Small aircraft multi-sensor fusion navigation and positioning system

Abstract

The invention discloses a multi-sensor fusion navigation and positioning system for a small aircraft, and relates to the technical field of aircraft navigation, guidance and control. The system comprises an environment feature extraction module, a situation confidence generation module, a noise covariance matrix adjustment module and a navigation state fusion module. The system aims to improve the precision and environmental adaptability of the navigation system; by extracting multi-dimensional environment characteristics such as global navigation satellite system signal quality and inertia measurement unit dynamics in real time, the system can accurately perceive dynamic changes of a flight environment, core parameters of a fusion filter are adjusted according to the dynamic changes, and a filter model is continuously matched with an actual physical environment, so that the final precision of navigation and positioning is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aircraft navigation, guidance and control, in particular to a small aircraft multi-sensor fusion navigation and positioning system. BACKGROUND

[0002] In the field of small aircraft navigation technology, multi-sensor fusion systems, especially those based on Kalman filter, integrate data from global navigation satellite system (GNSS) and inertial measurement unit (IMU) to achieve optimal estimation of aircraft state; In these fusion systems, their performance is highly dependent on the accurate description of system model and sensor noise characteristics, which is usually done through pre-set process noise covariance matrix Q and observation noise covariance matrix R; existing technical solutions generally set these two core parameter matrices as fixed values, which can only reflect the system uncertainty under specific or ideal flight environment through offline calibration; Such static parameter configuration has inherent defects; when the actual environment of the aircraft changes dynamically, such as entering a city canyon with severely attenuated GNSS signals from an open airspace with unobstructed GNSS signals, or the aircraft's maneuverability increases, causing the quality of IMU data to decrease, the fixed noise matrix can no longer accurately represent the real noise statistical characteristics under the current environment; this model mismatch causes the filter to improperly allocate the degree of trust for each sensor, resulting in a decline in the performance of the fusion solution; Therefore, how to enable the core parameters of the navigation filter to adapt to the dynamically changing flight environment in real time and autonomously, so as to continuously maintain the accuracy, continuity and reliability of the navigation solution, has become a technical problem to be solved. SUMMARY

[0003] To solve the above technical problems, the present application discloses a small aircraft multi-sensor fusion navigation and positioning system, in particular, the technical solution of the present application comprises: An environment feature extraction module for processing sensor data to generate an environment feature vector; A context confidence generation module for generating a context confidence vector based on the environment feature vector and pre-set inference rules; A noise covariance matrix adjustment module for combining the context confidence vector and a pre-set reference noise covariance matrix to generate an adaptive observation noise covariance matrix and an adaptive process noise covariance matrix; A navigation state fusion module for performing fusion filtering on the navigation state based on the adaptive observation noise covariance matrix and the adaptive process noise covariance matrix.

[0004] Preferably, the environment feature extraction module extracts GNSS signal quality features, IMU data dynamic features and multi-source data consistency features, and combines the features to generate the environment feature vector.

[0005] Preferably, the context confidence generation module is specifically used for: generating a fuzzy language variable based on the environment feature vector and a preset membership function; generating an original activation score for each predefined environment context according to the fuzzy language variable and a preset rule base; normalizing the original activation score to generate the context confidence vector.

[0006] Preferably, the normalization of the original activation score comprises: processing the original activation scores of all contexts by using a Softmax function to generate the context confidence vector.

[0007] Preferably, the noise covariance matrix adjustment module generates the adaptive observation noise covariance matrix, comprising: weighting and summing a preset reference observation noise covariance matrix by taking the confidence of each predefined environment context as a weight to obtain the adaptive observation noise covariance matrix.

[0008] Preferably, the noise covariance matrix adjustment module generates the adaptive process noise covariance matrix, comprising: adjusting a preset additional process noise covariance matrix according to the confidence of a specific environment context; combining the adjusted additional process noise covariance matrix with a preset reference process noise covariance matrix to obtain the adaptive process noise covariance matrix.

[0009] Preferably, the navigation state fusion module is specifically used for: using an extended Kalman filter; wherein the adaptive observation noise covariance matrix is used when calculating the Kalman gain, and the adaptive process noise covariance matrix is used when updating the state prediction covariance.

[0010] Preferably, the reference process noise covariance matrix and the additional process noise covariance matrix are obtained based on offline calibration in a simulation environment of a specific context.

[0011] Compared with the prior art, the present application has the following beneficial effects: 1.The application improves the precision and environmental adaptability of the navigation system; by extracting real-time global navigation satellite system signal quality, inertial measurement unit dynamics, and multi-source data consistency, etc. as multi-dimensional environmental characteristics, the dynamic changes of the flight environment can be accurately perceived, and the core parameters of the fusion filter can be adjusted accordingly, so that the filter model continuously matches the actual physical environment, thereby significantly improving the final accuracy of navigation and positioning; 2.The application ensures the continuity and smoothness of the navigation solution; the system uses fuzzy logic reasoning and normalization processing to convert discrete environmental characteristics into continuous situation confidence descriptions; when adjusting the observation noise parameters, the reference model of different situations is smoothed and weighted to ensure that the core filter parameters can change continuously and smoothly in the environmental transition zone, effectively avoiding navigation solution oscillation or jumping caused by parameter mutation, and ensuring the continuity of the flight state output; 3.The application enhances the robustness and reliability of the system in harsh environments; the system identifies sensor abnormalities through multi-source data consistency characteristics, and accurately compensates for specific negative situations such as high vibration through targeted incremental adjustment of process noise; this mechanism enables the system to reduce its reliance on unreliable information when encountering signal interference, sensor performance degradation, and other adverse situations, effectively suppressing error divergence, thereby maintaining the reliability of navigation performance in complex environments; 4.The application realizes the unity of navigation performance optimality and robustness; unlike traditional solutions that use a single conservative parameter, the system uses a basic model with targeted incremental adjustment; this strategy ensures that the model's uncertainty is only increased when specific negative situations are detected, and the optimal parameter configuration is maintained in normal flight environments; this design avoids sacrificing system precision in regular conditions to cope with occasional harsh conditions, achieving optimal and robust navigation system performance in different conditions. BRIEF DESCRIPTION OF DRAWINGS

[0012] The application will be further explained in conjunction with the accompanying drawings and examples: Figure 1 is a structural diagram of the system of the application. DETAILED DESCRIPTION

[0013] To make the purpose, technical scheme and advantages of the application clearer, the application will be further described in detail below in conjunction with specific examples.

[0014] Example 1: Please refer to Figure 1 A small aircraft multi-sensor fusion navigation and positioning system, comprising: An environmental feature extraction module for processing sensor data to generate an environmental feature vector; The context confidence generation module is used to generate a context confidence vector based on the environmental feature vector and preset inference rules. The noise covariance matrix adjustment module is used to combine the context confidence vector with the preset baseline noise covariance matrix to generate an adaptive observation noise covariance matrix and an adaptive process noise covariance matrix. The navigation state fusion module is used to perform fusion filtering on the navigation state based on the adaptive observation noise covariance matrix and the adaptive process noise covariance matrix.

[0015] This invention provides a multi-sensor fusion navigation and positioning system for small aircraft. The system aims to solve the technical problem that the fixed navigation filter parameters in the prior art cannot adapt to dynamic flight environments. By sensing environmental changes in real time and dynamically adjusting the core parameters of the filter, the system can achieve adaptive optimization of navigation performance in changing flight environments. Its overall architecture includes an environmental feature extraction module, a context confidence generation module, a noise covariance matrix adjustment module, and a navigation state fusion module, forming a complete and self-consistent technical closed loop. The environmental feature extraction module aims to process raw data from multiple sensors to generate an environmental feature vector with clear physical meaning that can quantitatively characterize the external environment and the sensor's own operating state. In this embodiment, the module acquires real-time data streams from airborne sensors such as GNSS receivers and IMUs. By processing this data stream with specific algorithms, it extracts key indicators reflecting GNSS signal quality, aircraft dynamic characteristics, and consistency among multiple data sources. These indicators are then combined into a multi-dimensional environmental feature vector, providing direct and quantitative input for subsequent environmental context cognition. The purpose of the scenario confidence generation module is to generate a scenario confidence vector describing the probability that the current aircraft is in various predefined environmental scenarios based on the aforementioned environmental feature vector and through a lightweight inference model. In this embodiment, the module internally presets multiple environmental scenarios covering typical flight scenarios, such as open airspace, urban canyons, high-vibration environments, or GNSS deception interference. The module receives the feature vector output by the environmental feature extraction module and calculates it according to the preset inference rules, outputting a confidence score for each predefined environmental scenario. The confidence scores of all scenarios together constitute a vector, which strictly satisfies the characteristics of a probability distribution, thereby achieving a continuous and quantitative description of complex flight environments. The noise covariance matrix adjustment module aims to transform the aforementioned situation confidence vector into core parameters that the navigation filter can directly use, namely, the adaptively adjusted process noise covariance matrix and observation noise covariance matrix. As an intermediary link connecting environmental perception and filtering control, in this embodiment, after receiving the situation confidence vector, this module, combined with a preset set of benchmark noise covariance matrices, generates adaptive observation noise covariance matrices and adaptive process noise covariance matrices respectively through a specific weighted fusion and incremental adjustment algorithm. These generated matrices are no longer fixed but dynamically adjusted according to the real-time changes in the situation confidence vector, accurately reflecting the uncertainty levels of the sensor measurements and the system state transition model under the current environment. The navigation state fusion module aims to optimally estimate the aircraft's navigation state, such as position, velocity, and attitude, based on the aforementioned adaptively adjusted noise covariance matrix. In this embodiment, the module employs a recursive fusion filtering algorithm. In the core calculation loop of the filter, it no longer uses fixed, pre-set noise parameters, but directly uses the adaptive observation noise covariance matrix and the adaptive process noise covariance matrix generated in real time by the noise covariance matrix adjustment module. By deeply integrating the results of environmental perception into the mathematical core of the filter, the behavior of the entire navigation system can adapt to changes in the external environment in real time and autonomously. The system defined in this embodiment constructs a complete information processing link from environmental perception to model inference and parameter adaptation through the collaborative work of the above four modules. The system solves the inherent defect of traditional navigation filters that cannot adapt to dynamic environments due to the use of static parameters. It enables the system to autonomously, smoothly and optimally adjust its fusion strategy in complex and ever-changing situations, such as when GNSS signals change from strong to weak or when aircraft maneuverability changes from weak to strong, thereby significantly improving the accuracy, continuity and reliability of navigation and positioning.

[0016] Example 2: The environmental feature extraction module extracts GNSS signal quality features, IMU data dynamic features, and multi-source data consistency features, and combines these features to generate an environmental feature vector.

[0017] This embodiment further defines the specific implementation of the environmental feature extraction module based on the system in Embodiment 1; the feature vector extracted by this module is not an arbitrary combination of parameters, but is designed closely around the key factors affecting the performance of navigation filtering. To further clarify, the environmental feature extraction module extracts GNSS signal quality features, IMU data dynamic features, and multi-source data consistency features, and combines these features to generate an environmental feature vector; GNSS signal quality characteristics refer to quantitative indicators that directly reflect the received strength and quality of GNSS signals. Their function is to determine whether the GNSS signal is obstructed, interfered with, or in an environment with severe multipath effects. In this embodiment, this characteristic can specifically be the average carrier-to-noise ratio (C / N0) of all available GNSS satellite signals, denoted as... The source of this feature is the raw observations output in real time by the airborne GNSS receiver; IMU data dynamics refers to an indicator that quantifies the maneuver intensity of an aircraft and the vibration level of its environment. Its function is to assess the degree to which the IMU's performance may degrade due to high-frequency vibration or severe maneuvers. In this embodiment, this characteristic can be achieved by performing a Short-Time Fourier Transform (STFT) on the output data of the IMU gyroscope or accelerometer and calculating its spectral entropy, denoted as... This feature is obtained by performing signal processing operations on the raw output data of the IMU. Multi-source data consistency characteristics refer to an index that measures the degree of consistency between information from different sensor sources. Its function is to detect anomalies such as sensor malfunctions or spoofing attacks. In this embodiment, this characteristic can specifically be the normalized residual norm between GNSS measurements and predicted measurements converted from IMU state predictions to the observation space, such as Mahalanobis distance, denoted as... When GNSS signals are subjected to spoofing attacks, their output will deviate significantly and continuously from the IMU's autonomous calculations, causing an abnormal increase in the residual norm. This characteristic originates from the calculation of intermediate variables within the navigation state fusion module. The aforementioned characteristic parameters collectively constitute the time-varying parameters. Environmental feature vector:

[0018] By extracting and combining these three complementary features with clear physical meaning, the system can more comprehensively and accurately characterize the complex environment in which the aircraft is located. Compared with a single feature, this feature combination can distinguish more ambiguous situations, such as whether the GNSS signal is weak or the IMU vibration is large, providing a solid data foundation for the accurate generation of subsequent situation confidence, thereby improving the depth and breadth of the entire system's environmental perception.

[0019] Example 3: The context confidence generation module is specifically used for: Based on environmental feature vectors and preset membership functions, fuzzy linguistic variables are generated. Based on fuzzy linguistic variables and a pre-defined rule base, an initial activation score is generated for each predefined environmental context. The original activation scores are normalized to generate a contextual confidence vector.

[0020] The original activation scores are normalized, including: The Softmax function is used to process the raw activation scores of all scenarios to generate scenario confidence vectors.

[0021] This embodiment further defines the specific implementation of the context confidence generation module based on the system in Embodiment 1. To adapt to the airborne resource limitations of small aircraft, this embodiment uses a lightweight and efficient fuzzy logic reasoning system to generate the context confidence vector. The workflow of the context confidence generation module is described below: The module generates fuzzy linguistic variables based on environmental feature vectors and preset membership functions; the membership function refers to taking a precise real number as input, such as a feature value. This is a function that maps to one or more fuzzy sets, such as high, medium, and low, and its value is between 0 and 1. Its function is to transform quantified feature data into fuzzy language descriptions that conform to logical reasoning habits. The specific forms of these functions, such as triangles, trapezoids, and Gaussians, are set offline based on expert experience and statistical analysis of a large amount of flight data. Specifically, the distribution range of characteristic values ​​can be determined by performing histogram statistics on a large amount of actual or simulated flight data; for example, statistical analysis can reveal... The signal quality range is mainly concentrated in [35,40] dB-Hz, and the transition range is approximately 5 dB-Hz. Therefore, the core range of the top edge of the trapezoidal membership function is set to [35,40], and the bottom edge is extended to both sides by 5 dB-Hz, thus determining the specific parameters of the function. For example, for input features The membership degree of a fuzzy set at high, medium, and low levels can be obtained through the membership function, such as... The membership degree is 0.8 for high and 0.2 for medium. To further illustrate, for the feature... The membership degree of this fuzzy set can be calculated using the trapezoidal membership function:

[0022] in, Represents input Value (unit: dB-Hz) It corresponds to its membership degree.

[0023] The module generates raw activation scores for each predefined environmental context based on fuzzy linguistic variables and a pre-defined rule base. The rule base refers to a set of IF-THEN logical rules predefined by domain experts, simulating the decision-making process of experts facing specific environmental characteristics. For example, a rule could be: IF ( is LOW AND ( (is LOW) THEN (Context is) ), i.e., urban canyons; another more complex rule could be: IF ( is LOW AND ( is HIGH) AND ( is LOW) THEN (Context is C3); IF ( The `isHIGH)THEN(Context is C4)` module inputs the fuzzified feature variables into the rule base, and then uses fuzzy inference, such as the Mamdani or Sugeno methods, to perform fuzzy reasoning for each predefined context. Calculate an original activation score ; The module normalizes the original activation scores to generate a context confidence vector. To ensure that the output context confidences have the characteristics of a probability distribution (i.e., the sum of all confidences is 1) for subsequent weighted calculations, the original activation score vector needs to be normalized.

[0024] In this embodiment, normalization is performed; the Softmax function is used to process the raw activation scores of all scenarios to generate the final scenario confidence vector. To dynamically generate context confidence vectors The Softmax normalization function used in this embodiment is calculated as follows:

[0025] in, Indicates the current environmental characteristics Under these circumstances, the system determines that the aircraft is in a certain situation. The final confidence score is a dimensionless floating-point number, ranging from 0 to 1, and is calculated using this formula. For the corresponding scenario At any moment The original activation score is a dimensionless floating-point number, calculated by the preceding fuzzy logic reasoning steps; The total number of predefined environmental scenarios, a positive integer, is a system preset parameter; The temperature coefficient is a positive adjustable parameter, a positive floating-point number, and a system-preset hyperparameter. Its value is determined offline based on the expected response characteristics of the system during scenario switching. Using the Softmax function instead of simple linear normalization amplifies the differences between activation scores, making the confidence level more biased towards the highest-scoring scenario while preserving awareness of other possible scenarios; temperature coefficient The introduction of this provides a flexible means to adjust the smoothness of the output probability distribution; This formula is in the final stage of the context confidence generation module, and its input is the original activation score vector obtained from fuzzy inference. Its output is a normalized context confidence vector.

[0026] The information is then passed to the next noise covariance matrix adjustment module. By introducing the Softmax function for normalization, this embodiment ensures that the output situation confidence vector not only strictly satisfies the probability axiom, but also can adjust the decisiveness of its decision as needed. This allows the subsequent noise matrix adjustment to be carried out on an input with good mathematical properties and clear physical meaning, thereby greatly improving the smoothness and robustness of the entire adaptive adjustment mechanism. By combining the above-mentioned fuzzy inference and Softmax normalization, this system can achieve the transformation from continuously changing, precise sensor data to continuously changing, probabilistic situation descriptions with low computational cost. Compared with the traditional hard threshold-based switching scheme, this soft switching mechanism can avoid abrupt changes and oscillations in navigation solutions at situation boundaries, ensuring a smooth transition in navigation performance.

[0027] Example 4: The noise covariance matrix adjustment module generates an adaptive observation noise covariance matrix, including: The confidence level of each predefined environmental scenario is used as a weight to perform a weighted summation on the preset benchmark observation noise covariance matrix to obtain the adaptive observation noise covariance matrix.

[0028] This embodiment, based on the system in Embodiment 1, focuses on how the noise covariance matrix adjustment module generates an adaptive observation noise covariance matrix. The specific implementation method has been further specified; The method for generating the adaptive observation noise covariance matrix in this module is to use the confidence level of each predefined environmental scenario as a weight to perform a weighted summation on the preset benchmark observation noise covariance matrix. To dynamically calculate the adaptive observation noise covariance matrix based on contextual confidence. This embodiment uses an adaptive calculation formula based on model weighting, and its calculation method is as follows:

[0029] in, For a moment The adaptive observation noise covariance matrix is ​​the final output of this calculation, which directly reflects the degree of uncertainty of the measurement values ​​of each sensor in the current environment. It is the covariance matrix. For example, the variance of the position observation is in meters², and it is calculated by this formula. The context confidence generation module outputs the aircraft's time at time... In the context The confidence level, which is used here as a weighting coefficient, is a dimensionless floating-point number with a value range of 0 to 1, and is calculated from the previous steps; For each predefined environment scenario The corresponding benchmark observation noise covariance matrix represents the noise level under ideal, pure conditions. In this scenario, the statistical characteristic of sensor observation noise is the covariance matrix, which is a preset parameter matrix. Its value is obtained by accurately simulating the scenario. The results are obtained by conducting numerous repeated experiments in a real or semi-physical simulation environment, collecting sensor data, and performing offline statistical calibration; for example... The variance value corresponding to GNSS positions in open airspace is relatively small, while This value increases significantly in urban canyons; to enable those skilled in the art to implement this method, a method for obtaining the benchmark observation noise covariance matrix is ​​provided. Example of specific calibration method: Step 1: Build a hardware-in-the-loop simulation environment that includes a high-rise building model. This environment can simulate the multipath effect and blockage of GNSS signals. Step 2: Plan 100 representative flight paths in this environment and simulate them, while recording the actual position of the aircraft output by the simulation system and the position output by the simulated GNSS receiver; Step 3: Calculate the error sequence between the GNSS position output and the actual position in each simulation; Step 4: Perform statistical analysis on all error sequences obtained in Step 3 and calculate their covariance matrix; this covariance matrix is ​​then labeled as... The baseline matrix for other scenarios can be obtained using a similar method. The total number of predefined environmental scenarios, a positive integer, is a system preset parameter; Real-world flight environments are often not singular or pure, but rather a mixture or transitional state of multiple scenarios. By using real-time scenario confidence as weights, a weighted average of the baseline noise models representing various pure scenarios can be generated to produce an equivalent noise model that best matches the current mixed environment. This formula is executed in the noise covariance matrix adjustment module, and its input is the context confidence vector. The output is The matrix is ​​then passed to the navigation state fusion module to calculate the Kalman gain. The essence of this calculation is based on all preset benchmark noise models. In the constructed model space, dynamic interpolation is performed based on real-time context confidence; when the environment smoothly transitions from one context to another, Continuous changes occur, thus It also changes smoothly and continuously; for example, when an airplane flies from an open area. Approaching 1, it flew into the city canyon. Gradually increase It will smoothly from Transition to This allows the navigation filter to gradually reduce its reliance on GNSS observations, avoiding navigation oscillations caused by parameter mutations and ensuring the smoothness and reliability of navigation. By using this weighted summation method, the system can generate an observation noise covariance matrix that precisely matches the current complex environment, enabling refined and dynamic management of the trust level of sensor measurements. This greatly improves the adaptability and robustness of the fusion filter in areas of environmental change.

[0030] Example 5: The noise covariance matrix adjustment module generates an adaptive process noise covariance matrix, including: The preset additional process noise covariance matrix is ​​adjusted based on the confidence level of a specific environmental situation. The adjusted additional process noise covariance matrix is ​​combined with the preset baseline process noise covariance matrix to obtain the adaptive process noise covariance matrix.

[0031] The baseline process noise covariance matrix and the additional process noise covariance matrix are obtained through offline calibration in a simulation environment under specific conditions.

[0032] How does the noise covariance matrix adjustment module generate an adaptive process noise covariance matrix? The specific implementation method and the source of the matrices used were further specified; The method for generating an adaptive process noise covariance matrix in this module is to add additional process noise with a specific target based on the confidence level of a particular environmental context, on top of a universal baseline process noise covariance matrix. The method includes: adjusting a preset additional process noise covariance matrix according to the confidence level of the particular environmental context; and then combining the adjusted additional process noise covariance matrix with the preset baseline process noise covariance matrix. To achieve accurate response to specific negative situations, this embodiment employs an adaptive calculation formula based on incremental adjustment, the calculation method of which is as follows:

[0033] in, For a moment The adaptive process noise covariance matrix, which is the final output of this calculation, describes the uncertainty of the current aircraft state transition model. It is the covariance matrix, which is calculated by this formula. The reference process noise covariance matrix represents the uncertainty of the basic model under ideal, stable flight conditions. This matrix is ​​obtained through offline calibration in a specific scenario, such as a simulation environment of stable flight. Alternatively, it can be initially set based on the aircraft's dynamic model and the noise indicators specified in the IMU device manual, such as angle random walk and velocity random walk. It is a set of specific environmental scenarios that can significantly affect process noise, and is a library of all predefined scenarios. A subset, preset by the system, for example This refers to high-vibration environments, as high vibrations can significantly degrade IMU performance and increase its random drift. For a specific context in Belongs to set The real-time confidence level, which is used here as an adjustment weight, is a dimensionless floating-point number and is calculated by the context confidence generation module. For context The additional process noise covariance matrix describes the situation. When this occurs, the incremental part of the uncertainty in the system model is the covariance matrix; this matrix is ​​also based on offline calibration in a simulation environment under specific circumstances, such as a vibration table test environment simulating high vibration. For the corresponding scenario The dimensionless adjustment factor is a positive floating-point number and is a preset hyperparameter of the system. It is optimized and tuned through system-level simulation to achieve the best system response characteristics, such as minimizing the positioning drift rate in degraded mode. Process noise is mainly affected by internal factors such as the aircraft's own maneuvering and IMU performance degradation; these influencing factors are usually superimposed on the basic noise; therefore, by adopting a strategy of basic model plus target incremental adjustment, it is possible to accurately compensate for known and specific negative impact sources without interfering with the basic model. This formula is also executed in the noise covariance matrix adjustment module, with the input being the context confidence vector and the output being... The matrix is ​​then passed to the navigation state fusion module to update the state prediction covariance; the calculation logic is that the system always maintains a basic process noise. Only when a specific situation that would worsen the process model is detected. At that time, that is Only when the noise level is significantly greater than zero should the corresponding additional noise be injected as needed and proportionally. For example, when an aircraft enters a high-vibration environment... rise, The corresponding increase allows the filter to promptly increase its awareness of the uncertainty in its own state prediction, thereby treating the IMU's calculation results more cautiously in the state update step and effectively suppressing the rapid accumulation of navigation errors caused by IMU performance degradation. Through this precise and targeted incremental adjustment mechanism, the system effectively adjusts specific interference sources, avoiding the use of a conservative and overly large Q matrix in all situations. This approach not only effectively suppresses error divergence in harsh environments but also maintains the optimality of the navigation solution in normal environments, achieving a balance between robustness and optimality under different operating conditions. At the same time, it clarifies that the reference and additional process noise covariance matrices are obtained through offline calibration, ensuring the objectivity and feasibility of these key parameters.

[0034] Example 6: The navigation state fusion module is specifically used for: An extended Kalman filter is used; Specifically, an adaptive observation noise covariance matrix is ​​used when calculating the Kalman gain, and an adaptive process noise covariance matrix is ​​used when updating the state prediction covariance.

[0035] This embodiment further defines the specific implementation of the navigation state fusion module based on the system in Embodiment 1, and clarifies the filter type used and the application method of the adaptive noise covariance matrix. The navigation state fusion module is specifically used to employ an extended Kalman filter (EKF). The EKF is a standard algorithm well-known to those skilled in the art for handling the state estimation problem of nonlinear systems. Its basic structure includes two steps: state prediction and state update, which will not be elaborated here. The innovation of this invention lies in the adaptive improvement made to two key steps during the execution of the EKF algorithm: an adaptive observation noise covariance matrix is ​​used when calculating the Kalman gain; to further clarify, the Kalman gain is calculated in the state update step of the EKF algorithm. The formula uses the observation noise covariance matrix; in this embodiment, it will be generated in real time by the noise covariance matrix adjustment module. Substituting this into the formula replaces the traditional fixed... The matrix; since the Kalman gain determines the filter's trust weight for new observations, this allows the weight to be dynamically adjusted as the environment changes; An adaptive process noise covariance matrix is ​​used when updating the state prediction covariance; to further clarify, in the state prediction step of EKF, the state prediction covariance matrix is ​​updated. The formula uses the process noise covariance matrix; in this embodiment, it will be generated in real time by the noise covariance matrix adjustment module. Substituting this into the formula replaces the traditional fixed... Matrix; Since the state prediction covariance reflects the uncertainty of the model prediction, this allows this uncertainty assessment to reflect changes in aircraft dynamics and IMU performance in real time. By and By deeply integrating into the core mathematical loop of the extended Kalman filter, this system achieves seamless coupling between environmental perception capability and state estimation algorithm; this makes the EKF no longer a static actuator, but an intelligent estimator with cognitive capabilities; it can autonomously and optimally balance its dependence on its own prediction model and external sensor measurements according to real-time changes in the environment, thereby maintaining high accuracy and high robustness of navigation performance in various complex and even adversarial environments. To verify the beneficial effects of the present invention, a simulation comparison experiment was conducted. In a typical simulation test of entering an urban canyon from open airspace, the navigation scheme of the present invention reduced the root mean square error of the position by about 40% compared with the scheme using the traditional fixed parameter extended Kalman filter, and no obvious positioning jump occurred at the scenario switching boundary, which verified the significant advantages of the present invention in improving positioning accuracy and continuity.

[0036] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A small aircraft multi-sensor fusion navigation and positioning system, characterized in that, The method comprises the following steps: An environment feature extraction module is used to process sensor data to generate an environment feature vector; A context confidence generation module is used to generate a context confidence vector according to the environment feature vector and preset inference rules; A noise covariance matrix adjustment module is used to combine the context confidence vector and a preset reference noise covariance matrix to generate an adaptive observation noise covariance matrix and an adaptive process noise covariance matrix; A navigation state fusion module is used to perform fusion filtering on a navigation state according to the adaptive observation noise covariance matrix and the adaptive process noise covariance matrix.

2. The system of claim 1, wherein, The environment feature extraction module extracts GNSS signal quality features, IMU data dynamic features, and multi-source data consistency features, and combines the features to generate the environment feature vector.

3. The system of claim 1, wherein, The context confidence generation module is specifically used to: Generate fuzzy language variables based on the environment feature vector and a preset membership function; Generate original activation scores for each predefined environment context according to the fuzzy language variables and a preset rule base; Perform normalization processing on the original activation scores to generate the context confidence vector.

4. The system of claim 3, wherein, The normalization processing on the original activation scores comprises: Processing the original activation scores of all contexts using a Softmax function to generate the context confidence vector.

5. The system of claim 1, wherein, The noise covariance matrix adjustment module generates the adaptive observation noise covariance matrix, which comprises: Performing weighted summation on a preset reference observation noise covariance matrix by taking the confidence of each predefined environment context as a weight to obtain the adaptive observation noise covariance matrix.

6. The system of claim 1, wherein, The noise covariance matrix adjustment module generates the adaptive process noise covariance matrix, which comprises: Adjusting a preset additional process noise covariance matrix according to the confidence of a specific environment context; Combining the adjusted additional process noise covariance matrix with a preset reference process noise covariance matrix to obtain the adaptive process noise covariance matrix.

7. The system of claim 1, wherein, The navigation state fusion module is specifically used to: Use an extended Kalman filter; Wherein, the adaptive observation noise covariance matrix is used when calculating the Kalman gain, and the adaptive process noise covariance matrix is used when updating the state prediction covariance.

8. The system of claim 6, wherein, The reference process noise covariance matrix and the additional process noise covariance matrix are obtained based on offline calibration in a simulation environment of a specific context.

Images