning, and the multimodal fusion and virtual twin collaborative control module. Preliminary data processing and short-term attitude solution are performed by a traditional inertial navigation system. The specific steps of combining the quantum annealing algorithm with the compressed sensing algorithm are as follows: Sp201. Use quantum annealing algorithm to initially optimize the raw data collected by quantum sensor. The raw data include acceleration and angular velocity raw data measured by accelerometer and gyroscope. Use quantum annealing algorithm to initially optimize these raw data. By setting different parameter combinations of quantum bits to represent data processing, simulate the annealing process of quantum system from high temperature to low temperature. During the annealing process, the state of quantum bits will change. Through continuous iteration, find the optimal parameter combination that minimizes data processing error and reduces data noise interference. The optimized data includes removing measurement noise and correcting zero bias. Sp202. Use the compressed sensing algorithm to sparsely represent and reconstruct the data processed by the quantum annealing algorithm. According to the sparse characteristics of the data, by designing a suitable measurement matrix, the high-dimensional data is projected into a low-dimensional space to reduce the amount of data and optimize data transmission and storage efficiency.

3. The flight attitude control method based on inertial navigation technology according to claim 1, characterized in that: The steps of the distributed multi-source collaborative inertial navigation are performed by a distributed multi-source collaborative inertial navigation module. While acquiring the motion state of the aircraft, the distributed multi-source collaborative inertial navigation module fuses the visual camera to acquire the surrounding visual image information, and acquires thermal infrared images at night and in low visibility environments through the infrared camera for identifying landmarks and obstacles. It is equipped with an air pressure sensor, and obtains the aircraft altitude information in combination with the air pressure data. The data fusion algorithm that combines the D-S evidence theory with the particle swarm optimization algorithm is used to process the multi-source sensor data, dynamically adjust the sensor data weight, and realize the all-round and accurate perception of the aircraft attitude. The fused high-precision attitude data is transmitted to the real-time error compensation module based on deep learning, the multi-modal fusion and virtual twin collaborative control module, and the distributed attitude data management module based on blockchain. The steps of the data fusion algorithm that combines the DS evidence theory with the particle swarm optimization algorithm are as follows: Sp301. Use DS evidence theory to preliminarily fuse the data collected by the inertial navigation unit, optical camera, infrared camera, and air pressure sensor, calculate the basic probability distribution function of each data source, and evaluate the reliability of the data. Use DS evidence theory to first determine the basic probability distribution function of each data source based on the characteristics and historical data of each sensor. For the inertial navigation unit, assign probability based on its measurement accuracy and stability, evaluate the reliability of the data, and then synthesize the evidence of the same target aircraft attitude from different sensors to obtain the preliminary results after fusion; Sp302. The particle swarm optimization algorithm is used to dynamically adjust the data weights after fusion of the DS evidence theory. The position of the particles represents the weight distribution scheme. Each particle represents a possible weight combination. Through information sharing and iterative search among particles, the degree of match between the fused data and the actual flight status is used as the fitness function to find the optimal weight combination and improve the accuracy of the fused data.

4. The flight attitude control method based on inertial navigation technology according to claim 1, characterized in that: The real-time error compensation step based on deep learning is performed by a real-time error compensation module based on deep learning. When the echo state network algorithm is combined with the deep belief network algorithm to train the fusion data, data features are automatically extracted, error laws are learned, and errors are predicted and compensated in real time during flight. The attitude data after error compensation is fed back to the multimodal fusion and virtual twin collaborative control module, and the compensation-related information is transmitted to the distributed attitude data management module based on blockchain for recording; The training steps of the echo state network algorithm combined with the deep belief network algorithm are as follows: Sp401, using the echo state network algorithm to build the initial network model, determine the network's reserve pool size and connection weight parameters, take the inertial navigation raw data and environmental parameters as input, and use its sparse connection characteristics and dynamic reserve pool to perform preliminary feature extraction on the input data. The extracted features include data change trends, periodic features, and correlation features with environmental parameters; Sp402. Use the features extracted by the echo state network as the input of the deep belief network. Through layer-by-layer greedy training and reverse fine-tuning, further mine the data features, learn the error rules of the habitual navigation system, and realize real-time error compensation.

5. The flight attitude control method based on inertial navigation technology according to claim 1, characterized in that: The energy collection and self-powered inertial system module performs energy collection and self-powered inertial processing steps. The piezoelectric material energy collector on the wing surface generates electrical energy using wing vibration. The thermoelectric power generation material on the fuselage skin generates electrical energy based on the temperature difference between the inside and outside of the fuselage. The engine air intake converts the kinetic energy of the airflow into electrical energy through the principle of electromagnetic induction. The collected electrical energy is processed by a rectifier and a voltage stabilizing circuit and then stored in an energy storage element, which then powers the inertial navigation system through the voltage stabilizing circuit. The firefly algorithm combined with the adaptive dynamic programming algorithm will optimize the energy conversion and storage strategy of the energy collection device, improve the energy conversion efficiency, continuously provide stable power supply to each power module, and ensure normal operation. The optimization steps of the algorithm are as follows: Sp501. Use the firefly algorithm to optimize the working parameters of the energy collection device, where the energy collection device includes a piezoelectric vibration energy collector, a temperature difference power generation module, and an electromagnetic induction airflow energy collector. The optimized working parameters mainly include the resonant frequency of the piezoelectric vibration energy collector, the material ratio of the temperature difference power generation module, and the number of coil turns of the electromagnetic induction airflow energy collector. Simulate the luminescence and movement behavior of fireflies, take energy conversion efficiency as the objective function, find the optimal parameter combination for energy collection, and improve energy conversion efficiency. Sp502. Adopt an adaptive dynamic programming algorithm to dynamically adjust the charging and discharging strategies of energy storage elements according to energy collection and consumption to adapt to the energy requirements under different flight conditions such as take-off, cruising, landing and maneuvering flight. During the take-off phase, the energy demand is large, which is mainly used to start the engine and increase the altitude of the aircraft. The energy storage elements need to discharge quickly to meet the high power demand. During the cruising phase, the energy demand is relatively stable. The energy collection device charges the energy storage elements while meeting the current power consumption. During the landing phase, the energy demand is reduced, but the power supply of key equipment must be guaranteed. During maneuvering flight, the energy demand fluctuates greatly, and the charging and discharging strategies are adjusted according to the intensity of the maneuvering action.

6. The flight attitude control method based on inertial navigation technology according to claim 1, characterized in that: The distributed posture data management module based on blockchain performs posture data management steps, adopts distributed ledger technology to achieve data security and efficient transmission, and uses homomorphic encryption algorithm combined with zero-knowledge proof algorithm to achieve efficient data verification and sharing while ensuring data security. Each module can obtain authenticated and reliable data from the blockchain according to needs. The implementation steps of the homomorphic encryption algorithm combined with the zero-knowledge proof algorithm are as follows: Sp601, use homomorphic encryption algorithm to encrypt posture data, calculate and process data in ciphertext state, and ensure the security of data during transmission and storage; Sp602 uses zero-knowledge proof algorithm to prove the authenticity and integrity of data to other sensor nodes on the aircraft and ground control center nodes without leaking the specific content of the data, thereby achieving efficient data verification and sharing.

7. The flight attitude control method based on inertial navigation technology according to claim 1, characterized in that: The multimodal fusion and virtual twin collaborative control module executes the multimodal fusion and virtual twin collaborative control steps, inputs the multimodal data features into the virtual model, and uses the dynamic equations and flight mechanics principles in the model to build a real-time virtual twin model to simulate the real-time flight state of the aircraft, and compare it with the preset normal flight attitude range. When it exceeds the range, it is predicted to be an abnormal attitude. After analyzing the cause of the abnormality, according to the preset control strategy library, combined with fuzzy logic reasoning, a feedback optimization strategy is generated to adjust the aircraft control parameters, realize attitude adjustment, and transmit the attitude adjustment instruction to the actuator of the aircraft. At the same time, the attitude data record is fed back to other modules. The processing steps of the variational mode decomposition algorithm combined with the fuzzy neural network algorithm are as follows: Sp701, use variational modal decomposition algorithm to decompose multimodal data such as inertial navigation data, lidar point cloud data, millimeter wave radar data and visual image data. For inertial navigation data, decompose it into high-frequency dynamic change mode and low-frequency trend mode according to the frequency characteristics of the signal. Lidar point cloud data is decomposed into near-field target mode, mid-field environment mode and far-field background mode according to spatial distribution characteristics. Millimeter wave radar data is decomposed into stationary target mode and moving target mode according to target characteristics. Visual image data is decomposed into edge mode, texture mode and target mode according to texture and shape characteristics. Feature information of each mode is extracted, including rapid posture change characteristics in the high-frequency dynamic change mode of inertial navigation, and target distance and shape characteristics in the near-field target mode of lidar. Sp702. Input the extracted feature information into the fuzzy neural network. The fuzzy logic fuzzifies the feature information by setting fuzzy rules and converts it into fuzzy language variables. The neural network uses its powerful learning ability to further process and fuse the fuzzified features. According to the degree of correlation between the input features and the posture abnormality, it determines whether there is a posture abnormality and makes a prediction, thereby improving the virtual twin model's prediction accuracy for aircraft posture abnormalities.

8. The flight attitude control method based on inertial navigation technology according to claim 1, characterized in that: The terahertz communication and inertial navigation fusion anti-interference module performs the terahertz communication and inertial navigation fusion anti-interference steps, adopts a low-density parity check code algorithm combined with a fractional-order Fourier transform algorithm, improves the anti-interference ability and data transmission rate of the terahertz communication link, and ensures the accuracy and timeliness of data interaction between modules. The application steps of the low-density parity check code algorithm combined with the fractional-order Fourier transform algorithm are as follows: Sp801, using a low-density parity check code algorithm to encode the transmitted posture data, increase the redundant information of the data, and improve the error correction capability of the data during transmission. During encoding, the original posture data is grouped according to the set check matrix, and redundant check bits are added to each group of data, so that the receiving end can detect and correct the error bits that occur during the transmission process according to the check bits, increase the redundant information of the data, and improve the error correction capability of the data during transmission; Sp802. Use the fractional Fourier transform algorithm to transform the terahertz communication signal. Select the fractional order according to the channel characteristics such as multipath fading and noise interference, transform the signal from the time domain to the fractional Fourier domain, adjust the time-frequency distribution of the signal, and enhance the signal's anti-interference ability and transmission rate.

9. The flight attitude control method based on inertial navigation technology according to claim 1, characterized in that: The data interaction and collaborative control mechanism between the modules of the system are as follows: Sp901. By building a dedicated data transmission channel on the aircraft's avionics system bus, using a high-speed serial communication protocol, and establishing a unified data bus, each module transmits and interacts data through the data bus to ensure data consistency and timeliness; Sp902, formulate a data interaction protocol, clarify the responsibilities and processes of each module in the process of data sending, receiving and processing, and ensure the stable operation of the system. The data interaction protocol includes data format definition, transmission rate setting and data verification rules. In the process of data sending, the sending module is responsible for packaging the data according to the protocol format, adding the check bit and sending it. The receiving module is responsible for receiving the data and verifying the data integrity. If the verification passes, the data is parsed. The processing module performs corresponding processing according to the parsed data. The processing steps include attitude solution and error compensation to ensure the stable operation of the system; Sp903. Assign task priorities according to the urgency and importance of each module's tasks. When the aircraft encounters an emergency, the inertial navigation and attitude control module tasks have the highest priority, and system computing resources and data transmission bandwidth are allocated first. During normal flight, data recording and non-critical sensor data processing modules have lower priority. Through the task scheduler, CPU computing time and memory resources and other system resources are dynamically allocated according to priority to achieve efficient collaborative control.

Images