Unmanned aerial vehicle redundancy sensor self-arbitration method and flight control board

Abstract

The application discloses a kind of redundancy sensor self-arbitration method and flight control board of unmanned plane.The flight control board is equipped with at least two inertial measurement units and at least three magnetic field sensors.The method first carries out cross consistency detection to inertial measurement data and excludes invalid unit in combination with the independent health evidence of each unit, forms credible inertial reference;Then according to the length deviation of each magnetic field vector relative to local geomagnetic field, the change deviation of magnetic heading change relative to short-term inertial heading change and the heading consistency between candidate magnetic field sensors, sequentially determine effective magnetic field sensor;Through credible, invalid and transition health state management sensor, make the sensor recovered through detection or arbitration participate in attitude and heading estimation again according to gradual change weight.Single-path effectiveness determination precedes inter-sensor consistency determination, can exclude disturbed magnetic field sensor consistent with each other but deviating from credible inertial reference, thereby improving the reliability of attitude and heading input under strong magnetic interference and sensor abnormal condition.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) flight control technology, specifically to a UAV redundancy sensor self-arbitration method and flight control board. Background Technology

[0002] Currently, attitude and heading estimation for drones typically relies on inertial measurement units (IMUs) and magnetic field sensors. IMUs output three-axis angular velocities and three-axis specific forces, supporting attitude recursion at high update rates; magnetic field sensors output three-axis magnetic field vectors, providing absolute or relative heading constraints after attitude compensation. In swarm performance scenarios with dozens or even more drones operating densely, abnormal heading of a single drone not only affects its own trajectory but may also disrupt formation spacing. Therefore, the flight control system needs to continuously assess the reliability of the data entering the estimator.

[0003] Inertial measurement units (IMUs) may generate abnormal data due to temperature drift, strong vibration, communication interruption, output saturation, register malfunction, or device failure. When using only a single IMU, the system lacks real-time reference. With two or more IMUs, simple averaging or fixed master-slave switching may still introduce outliers into the estimator, or cause attitude input jumps during switching. Especially with only two units, inconsistency between them does not directly indicate which unit has failed; attribution must be based on independent health data from each unit.

[0004] Magnetic field sensors are susceptible to interference from the drone's surrounding environment and its own electrical equipment, and the interference experienced can vary depending on the location and time. When a drone passes near steel beams, speaker magnets, or temporary cables, the magnitude and direction of a particular magnetic field vector may change simultaneously; changes in current from motors, ESCs, and power distribution wiring can also generate time-varying magnetic fields. While majority voting among multiple magnetic field sensors can identify isolated anomalies, when two adjacent external magnetic field sensors are simultaneously within the range of the same interference source, they may exhibit highly correlated and consistent erroneous changes. If sensor consistency is taken as a sufficient condition, this shared deviation will form an erroneous majority. On the other hand, if all magnetic field sensors are temporarily unavailable, directly continuing to update magnetic field measurements will inject contamination measurements into the heading status, while a direct hard recovery can easily cause abrupt changes in heading. Summary of the Invention

[0005] This invention aims to solve the technical problems of redundant inertial and magnetic field sensors having difficulty in determining reliable inputs under abnormal or common magnetic interference conditions, and the estimation discontinuity caused by sensor elimination and recovery processes. It provides a self-arbitration method and flight control board that uses a reliable inertial reference as a non-magnetic constraint, prioritizes single-path magnetic field validity detection over inter-channel consistency detection, and is combined with gradual recovery of health status.

[0006] To achieve the above objectives, this invention provides a self-arbitration method for redundant sensors in unmanned aerial vehicles (UAVs). The UAV flight control board is equipped with at least two inertial measurement units and at least three magnetic field sensors. The method includes: S1, perform cross-consistency detection on the inertial measurement data output by each inertial measurement unit. When the inertial measurement data of at least two inertial measurement units meet the preset continuous inconsistency condition, determine the invalid inertial measurement unit based on the independent health evidence of each inertial measurement unit and exclude its data. Form a reliable inertial reference based on the inertial measurement data that has passed the detection. S2, determine the short-term heading change reference based on the reliable inertial reference, and determine the candidate magnetic field sensor based on the magnitude deviation of the magnetic field vector output by each magnetic field sensor relative to the local geomagnetic field and the deviation between the magnetic heading change corresponding to each magnetic field sensor and the short-term heading change reference. S3. Determine the effective magnetic field sensor based on the number of candidate magnetic field sensors and the heading consistency among them; S4. Based on the cross-consistency detection and arbitration results of each magnetic field sensor, maintain the health status of each inertial measurement unit and each magnetic field sensor, input the reliable inertial reference and the valid magnetic field sensor data in the reliable state into the attitude and heading estimator, and exclude the sensor data in the invalid state; when an inertial measurement unit in the invalid state passes the cross-consistency detection again, or a magnetic field sensor in the invalid state passes the arbitration again, switch the sensor to the transition state, and let its data participate in the attitude and heading estimation again according to the gradual weight.

[0007] To optimize the above technical solution, the specific measures also include: Furthermore, the cross-consistency check in step S1 includes: Calculate the difference in angular velocity data and the difference in specific force data for the corresponding axes of different inertial measurement units; When the difference in the angular velocity data or the difference in the force data is continuous When the difference threshold is exceeded in a single detection cycle, the inertial measurement data is determined to be persistently inconsistent. A pre-defined positive integer; Based on the self-test status, communication status, output saturation status, output constant status, and noise anomaly status of each inertial measurement unit, independent health evidence is formed for each inertial measurement unit. When the inertial measurement data is consistently inconsistent and the independent health evidence of one inertial measurement unit is abnormal, that inertial measurement unit is identified as an invalid inertial measurement unit.

[0008] Further, in step S2, candidate magnetic field sensors are determined based on the magnitude deviation of the magnetic field vector output by each magnetic field sensor relative to the local geomagnetic field, including: The geomagnetic field strength is obtained based on the location information of the UAV, and the geomagnetic field strength is corrected based on the magnetic field calibration results of the UAV's location to obtain the geomagnetic field reference value. Calculate the magnitude deviation of each magnetic field vector relative to the geomagnetic field reference value, and retain magnetic field sensors whose magnitude deviation does not exceed the magnitude tolerance as candidate magnetic field sensors.

[0009] Further, in step S2, candidate magnetic field sensors are determined based on the deviation between the magnetic heading change corresponding to each magnetic field sensor and the short-term heading change reference, including: The yaw axis angular velocity is obtained from the detected inertial measurement data and integrated over a preset time period to obtain the short-term heading change reference. The magnetic heading change corresponding to each magnetic field sensor is determined based on the magnetic field vector output by each magnetic field sensor within a preset time period. Calculate the deviation between each magnetic heading change and the short-term heading change reference, and retain the magnetic field sensors whose deviation does not exceed the change tolerance as candidate magnetic field sensors.

[0010] Further, step S3 includes: when there are at least two candidate magnetic field sensors, calculating the pairwise difference between the heading angles corresponding to each candidate magnetic field sensor; When the difference between the heading angles corresponding to at least two candidate magnetic field sensors does not exceed the consistency tolerance, and the difference between the heading angle corresponding to another candidate magnetic field sensor and the heading angles corresponding to the at least two candidate magnetic field sensors exceeds the consistency tolerance, the at least two candidate magnetic field sensors are determined as valid magnetic field sensors. When only one candidate magnetic field sensor exists, the modulus deviation and variation deviation of that candidate magnetic field sensor are continuous. If each detection cycle does not exceed the corresponding tolerance, the candidate magnetic field sensor is determined to be a valid magnetic field sensor. A pre-defined positive integer; When multiple effective magnetic field sensors exist, the fusion weight is determined based on the health status, modulus deviation, and variation deviation of each effective magnetic field sensor, and the heading reference is formed based on the magnetic field vectors of the multiple effective magnetic field sensors.

[0011] Furthermore, the health status of each sensor in step S4 includes a trusted state, an invalid state, and a transitional state; When the sensor is in a trusted state, it is in continuous If the sensor fails the corresponding test or arbitration within a test cycle, the sensor will be changed from a trusted state to an invalid state. When the sensor is in an invalid state, in continuous When the sensor passes the corresponding test or arbitration within each test cycle, the sensor is changed from an invalid state to a transitional state. In the transition state, the sensor data is gradually incorporated into the attitude and heading estimation according to the gradual weight that increases over time, and the sensor is converted to a reliable state after the gradual weight reaches the target weight. in, and These are pre-defined positive integers.

[0012] Further, in step S4, the attitude and heading estimator is a state estimation filter, the reliable inertial reference is used for state prediction of the state estimation filter, and the effective magnetic field sensor data is used for the heading measurement update of the state estimation filter; When there are no valid magnetic field sensors in a reliable or transitional state, stop using magnetic field sensor data to update the heading vector measurement, maintain short-term heading estimation based on the reliable inertial reference, and increase the process noise corresponding to the heading estimation. When at least one magnetic field sensor passes arbitration again, it will re-participate in the navigation vector measurement update according to the gradient weight corresponding to that magnetic field sensor.

[0013] Furthermore, in step S4, when arbitrating each magnetic field sensor, the single-path validity of each magnetic field sensor is first determined based on the modulus deviation and variation deviation, and then the heading consistency of the candidate magnetic field sensors that have passed the single-path validity determination is determined. A magnetic field sensor that fails to pass the modulus deviation judgment or change deviation judgment shall not be determined as a valid magnetic field sensor because its heading is consistent with that of other magnetic field sensors. When at least two magnetic field sensors have the same heading, but the magnetic heading changes corresponding to the at least two magnetic field sensors are inconsistent with the short-term heading change reference, the at least two magnetic field sensors are excluded.

[0014] The present invention also provides a UAV flight control board with a redundant sensor self-arbitration function, comprising: a flight control circuit board; The main control processor and at least two inertial measurement units are mounted on the flight control circuit board. A magnetic field sensor group including an onboard magnetic field sensor and at least two external magnetic field sensors; and an external sensor interface connecting the external magnetic field sensors to the main control processor; The main control processor is used for: Cross-consistency checks are performed on the inertial measurement data. Invalid inertial measurement units are identified based on the independent health evidence of each inertial measurement unit. A reliable inertial reference is formed based on the inertial measurement data that pass the check. Based on the reliable inertial reference, a short-term heading change reference is determined. Based on the magnitude deviation of the magnetic field vector output by each magnetic field sensor relative to the local geomagnetic field and the deviation between the magnetic heading change corresponding to each magnetic field sensor and the short-term heading change reference, candidate magnetic field sensors are determined. Based on the number of candidate magnetic field sensors and the heading consistency among the candidate magnetic field sensors, effective magnetic field sensors are determined. The trusted inertial reference and the valid magnetic field sensor data in the trusted state are input into the attitude and heading estimator. Sensor data that fails the detection or arbitration are excluded, and the data is converted to a transition state and re-participated in attitude and heading estimation according to the gradual weight.

[0015] Furthermore, at least two of the external magnetic field sensors are respectively disposed on different external magnetic compass plates, each external magnetic compass plate including a circuit board cantilever extending in a direction away from the electromagnetic interference source of the UAV body, and the external magnetic field sensor is disposed at the far end of the corresponding circuit board cantilever. At least two of the external magnetic field sensors are connected to the main control processor through a bus connection structure with channel isolation addressing capability, so that each external magnetic field sensor can be addressed and read by the main control processor.

[0016] The beneficial effects of this invention are as follows: By combining cross-consistency detection of inertial measurement data with independent health evidence of each inertial measurement unit, invalid inertial measurement units can be identified more accurately, reducing the possibility of abnormal inertial data entering the attitude and heading estimator; by sequentially performing magnetic field modulus detection, comparison of magnetic heading change and short-term heading change reference, and heading consistency judgment among candidate magnetic field sensors, single-path magnetic interference and common magnetic interference experienced by multiple magnetic field sensors can be identified; through continuous detection counting, transition state, and gradual weighting, the sensors can switch smoothly during elimination and recovery processes; when all magnetic field sensors are temporarily unavailable, magnetic field measurement updates are stopped and a reliable inertial reference is used to maintain short-term heading, while increasing the corresponding process noise reflects the decrease in heading estimation reliability, preventing disturbed magnetic field data from continuing to affect heading estimation. Furthermore, placing external magnetic field sensors at the far end of the circuit board cantilever and addressing them separately helps reduce electromagnetic interference from the aircraft and provides independent data for the path-by-path detection and arbitration of each magnetic field sensor. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the self-arbitration method for redundant sensors in UAVs according to the present invention. Figure 2 This is a schematic diagram of the cross-consistency detection process of the inertial measurement unit of the present invention; Figure 3 This is a schematic diagram of the sequential arbitration process for the magnetic field sensor provided by the present invention; Figure 4 This is a schematic diagram of the sensor health status transfer process provided by the present invention; Figure 5 This is a schematic diagram of the common magnetic interference identification logic provided by the present invention; Figure 6 The present invention provides computer simulation heading curves for three types of magnetic interference scenarios. Detailed Implementation

[0018] The invention will now be described in further detail with reference to the accompanying drawings.

[0019] It should be noted that terms such as "upper," "lower," "left," "right," "front," and "back" used in this invention are merely for clarity of description and are not intended to limit the scope of the invention. Changes or adjustments to these relative relationships, without substantially altering the technical content, should also be considered within the scope of the invention. The chip model, sampling frequency, threshold, and duration provided in this invention are examples of how the invention can be implemented and are not intended to limit the scope of protection. Other processors, inertial measurement units, magnetic field sensors, or state estimators with the same functions can also execute the method of this invention and are within the scope of protection of this invention.

[0020] like Figure 1-6 As shown in the figure, this invention discloses a UAV flight control board with redundant sensor self-arbitration function, including a flight control circuit board, a main control processor, at least two inertial measurement units, a magnetic field sensor group consisting of an onboard magnetic field sensor and at least two external magnetic field sensors, and an external sensor interface. The external magnetic field sensors are respectively disposed at the far ends of the cantilever arms of the printed circuit boards of different magnetic compasses to reduce electromagnetic interference from the UAV body; each external magnetic field sensor is connected to the main control processor through an independent bus branch or different downstream channels of a bus multiplexer, enabling the main control processor to address and read each external magnetic field sensor separately, thereby providing mutually distinguishable magnetic field data for the path-by-path validity detection and arbitration of each magnetic field sensor.

[0021] In this embodiment, the main control processor is connected to the first inertial measurement unit, the second inertial measurement unit, and the first, second, and third magnetic field sensors. The first magnetic field sensor is mounted on the flight control circuit board, while the second and third magnetic field sensors are mounted on two external magnetic compass boards. The main control processor sequentially runs a data preprocessing module, an inertial consistency detection module, a single-path magnetic field detection module, a candidate consistency module, a health status management module, and an attitude and heading estimator. Each module can be implemented by a software task on the same processor, or it can be implemented by a division of labor between the processor and programmable logic.

[0022] The self-arbitration method for redundant sensors in unmanned aerial vehicles (UAVs) disclosed in this invention first performs cross-consistency checks on the inertial measurement data. When data inconsistency persists, the method not only relies on the differences between channels but also reads independent health evidence such as self-testing, communication, saturation, constant output, and noise anomalies of each unit. This identifies attributable invalid inertial measurement units and excludes their data. A reliable inertial reference is then formed using the data that passes the checks. This reliable inertial reference includes at least angular velocities and specific forces that can be used for attitude prediction, and short-term heading changes independent of the magnetic field can be obtained from the yaw axis angular velocity.

[0023] In this embodiment, the inertial measurement data refers to the angular velocity data and specific force data output by the inertial measurement unit, wherein the angular velocity data is the three-axis angular velocity output by the gyroscope, and the specific force data is the three-axis specific force output by the accelerometer.

[0024] In this embodiment, the data is first preprocessed, and each inertial measurement unit outputs an angular velocity vector. and force vector ,in, Let x, y, and z be the three-axis angular velocity vector of the j-th inertial measurement unit in the k-th detection cycle; j is the channel number of the inertial measurement unit; k is the detection cycle number; and x, y, and z are the three axes of the body coordinate system. Let be the triaxial force vector of the j-th inertial measurement unit in the k-th detection cycle; , , These correspond to the specific forces along the three axes. Each magnetic field sensor outputs a magnetic field vector. ,in, Let be the magnetic field vector of the i-th magnetic field sensor in the k-th detection cycle; i is the channel number of the magnetic field sensor. , , These correspond to the three-axis magnetic field components. The main control processor adds timestamps to the sampled data, interpolates or maintains data with different output data rates to a unified detection time, and uses the mounting matrices of each sensor to unify their coordinates to the UAV's body coordinate system. If the time deviation exceeds the allowable value, the communication or timing evidence of the corresponding channel is marked as abnormal to avoid misjudging asynchronous motion as sensor differences.

[0025] Before arbitration, the inertial data is compensated for with zero bias and scaling factor, and the magnetic field data is compensated for with hard magnetic bias and soft magnetic matrix. Magnetic field compensation is as follows: ,in, Let be the original magnetic field vector of the i-th magnetic field sensor; This is the hard magnetic bias vector; This is a correction matrix for soft magnetic properties, proportionality, and installation errors. This is the compensated magnetic field vector. The calibration parameters are obtained through multi-attitude rotation sampling and fitting after the UAV is assembled and stored in non-volatile memory.

[0026] The main control processor calculates the difference between the corresponding axes of different inertial measurement units in each detection cycle. In this embodiment, the gyroscope difference is... The difference between the two forces is ,in, Corresponding to the machine body coordinate axis; For two inertial measurement units in the first... The absolute difference in angular velocity of the shaft; For two inertial measurement units in the first... The absolute difference in specific force between shafts. When any shaft Greater than the gyroscope difference threshold ,or Greater than the force difference threshold When this period is not consistent, it is considered inconsistent; only continuous periods are considered inconsistent. The condition of continuous inconsistency is met only if all detection cycles are inconsistent. Continuous counting can filter out individual sampling spikes.

[0027] The difference threshold can be determined jointly by stationary noise and flight maneuver envelope. For example, after the sensor is installed, fault-free stationary data can be collected, and the mean of the differences between each axis can be calculated. and standard deviation , and then The larger of the allowable engineering difference lower limit and the lower limit is used as the initial threshold, where c is a positive number that covers expected random fluctuations; then normal maneuver data is used to confirm that the threshold will not misjudge the allowable time alignment error as a fault. According to the testing cycle and allowed time for abnormal confirmation Set as .

[0028] With only two inertial measurement units (IMUs), persistent inconsistency only proves that at least one of them is abnormal, but cannot identify the faulty unit alone. Therefore, the main control processor synchronously maintains independent health evidence for each unit. , . This includes at least chip self-test bit, bus communication timeout or verification error, range saturation, long-term constant output during defined machine movement, and abnormal sliding window noise variance. If the two units are consistently inconsistent and only... If there is an anomaly, exclude the first unit; if only If an anomaly is detected, the second unit is excluded. If both independent pieces of evidence are normal or both are abnormal and cannot be uniquely attributed, the status "Consistency Abnormal but Faulty Party Undetermined" is output, and the two abnormal readings are not simply averaged; the previous reliable inertial reference can be temporarily maintained, and the downgraded status is reported to the flight control safety management module. When setting up more than three inertial measurement units, pairwise clustering can also be used to determine the consistency set based on independent evidence.

[0029] One or more inertial measurement units (IMUs) are used to form a reliable inertial reference. If only one channel is reliable, that channel is selected directly; if multiple channels are reliable, they can be normalized and weighted according to the calibration noise variance or health score of each channel. The reliable inertial reference includes angular velocity. And comparison It is fed into the attitude and heading estimator for state prediction and used as a reference for short-term heading changes in subsequent magnetic field detection.

[0030] For each magnetic field sensor, first calculate the deviation of its calibrated magnetic field vector magnitude from the local geomagnetic reference value, then calculate the deviation of its magnetic heading change over a preset time period from the short-term heading change given by a reliable inertial reference. The magnetic field sensor is included in the candidate set only if the magnitude deviation and the deviation satisfy their respective tolerances. The local geomagnetic reference value can be determined jointly by the geomagnetic model value corresponding to its geographical location and the calibration value of the clean area, taking into account regional geomagnetic differences and fixed deviations in the installation environment.

[0031] Local geomagnetic field reference value This can be obtained in two stages. First, the initial value of the total field strength is obtained from the geomagnetic model based on the latitude, longitude, altitude, and date of the drone's location. Then, a clean area away from steel beams, speakers, cables, and vehicles was selected within the site. The drone was then used to collect the compensated magnetic field vector in multiple attitudes, and the median of the magnitude was taken. If the data for the cleaned area is stable, then... or and The weighted result is used as , , The weights assigned to the site location satisfy 0 ≤ ≤1. If site markings are unavailable, use The modulus tolerance is relaxed accordingly. The field calibration values ​​are saved along with their acquisition time, location, and dispersion to facilitate determination of whether recalibration is needed.

[0032] For the i-th magnetic field sensor, the modulus deviation can be taken as... Normalization can also be used. ,in, Let be the absolute deviation of the magnetic field modulus of the i-th magnetic field sensor; To compensate for the magnitude of the magnetic field vector; This represents the normalized magnetic field modulus deviation. When... Not exceeding the modulus tolerance The system uses modulus length detection. This detection is suitable for identifying local ferromagnets and strong magnetic sources that significantly increase or decrease the total field strength. However, for common interferences that only change direction or cause similar deviations between two adjacent sensors, the main control processor also performs motion consistency detection. First, the compensated magnetic field vector is transformed to the horizontal plane using the current attitude estimate. Then, the magnetic heading is calculated based on the horizontal magnetic field components. At that time, the annular angle difference is used to calculate the position of the i-th magnetic field sensor in the nearest... Magnetic heading change within a detection cycle ,in, , The horizontal component of the compensated magnetic field vector obtained after attitude compensation; Use a horizontal coordinate system; It is the arctangent function in the four quadrants; For the i-th magnetic field sensor in the i-th... Magnetic heading change between the first detection cycle and the kth detection cycle; For the annular angle difference calculation, the angle is normalized to ( [180°, 180°], to avoid spurious large differences when crossing ±180°. In this embodiment, the forward direction of the aircraft is the positive x-axis, the rightward direction is the positive y-axis, and the downward direction is the positive z-axis, with the heading angle positive in the direction of increasing from north to east. At the same time, the yaw axis angular velocity in the reliable inertial reference is subtracted from the current estimated zero yaw, and then integrated within the same time window to obtain the result. ,in, Short-term heading changes given for a reliable inertial reference; The detection cycle number within the window; The heading angular velocity is determined based on a reliable inertial reference and the current attitude; The time interval between two adjacent detection moments, i.e., the length of a single detection cycle, and the time length corresponding to the changing detection window are... When the roll and pitch angles of the UAV are small, the yaw axis angular velocity in the reliable inertial reference can be approximated by subtracting zero yaw from the yaw rate. When there is a large roll or pitch, the aircraft system angular velocity is converted into yaw angular velocity based on the current attitude. The variation deviation is... , Let be the short-time magnetic-inertial variation deviation of the i-th magnetic field sensor. Not exceeding the short-time magnetic-inertial variation tolerance At that time, motion consistency is detected. Magnetism-inertia short-time variation tolerance. This refers to the maximum absolute angle difference that can exist between the magnetic heading change measured by the magnetic field sensor and the short-term heading change determined by the reliable inertial reference within the same change detection window. Only the i-th magnetic field sensor simultaneously satisfies the condition in the k-th detection cycle. and Only when the magnetic field is considered a candidate magnetic field sensor is the appropriate timeframe determined. The short-time window needs to be long enough for the actual rotation to exceed the magnetic field noise, yet short enough to make the gyroscope's zero-bias accumulation negligible. This can be achieved by statistically analyzing normal turning data from a clean environment. Distribution and determination and If the drone is approximately stationary, Even when close to zero, the magnetic heading oscillation caused by time-varying magnetic interference will still cause... Exceeding limits.

[0033] A second-level arbitration is performed on the candidate set. When there are at least two candidate magnetic field sensors, the pairwise annular angle difference of the candidate magnetic headings is calculated. , Let be the heading angle difference between the i-th and j-th candidate magnetic field sensors; The channels are connected to the same consistent set. The heading consistency tolerance refers to the maximum permissible annular angle difference between the heading angles corresponding to any two candidate magnetic field sensors within the same detection cycle. In a three-path scenario, if the difference between the heading angles corresponding to two candidate magnetic field sensors does not exceed the consistency tolerance, and the difference between the heading angle corresponding to another candidate magnetic field sensor and the heading angles corresponding to the two candidate magnetic field sensors exceeds the consistency tolerance, then the two paths that are consistent and have passed single-path detection are determined as the valid set; if all three paths are consistent, all can be considered as valid channels. If multiple sets of the same size but inconsistent with each other appear, a decision is made based on single-path deviation, historical health status, or continued observation, rather than directly elevating any one set to trustworthiness.

[0034] When only one candidate magnetic field sensor remains, the system does not immediately reject the channel due to the lack of a majority. Instead, it requires that the modulus deviation and variation deviation of the candidate magnetic field sensor be within a continuous range. Each detection cycle is confirmed to be valid only after single-channel detection. Multiple valid magnetic field sensors can be weighted according to their health status, modulus deviation, and variation deviation to form a heading reference. This single-channel-first, then inter-channel sequence ensures that even if two sensors are consistent under common interference, they will still be excluded before a majority vote if their magnetic heading changes are inconsistent with the reliable inertial reference.

[0035] In this embodiment, = , The number of consecutive confirmation cycles when there is only one candidate magnetic field sensor; This is the confirmation time for a single candidate. The fusion weight for multiple valid channels can be taken as follows: , The normalized fusion weights are for the i-th effective magnetic field sensor; This represents the set of effective magnetic field sensors for the k-th detection cycle. The summation sequence number is the effective magnetic field sensor number; Let be the quality score of the i-th effective magnetic field sensor in the k-th detection cycle, and >0; It is directly proportional to the health status coefficient and inversely proportional to the bias and noise variance, in this embodiment. ,in, Let be the health status coefficient of the i-th magnetic field sensor in the k-th detection cycle; when the magnetic field sensor is in a reliable state... =1, when it is in a transition state =α(r), when it is in an invalid state =0, and magnetic field sensors in an invalid state do not participate in the fusion; A preset positive number to prevent the denominator from being zero or too small; These are the non-negative weighting coefficients for the magnetic field modulus deviation term; For the short-time variation deviation term of magnetic-inertial change, there is a non-negative weighting coefficient. These are the non-negative weighting coefficients for the noise variance term; The calibration noise variance corresponding to the magnetic field observations involved in the fusion in the i-th magnetic field sensor; Is with A pre-defined reference noise variance with the same dimensions, and >0, used for Perform dimensionless normalization. During fusion, each magnetic field vector can be normalized first and then weighted, or the sine and cosine of each heading can be weighted separately to avoid the angle looping problem.

[0036] Assuming two external magnetic compass boards are simultaneously placed near the same stage speaker or the same high-current cable, the two readings are subject to relevant time-varying disturbances, causing... and They are close to each other. If a majority vote followed by quality inspection is adopted, the two may constitute an erroneous majority. This invention first calculates them separately. , and , Even if the common disturbance does not cause the modulus length to exceed the limit, as long as the changes in the two magnetic headings are different... If the signal fails, it will not be included in the candidate set. Therefore, the two cannot cover single-path failures through mutual consistency. If the remaining onboard magnetic field sensor passes the test, it is used after single-path confirmation; if it also fails, magnetic field updates are stopped and the heading is temporarily maintained by a trusted inertial reference.

[0037] For each inertial measurement unit and magnetic field sensor, maintain a reliable state, an invalid state, a transitional state, and a failure count. Recovery count And transition count r.

[0038] In this embodiment, an invalid state refers to a situation where the data currently output by the sensor does not meet the conditions for use as input for attitude and heading estimation. The sensor's health status can be determined as invalid when it experiences at least one of the following: malfunction, interference, communication offline, or timing anomaly. The malfunction includes sensor self-test anomalies, output saturation, continuously constant output, abnormal noise levels, or other hardware or data anomalies attributable to the sensor. The interference mainly refers to the magnetic field sensor failing magnetic field modulus detection, magneto-inertial short-time change consistency detection, or heading consistency judgment among candidate magnetic field sensors. Communication offline includes the sensor not responding within a preset time, bus communication timeout, data verification failure, or the data ready signal not being triggered. Timing anomalies include timestamp deviations of sensor data exceeding a preset timing tolerance, abnormal data arrival order, or sampling time intervals between adjacent data exceeding the allowable range. The above situations can occur individually or simultaneously.

[0039] When an anomaly can be attributed to a specific sensor based on independent health evidence or magnetic field sensor arbitration results, that sensor is deactivated and its current output data is no longer used as new input for attitude and heading estimation. When multiple inertial measurement units (IMUs) exhibit persistent inconsistencies but the faulty unit cannot be uniquely identified, resulting in an abnormal output consistency and unidentified faulty unit, a degraded state is maintained, but no single IMU is directly deactivated based on this. Sensors in an inactivated state will resume normal operation once communication and data timing are restored. After each detection cycle passes the corresponding cross-consistency test or magnetic field arbitration, it enters a transition state and then gradually re-participates in attitude and heading estimation according to the gradual weight.

[0040] If the corresponding test or arbitration fails in the current cycle under trusted conditions. Add 1, when passing Reset to zero; achieve It then transitions to an invalid state and is excluded from the estimated input. In the invalid state, during this cycle, when passing through detection or arbitration... Add 1, in case of failure Reset to zero; achieve It then transitions to a transitional state. Recovery here occurs when the sensor re-passes the original cross-consistency detection or magnetic field arbitration continuously.

[0041] During the transition period, the following methods can be adopted: As a gradient coefficient, its value ranges from 0 to... ≤1, This represents the number of consecutive detection or arbitration cycles after the sensor enters the transition state. The first detection cycle after entering the transition state is r=1. The preset transition period number is used. If the target fusion weight is... Then the actual weight of the channel is Alternatively, the measurement noise covariance of this channel can be set to a value in the state estimation filter. This causes the measurement influence to increase from weak to strong. Let be the target measurement noise covariance when the i-th magnetic field sensor is in a reliable state. If the value reaches 1 and passes the test continuously during the transition period, it becomes trusted; if it fails again during the transition period, it returns to an invalid state. When the inertial channel is restored, the angular velocity and specific force of the old trusted reference and the restored channel can be weighted as described above; when the magnetic field channel is restored, the weights of the navigation vector measurement innovation or magnetic field fusion can be weighted.

[0042] and All are pre-set positive integers, all are counted based on the normal detection cycle, and are set according to the target time. and , For failure confirmation time, This refers to the recovery confirmation time. Typically, the recovery confirmation time can be longer than the elimination confirmation time to avoid the sensor repeatedly entering and exiting at the edge of interference. However, the specific value should be determined through bench testing based on the flight control update rate, sensor noise, allowable fault response time, and the rate of change of site interference.

[0043] Attitude and heading estimators can employ extended Kalman filters, error-state Kalman filters, or other state estimation filters. A reliable inertial reference is used for state prediction by the state estimation filter, and valid magnetic field data is used for heading vector updates. The state vector can include attitude error, velocity error, position error, gyro bias, and accelerometer bias. The prediction phase uses... and Advance attitude and other states; during the magnetic field update phase, the heading or triaxial magnetic field direction provided by the effective magnetic field is used as the measurement. No measurement updates are constructed for ineffective sensors. For transitional sensors, gradual weighting or amplified measurement covariance is used, thereby gradually applying measurement innovation to the heading state.

[0044] When there is no reliable or transitional effective magnetic field sensor, magnetic field measurement updates are stopped. A short-term heading is maintained based on a reliable inertial reference, and heading-related process noise is increased. The process noise is adjusted to increase the filter covariance with the duration of no magnetism, thus explicitly reflecting heading uncertainty. At least one magnetic field sensor continuously... After each cycle is re-arbitrated, its state first enters a transition phase, and then gradually restores magnetic field constraints, avoiding the complete correction of heading differences formed during long periods of non-magnetic conditions within a single cycle.

[0045] The present invention also discloses a UAV flight control board with a redundant sensor self-arbitration function, comprising: a flight control circuit board; The flight control circuit board includes a main control processor and at least two inertial measurement units; a magnetic field sensor group comprising an onboard magnetic field sensor and at least two external magnetic field sensors; and an external sensor interface connecting the external magnetic field sensors to the main control processor. At least two of the external magnetic field sensors are respectively mounted on different external magnetic compass boards, each external magnetic compass board including a circuit board cantilever extending away from electromagnetic interference sources on the UAV body, with the external magnetic field sensor located at the far end of the corresponding circuit board cantilever. At least two of the external magnetic field sensors are connected to the main control processor via a bus connection structure with channel isolation addressing capability, enabling the main control processor to address and read each external magnetic field sensor individually. The main control processor is used for: Cross-consistency checks are performed on the inertial measurement data. Invalid inertial measurement units are identified based on the independent health evidence of each inertial measurement unit. A reliable inertial reference is formed based on the inertial measurement data that pass the check. Based on the reliable inertial reference, a short-term heading change reference is determined. Based on the magnitude deviation of the magnetic field vector output by each magnetic field sensor relative to the local geomagnetic field and the deviation between the magnetic heading change corresponding to each magnetic field sensor and the short-term heading change reference, candidate magnetic field sensors are determined. Based on the number of candidate magnetic field sensors and the heading consistency among the candidate magnetic field sensors, effective magnetic field sensors are determined. The trusted inertial reference and the valid magnetic field sensor data in the trusted state are input into the attitude and heading estimator. Sensor data that fails the detection or arbitration are excluded, and the data is converted to a transition state and re-participated in attitude and heading estimation according to the gradual weight.

[0046] To ensure real-time data acquisition from multiple inertial measurement units (IMUs), the first and second IMUs each send a sampling completion indication to the main control processor via their respective data ready signals. Upon detecting any data ready signal, the main control processor records the corresponding sampling timestamp and initiates the direct memory access channel corresponding to the bus of that IMU. Through continuous read operations, it transfers the three-axis angular velocity data, three-axis specific force data, and data status information of that IMU to the corresponding memory buffer. Different IMUs employ independent bus branches, chip select signals, and memory buffers, ensuring that the data reading process of one sensor does not block the data acquisition of another. After direct memory access is complete, the main control processor marks the data frame in the corresponding buffer as available based on the transmission completion flag. This reduces the computation time spent by the main control processor waiting for bus transmission, reserving processing time for subsequent cross-consistency checks and magnetic field sensor arbitration.

[0047] Direct memory access (DMI) ensures timely data transfer from each channel, while synchronization between multiple sensor data is achieved through timestamps and a unified detection time. The main control processor reads the most recent complete data frame from each channel within each detection cycle and interpolates, extrapolates, or maintains it to the unified detection time based on the sampling timestamp of each data frame. When the timestamp difference between different channels does not exceed a preset timing tolerance, cross-consistency detection is performed using the time-aligned data. If the data readiness signal for a channel is not triggered within a preset time, DMI is not completed within the allowed time, or the timestamp deviation of the channel's data exceeds the timing tolerance, the channel is marked as having a communication or timing anomaly, and the use of that channel's data for the current cycle for inter-channel difference calculation is stopped, thus avoiding misjudging motion differences caused by sampling asynchrony as sensor malfunctions.

[0048] For magnetic field sensors with output data rates lower than those of the inertial measurement unit, reading can be initiated by the main control processor according to a unified scheduling cycle, or triggered by the data readiness signal of the magnetic field sensor. For communication buses that support direct memory access, magnetic field data transfer can also be completed using the direct memory access channel. Each magnetic field data stream is appended with a sampling timestamp and aligned to the corresponding detection time before calculating the change in magnetic heading and the heading difference between candidate magnetic field sensors.

[0049] In one embodiment, the main control processor of the flight control circuit board of the present invention is an STM32H743VIT6; the first and second inertial measurement units are BMI088 and BMI055 respectively, and are connected to independent serial peripheral interfaces with independent chip selects. Data ready signals from each inertial measurement unit trigger sampling, and the main control processor can configure direct memory access channels for the corresponding buses to batch transfer angular velocity and force registers to the memory buffer. Specific models and sampling frequencies are only provided as examples; other devices with the same data output and health status interface are also applicable.

[0050] The onboard magnetic field sensor uses the MMC5983MA, and the two external magnetic field sensors use the IST8310. Each external magnetic field sensor is located at the far end of a slender printed circuit board cantilever on an independent magnetic compass board, with the cantilever positioned away from motors, ESCs, step-down devices, and high-current power distribution traces. The two external sensors can be connected to two independent internal integrated circuit buses of the main control processor, or to different downstream channels of a bus multiplexer; even if the sensors have the same fixed slave address, the main control can still address, read, and isolate abnormal channels separately. The magnetic compass board can be equipped with an independent low-noise voltage regulator, and high-current loops are avoided in the wiring.

[0051] Before each experiment, complete the calibration in the following order: First, after installing the sensor, complete the inertia zero bias, scale factor, mounting matrix, and hard and soft magnetic field calibrations; second, statistically analyze the dual inertia difference, magnetic field modulus deviation, and magneto-inertial variation deviation on static and normal maneuver data to determine... , , , and Third, determine based on the testing cycle and fault response requirements. , , , and , The number of detection cycles required for a gradual transition. , The preset gradual transition time is used for the fusion weight to gradually increase from the initial weight to the target fusion weight after the sensor enters the transition state; fourth, verify that all channels are stable in a reliable state during interference-free flight; fifth, apply single-channel magnetic disturbance, dual-channel common magnetic disturbance, all magnetic field disturbance, inertial channel disconnection and output constant fault respectively to confirm the state transition and estimator degradation behavior.

[0052] The execution order of each detection cycle in the firmware is as follows: read data and perform time alignment; update communication and self-test evidence; perform cross-consistency detection and form a reliable inertial reference; use this reference to calculate short-term heading changes; perform magnetic field magnitude and change deviation detection on a path-by-path basis; perform heading consistency judgment only on the candidate set; update the health status and gradual weights of all sensors; and finally submit the reliable input and corresponding weights for this cycle to the attitude and heading estimator.

[0053] Experimental comparison: Duration 60s, step size 0.1s, actual heading set to... (degree), among which, The simulation time is 20, the amplitude of the sinusoidal heading change is 0.08, the angular frequency of the sinusoidal change is 0.18, and the linear heading rate of change is 0.18. The gyro angular velocity is composed of the true heading difference, a constant zero bias of 0.04° / s, and zero-mean Gaussian noise with a standard deviation of 0.05° / s. The magnitude of the three-way magnetic heading reference magnetic field is 48μT. The random number generator uses a fixed seed 20260416.

[0054] Three types of magnetic interference were injected between 20 and 40 seconds: In the single-path interference scenario, the second magnetic heading was superimposed with a disturbance of 18sin(0.7t) + 10 degrees, increasing its magnetic field modulus by 12 μT; in the dual-path interference scenario, the second and third paths were superimposed with disturbances of 16sin(0.58t) + 8 degrees and 16sin(0.58t) + 8.35 degrees, respectively, with both paths' magnetic field moduli increasing by 2sin(0.4t) μT; in the triple-path interference scenario, the first, second, and third magnetic headings were superimposed with disturbances of 12sin(0.55t) + 7 degrees and 15sin(0.65t), respectively. The disturbances of 5 degrees and 13sin(0.75t)+3 degrees increase the magnetic field modulus by 10μT, 13μT, and 16μT, respectively. In addition to the above disturbances, each magnetic heading is superimposed with zero-mean Gaussian noise with a standard deviation of 0.55°, and each magnetic field modulus is superimposed with zero-mean Gaussian noise with a standard deviation of 0.35μT.

[0055] The simulation model has a length tolerance of 7.2 μT and a short-time magnetic-inertial variation tolerance. Take 1.8°, candidate consistency tolerance Set a 6° angle, a 0.5s change detection window, and confirm failure. =2. Restore Confirmation =15. The heading estimation uses a complementary update for easy comparison: when an effective magnetic field is present, the gyroscope prediction is corrected with an update factor of 0.15; when no effective magnetic field is present, only the gyroscope prediction is used. Simulation results comparing magnetic interference are shown in Table 1.

[0056] Table 1 Simulation results of magnetic interference comparison The results show that in the single-channel perturbation scenario, both majority voting and the present invention can eliminate isolated anomalies. When all three module lengths are significantly abnormal, both module length screening and the present invention enter a short-term non-magnetic hold state. The main difference occurs in the dual-channel common perturbation scenario: the two perturbed magnetic fields are consistent with each other, causing only majority voting and module length subsequent majority voting to still follow the erroneous majority; the present invention utilizes the condition that the changes in magnetic headings of both channels are inconsistent with the reliable inertial reference to eliminate both before inter-channel voting, and the RMSE remains at a low level throughout the experiment.

[0057] In one embodiment, when there are more than two inertial measurement units (IMUs), a consistency map can be constructed first based on the corresponding axis differences, and then the largest consistent set can be selected by combining independent health evidence. When there are more than three magnetic field sensors, the candidate sets can be clustered, and the set that meets the minimum number and historical continuity can be selected. Modulus deviation, variation deviation, and consistency difference can be implemented using absolute thresholds, normalized thresholds, or statistical thresholds that adapt to noise estimation. Gradual weighting can be linear, S-shaped, or incremental based on the filter covariance, as long as it allows newly recovered sensors to participate in attitude and heading estimation with a smaller weight initially, and then their weight is gradually increased until it reaches the weight required for normal operation. Independent health evidence can also include power supply monitoring, temperature out-of-range readings, register configuration readback failures, or watchdog reset records.

[0058] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, equivalent substitutions made to the number of sensors, processor model, bus type, state estimator, and threshold calculation method should be considered within the scope of protection of the present invention.

Claims

1. A self-arbitration method for redundant sensors in a UAV, wherein the UAV flight control board is provided with at least two inertial measurement units and at least three magnetic field sensors, characterized in that, The method includes: S1, perform cross-consistency detection on the inertial measurement data output by each inertial measurement unit. When the inertial measurement data of at least two inertial measurement units meet the preset continuous inconsistency condition, determine the invalid inertial measurement unit based on the independent health evidence of each inertial measurement unit and exclude its data. Form a reliable inertial reference based on the inertial measurement data that has passed the detection. S2, determine the short-term heading change reference based on the reliable inertial reference, and determine the candidate magnetic field sensor based on the magnitude deviation of the magnetic field vector output by each magnetic field sensor relative to the local geomagnetic field and the deviation between the magnetic heading change corresponding to each magnetic field sensor and the short-term heading change reference. S3. Determine the effective magnetic field sensor based on the number of candidate magnetic field sensors and the heading consistency among them; S4. Based on the cross-consistency detection and arbitration results of each magnetic field sensor, maintain the health status of each inertial measurement unit and each magnetic field sensor, input the reliable inertial reference and the valid magnetic field sensor data in the reliable state into the attitude and heading estimator, and exclude the sensor data in the invalid state; when an inertial measurement unit in the invalid state passes the cross-consistency detection again, or a magnetic field sensor in the invalid state passes the arbitration again, switch the sensor to the transition state, and let its data participate in the attitude and heading estimation again according to the gradual weight.

2. The UAV redundancy sensor self-arbitration method according to claim 1, characterized in that, The cross-consistency check in step S1 includes: Calculate the difference in angular velocity data and the difference in specific force data for the corresponding axes of different inertial measurement units; When the difference in the angular velocity data or the difference in the force data is continuous When the difference threshold is exceeded in a single detection cycle, the inertial measurement data is determined to be persistently inconsistent. A pre-defined positive integer; Based on the self-test status, communication status, output saturation status, output constant status, and noise anomaly status of each inertial measurement unit, independent health evidence is formed for each inertial measurement unit. When the inertial measurement data is consistently inconsistent and the independent health evidence of one inertial measurement unit is abnormal, that inertial measurement unit is identified as an invalid inertial measurement unit.

3. The UAV redundancy sensor self-arbitration method according to claim 1, characterized in that, Step S2 involves determining candidate magnetic field sensors based on the magnitude deviation of the magnetic field vector output by each magnetic field sensor relative to the local geomagnetic field, including: The geomagnetic field strength is obtained based on the location information of the UAV, and the geomagnetic field strength is corrected based on the magnetic field calibration results of the UAV's location to obtain the geomagnetic field reference value. Calculate the magnitude deviation of each magnetic field vector relative to the geomagnetic field reference value, and retain magnetic field sensors whose magnitude deviation does not exceed the magnitude tolerance as candidate magnetic field sensors.

4. The UAV redundancy sensor self-arbitration method according to claim 1, characterized in that, Step S2 involves determining candidate magnetic field sensors based on the deviation between the magnetic heading change corresponding to each magnetic field sensor and the short-term heading change reference, including: The yaw axis angular velocity is obtained from the detected inertial measurement data and integrated over a preset time period to obtain the short-term heading change reference. The magnetic heading change corresponding to each magnetic field sensor is determined based on the magnetic field vector output by each magnetic field sensor within a preset time period. Calculate the deviation between each magnetic heading change and the short-term heading change reference, and retain the magnetic field sensors whose deviation does not exceed the change tolerance as candidate magnetic field sensors.

5. The UAV redundancy sensor self-arbitration method according to claim 1, characterized in that, Step S3 includes: When there are at least two candidate magnetic field sensors, calculate the pairwise difference between the heading angles corresponding to each candidate magnetic field sensor. When the difference between the heading angles corresponding to at least two candidate magnetic field sensors does not exceed the consistency tolerance, and the difference between the heading angle corresponding to another candidate magnetic field sensor and the heading angles corresponding to the at least two candidate magnetic field sensors exceeds the consistency tolerance, the at least two candidate magnetic field sensors are determined as valid magnetic field sensors. When only one candidate magnetic field sensor exists, the modulus deviation and variation deviation of that candidate magnetic field sensor are continuous. If each detection cycle does not exceed the corresponding tolerance, the candidate magnetic field sensor is determined to be a valid magnetic field sensor. A pre-defined positive integer; When multiple effective magnetic field sensors exist, the fusion weight is determined based on the health status, modulus deviation, and variation deviation of each effective magnetic field sensor, and the heading reference is formed based on the magnetic field vectors of the multiple effective magnetic field sensors.

6. The self-arbitration method for redundant sensors in unmanned aerial vehicles according to claim 1, characterized in that, In step S3, when arbitrating each magnetic field sensor, the single-path validity of each magnetic field sensor is first determined based on the modulus deviation and variation deviation, and then the heading consistency of the candidate magnetic field sensors that pass the single-path validity determination is determined. A magnetic field sensor that fails to pass the modulus deviation judgment or change deviation judgment shall not be determined as a valid magnetic field sensor because its heading is consistent with that of other magnetic field sensors. When at least two magnetic field sensors have the same heading, but the magnetic heading changes corresponding to the at least two magnetic field sensors are inconsistent with the short-term heading change reference, the at least two magnetic field sensors are excluded.

7. The UAV redundancy sensor self-arbitration method according to claim 1, characterized in that, In step S4, the health status of each sensor includes a trusted state, an invalid state, and a transitional state. When the sensor is in a trusted state, it is in continuous If the sensor fails the corresponding test or arbitration within a test cycle, the sensor will be changed from a trusted state to an invalid state. When the sensor is in an invalid state in continuous When the sensor passes the corresponding test or arbitration within each test cycle, the sensor is changed from an invalid state to a transitional state. In the transition state, the sensor data is gradually incorporated into the attitude and heading estimation according to the gradual weight that increases over time, and the sensor is converted to a reliable state after the gradual weight reaches the target weight. in, and These are pre-defined positive integers.

8. The UAV redundancy sensor self-arbitration method according to claim 1, characterized in that, In step S4, the attitude and heading estimator is a state estimation filter, the reliable inertial reference is used for state prediction of the state estimation filter, and the effective magnetic field sensor data is used for the heading measurement update of the state estimation filter. When there are no valid magnetic field sensors in a reliable or transitional state, stop using magnetic field sensor data to update the heading vector measurement, maintain short-term heading estimation based on the reliable inertial reference, and increase the process noise corresponding to the heading estimation. When at least one magnetic field sensor passes arbitration again, it will re-participate in the navigation vector measurement update according to the gradient weight corresponding to that magnetic field sensor.

9. A UAV flight control board with redundant sensor self-arbitration function, characterized in that, include: Flight control circuit board; The main control processor and at least two inertial measurement units are mounted on the flight control circuit board. A magnetic field sensor group including an onboard magnetic field sensor and at least two external magnetic field sensors; and an external sensor interface connecting the external magnetic field sensors to the main control processor; The main control processor is used for: Cross-consistency checks are performed on the inertial measurement data. Invalid inertial measurement units are identified based on the independent health evidence of each inertial measurement unit. A reliable inertial reference is formed based on the inertial measurement data that pass the check. Based on the reliable inertial reference, a short-term heading change reference is determined. Based on the magnitude deviation of the magnetic field vector output by each magnetic field sensor relative to the local geomagnetic field and the deviation between the magnetic heading change corresponding to each magnetic field sensor and the short-term heading change reference, candidate magnetic field sensors are determined. Based on the number of candidate magnetic field sensors and the heading consistency among the candidate magnetic field sensors, effective magnetic field sensors are determined. The trusted inertial reference and the valid magnetic field sensor data in the trusted state are input into the attitude and heading estimator. Sensor data that fails the detection or arbitration are excluded, and the data is converted to a transition state and re-participated in attitude and heading estimation according to the gradual weight.

10. The UAV flight control board according to claim 9, characterized in that, At least two of the external magnetic field sensors are respectively disposed on different external magnetic compass plates. Each external magnetic compass plate includes a circuit board cantilever extending in a direction away from the electromagnetic interference source of the UAV body. The external magnetic field sensor is disposed at the far end of the corresponding circuit board cantilever. At least two of the external magnetic field sensors are connected to the main control processor through a bus connection structure with channel isolation addressing capability, so that each external magnetic field sensor can be addressed and read by the main control processor.

Images