Mobile terminal high-precision satellite trajectory processing method based on Kalman filtering algorithm

Abstract

The invention discloses a mobile terminal high-precision satellite track processing method based on a Kalman filtering algorithm, and relates to the technical field of satellite navigation and positioning, and the method comprises the steps: carrying out the versioning receiving and caching of broadcast ephemeris and high-precision correction information at a terminal, and generating a correction candidate set and a state label according to a satellite; establishing a correction state filter according to the state label, executing version transition bridging correction, late measurement backfilling and forward repropagation, outputting time-delay smooth equivalent correction and uncertainty description quantity, and forming a terminal precision satellite track and clock error product; and conservatively fusing the output in a correction layer, and mapping the fused output into an operation index and a prediction upper bound to drive an operation state machine to degrade and recover. According to the method, the track continuity in a weak network and ephemeris updating asynchronous scene is improved, the kick and reconvergence are reduced, the availability and the operation quality of agricultural automatic operation are improved, and the risk of missed spraying, re-spraying and reworking is reduced.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of satellite navigation and positioning, in particular to a high-precision satellite trajectory processing method for a mobile terminal based on a Kalman filtering algorithm. BACKGROUND

[0002] In precision agriculture and automatic operation of agricultural machinery, tractors, transplanter, sprayer and harvester usually need to be installed with a vehicle-mounted controller, a tablet terminal or a mobile terminal to realize automatic steering, inter-row tracking, land boundary constraint, variable fertilization and pesticide application and repeated operation path through satellite navigation original observation. The above operations are mostly carried out in fields, inter-row orchards, hilly terraces, forest edges and facility agriculture coverage areas, and the operations have a long time, a low driving speed, and high requirements for the stability of transverse deviation, continuous heading and distance between adjacent operation trajectories in the control closed loop. The operation area is mostly located at the edge of a communication base station coverage or a cross-operator switching area, and there are factors such as shielding by the metal structure of agricultural machinery, reflection by a shed frame and attenuation by a tree crown. The continuous availability of high-precision positioning services directly affects the width utilization rate, input utilization rate and operation safety.

[0003] The existing high-precision positioning of a terminal is generally based on broadcast ephemeris superimposed with external corrections (orbit correction, clock difference correction and signal deviation correction), and a Kalman filter is used to realize state estimation and smoothing.

[0004] The correction information can be obtained from a cellular network or a satellite broadcast link. When the network is weak or disconnected, the correction message is often delayed, intermittently missing or out of order, and different types of corrections are not updated synchronously in terms of time and content. Broadcast ephemeris is updated periodically and has a version identifier, while the correction information is often related to a specific ephemeris version. When the ephemeris update and the correction update are inconsistent, the terminal will obtain a new ephemeris and old correction or old ephemeris and new correction in the same epoch, which will introduce systematic deviation in the calculation of satellite position and clock difference, resulting in abnormal filter innovation, weight mismatch or re-initialization. In order to ensure operation, some terminals will use the last epoch correction or short-term prediction when the network is short-term missing, and switch to a lower level positioning mode when the failure occurs. When the ephemeris version is switched, the message is delayed and backfilled, and the link is jittering, the time consistency and version consistency of the correction input will also be damaged. If such discontinuity or mismatch occurs during operation, it will cause positioning solution jump, trajectory drift, automatic driving exit or long-term re-convergence, and further cause missed spraying, repeated spraying, row spacing deviation, land boundary overrun and rework downtime.

[0005] Therefore, the technical problem to be solved at present is that when the network is weak or disconnected and the ephemeris and correction information updates are inconsistent, the correction message is delayed, missing, out of order and inconsistent in version, and the terminal side cannot continuously provide time-consistent and version-consistent satellite trajectory correction input for high-precision solution, resulting in discontinuous positioning and repeated convergence during operation. SUMMARY

[0006] The technical problems solved by the present application are as follows:

[0007] In view of the deficiencies of the prior art, the present application provides a mobile terminal high-precision satellite trajectory processing method based on a Kalman filtering algorithm, which performs version transition bridging correction, late measurement backfilling and forward repropagation, outputs time lag smoothing equivalent correction and uncertainty description quantity, forms terminal precise satellite trajectory and clock difference products, and conservatively fuses the outputs in the correction layer to map operation indicators and prediction upper bounds to drive operation state machine degradation and recovery. The method improves trajectory continuity in a weak network and unsynchronized ephemeris update scenario, reduces sudden jumps and re-convergence, improves agricultural automatic operation availability and operation quality, and solves the technical problems described in the background art.

[0008] The technical solutions are as follows:

[0009] To achieve the above object, the present application is implemented by the following technical solutions:

[0010] The mobile terminal high-precision satellite trajectory processing method based on the Kalman filtering algorithm comprises the following steps: receiving broadcast ephemeris and high-precision correction information, writing broadcast ephemeris version identifiers, message time stamps of high-precision correction information, associated version identifiers and field integrity flags into a correction record database and buffering according to satellites to form a satellite-by-satellite correction candidate set and a state label;

[0011] According to the state label, a correction state filter is established for each satellite, Kalman filtering prediction and update are adopted, a precise satellite state anchor and a broadcast ephemeris difference are generated to form a bridging correction vector during version transition, backfilling update and forward propagation are performed during late arrival or disorder, and a fixed time lag smoothing equivalent correction vector and an equivalent uncertainty description matrix are outputted;

[0012] Terminal precise satellite trajectory and clock difference products are generated by using the equivalent correction vector and the equivalent uncertainty description matrix, and the products are mapped into operation indicator vectors and prediction upper bounds, compared with operation threshold coefficients and recovery threshold coefficients to drive an operation state machine to output operation levels and degradation or recovery instructions.

[0013] Further, the correction record database establishes a circular buffer according to satellite numbers and a time index according to message time stamps, inserts corresponding positions according to the time index when receiving high-precision correction information, writes broadcast ephemeris version identifiers, associated version identifiers, field integrity flags, delay durations and correction categories into corresponding correction records, and eliminates correction records outside the buffer window according to time.

[0014] Further, the correction record is selected from the per-satellite correction candidate set and the state label is generated for each satellite at each epoch, wherein the version consistent is available when the associated version identifier is consistent with the broadcast ephemeris version identifier and the field is complete, the version transition is when the broadcast ephemeris version identifier has been updated while the correction record corresponds to the old version, the late or out-of-order is when the message timestamp is behind the current epoch and within the backfill window, and the real interruption is when there is no available correction record for consecutive epochs.

[0015] Further, the version transition is determined based on the version difference, and the version difference is determined according to the minimum difference of the wrap-around of the broadcast ephemeris version identifier and the associated version identifier; when the associated version identifier is missing, the corresponding correction record is marked as version identifier missing and its priority in the per-satellite correction candidate set is reduced.

[0016] Further, the state quantity of the correction state filter includes the orbit three-way correction, the clock correction and its drift term, and when the version consistent is available, the correction record is used as the measurement input to perform the update, and when the real interruption occurs, only the prediction is performed and the equivalent uncertainty description matrix is expanded, and the update process introduces the measurement injection coefficient according to the state label to adjust the measurement update strength.

[0017] Further, the precise satellite state anchor point is determined by the terminal precise satellite trajectory and clock error product of the last epoch, the bridge correction vector is formed by the difference between the satellite position and clock error obtained by the precise satellite state anchor point and the broadcast ephemeris propagation, and the position difference is projected according to the radial, tangential and normal unit vectors to generate the orbit three-way bridge correction.

[0018] Further, when the late or out-of-order occurs, the historical state of the correction state filter is saved in the cache window of the correction record library, the backfill update is performed according to the message timestamp to locate the historical epoch after receiving the late correction record, and the forward propagation is performed to the current epoch according to the recorded state label.

[0019] Further, when the terminal receives multiple source corrections, the compatibility screening is first performed based on the difference sequence of the terminal precise satellite trajectory and clock error product, then the covariance intersection method is used to obtain the fusion result for the equivalent correction vector and the equivalent uncertainty description matrix that pass the screening in the correction layer, and the one-dimensional search is used to determine the intersection weight and the positive definite processing is performed on the fusion matrix.

[0020] Further, the operation index vector is calculated based on the terminal precise satellite trajectory and clock error product in the operation line coordinate system, the prediction upper bound is determined by the operation uncertainty obtained by the sensitivity mapping of the equivalent uncertainty description matrix, and the maximum eigenvalue of the operation uncertainty is obtained by the power iteration as the prediction upper bound, wherein the sensitivity mapping is obtained by perturbing each component of the equivalent correction vector by a preset perturbation step and recalculating the operation index vector.

[0021] Further, the job state machine sets four states of keeping, degradation preparation, degradation execution, and recovery preparation, state migration is triggered by comparison results of the predicted upper bound and the job threshold coefficient and the recovery threshold coefficient, and requires that the migration is executed after a plurality of epochs satisfy the same triggering condition; the job state machine outputs a job level, a predicted sustainable job time, and an instruction corresponding to the current state, the instruction includes a degradation instruction and a recovery instruction and is written into a job log.

[0022] (Three) beneficial effects:

[0023] The application provides a mobile terminal high-precision satellite trajectory processing method based on a Kalman filtering algorithm, and has the following beneficial effects:

[0024] By versioning receiving and caching the broadcast ephemeris and high-precision correction information, generating a satellite-by-satellite correction candidate set and a state label, making the version consistent available, version transition, late or out-of-order, and real interruption separated, providing traceable, order-arrival-sequence-affected input and review for subsequent processing. According to the state label, a correction state filter is established, and a bridge correction vector is generated by differentiating the precise satellite state anchor point and the broadcast ephemeris during the version transition, so that the terminal's precise satellite trajectory and clock difference products are still continuous during the ephemeris asynchronous stage, avoiding trajectory mutation and resetting caused by correction mismatch.

[0025] For late or out-of-order correction records, backfilling and forward retransmitting to the current epoch according to the message timestamp, outputting a fixed time lag smoothed equivalent correction vector, and weak network backfill information entering the output according to time consistency, avoiding control side input transient jump.

[0026] Output the equivalent correction vector, and output the equivalent uncertainty description matrix at the same time, adjust the uncertainty growth according to the path semantics during the real interruption or version transition process, and avoid excessive confidence in the bridge, backfilling or prediction stage output according to the reliability of the upstream while the downstream.

[0027] When multiple sources of correction coexist, first, the difference sequence of the terminal precise satellite trajectory and clock difference product is compatible filtered based on compatibility, and the equivalent correction vector and the equivalent uncertainty description matrix are conservatively fused at the correction layer, to avoid the breakage jitter caused by hard switching at the result layer.

[0028] Map the terminal precise satellite trajectory and clock difference product and the equivalent uncertainty description matrix into a job index vector and a predicted upper bound, and then compare them with the job threshold coefficient and the recovery threshold coefficient to drive the job state machine to output the job level and the degradation or recovery instruction, and record the job log, so that the state label, the bridge backfilling and the conservative fusion are mutually coordinated in the job closed loop. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1A flowchart of a high-precision satellite trajectory processing method for a mobile terminal according to the present application. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0031] Please refer to Figure 1 The present application provides a high-precision satellite trajectory processing method for a mobile terminal based on a Kalman filtering algorithm, comprising: step one, organizing broadcast ephemeris and high-precision correction information into traceable records with version, time and integrity on the mobile terminal side, and forming a satellite-by-satellite correction candidate set and a state label that can directly trigger subsequent processing at each epoch, so as to decompose weak network phenomenon into a state that can be handled, rather than being classified as interruption in general.

[0032] In an agricultural automatic operation site, the transmission link of correction information may be temporarily unavailable, intermittently available or message congestion backfilling, resulting in the correction information of the same satellite entering the terminal in the form of delayed arrival, out-of-order arrival and segmented missing. At the same time, the broadcast ephemeris is updated periodically, and the correction information often carries the associated ephemeris version identifier. When the ephemeris update and the correction update are not synchronized, the terminal may simultaneously hold input combinations with inconsistent current broadcast ephemeris version identifier and correction associated ephemeris version identifier in the same epoch.

[0033] If the terminal only uses whether the correction is received as the basis for judgment, the version transition period and the late backfilling period will be misjudged as interruption, and then unnecessary reset and switching will be introduced in the subsequent filtering and control link. Therefore, step one needs to first unify the input objects and solidify them into traceable records, and then output a state label through consistency diagnosis as a trigger condition for step two to select cross-version bridging, late backfilling update or pure prediction.

[0034] At the beginning of each epoch, the mobile terminal first solidifies the broadcast ephemeris account by satellite, which provides a reference benchmark for subsequent consistency diagnosis. The account does not pursue to be expanded into a complex model, but clearly archives the broadcast ephemeris of each satellite as the current available reference object, so that all correction records in the same epoch can be traced back to the same reference version.

[0035] If the reference benchmark is not solidified first, the reference ephemeris may drift due to decoding sequence differences in the same epoch, and then the same correction record may get different conclusions at different judgment times, thereby destroying the stability of the state label from the source.

[0036] After receiving the broadcast ephemeris, the mobile terminal first completes the verification and field extraction, and then writes the satellite identification, broadcast ephemeris version identification, broadcast ephemeris applicable time period, and broadcast ephemeris generation time into the broadcast ephemeris table, and generates a unique reference pointer within the epoch for the satellite, which is replaced only when the ephemeris version is updated.

[0037] The mobile terminal uses a fixed-length record area indexed by satellite identification to save the broadcast ephemeris table, and uses a version identification change trigger writing method to avoid repeated coverage; at the same time, the broadcast ephemeris table is bound to the epoch time, so that all subsequent diagnoses are based on the broadcast ephemeris table of the epoch.

[0038] In use, the broadcast ephemeris table is first solidified by satellite, and then the version identification change trigger method is used to maintain consistency within the epoch. The broadcast ephemeris forms a stable reference within each epoch, which facilitates subsequent consistency comparison between correction records and reference versions; avoids contradictory diagnostic conclusions for the same correction record in the same epoch due to changes in reference objects.

[0039] After completing the solidification of the broadcast ephemeris table, the mobile terminal uniformly writes high-precision correction information into the correction record library, and establishes a circular buffer area by satellite granularity, so that even if the correction information arrives out of order, it can still be written into the corresponding time position.

[0040] After the accumulation of messages in a weak network, the terminal will receive multiple correction records for the same satellite in a short period of time. If written in the order of reception, the historical corrections will be erased, making it impossible to determine whether an epoch can be effectively corrected through backfilling. When analyzing each correction information, the mobile terminal fixedly extracts the satellite identification, correction category, correction record message time, correction associated ephemeris version identification, field integrity flag, and delay amount, and writes them into a correction record; the delay amount is used to depict the time offset of the correction record relative to the current epoch, and is taken as the absolute value of the difference between the epoch time and the correction record message time.

[0041] The mobile terminal establishes a circular buffer area according to the first index, and a time index structure according to the second index, which is the message time of the correction record. When arriving out of order, the correction record is inserted into the second message time; the time index structure can use a balanced binary search tree or a skip list, so that the insertion and search have an upper bound, and at the same time, during the wraparound process of the circular buffer area, old records that are too long are eliminated. In use, a circular buffer area is established by satellite granularity, a time index structure is established by the message time of the correction record, and out-of-order insertion and windowed elimination are supported; the correction record still retains the time semantics after weak network backfilling, and subsequent division into late backfilling and real missing, correction record writing rules are available, avoiding the loss of historical information caused by overwrite.

[0042] After the broadcast ephemeris table and the correction record database are inputted, the mobile terminal needs to establish a satellite-by-satellite correction candidate set for each satellite, and assign a continuous and comparable confidence measure to all correction records in each candidate set, to avoid using binary logic to switch between available and unavailable. The confidence measure is used for sorting and screening, so that the subsequent state tag generation has stable input rather than accidental arrival order. There may be multiple available correction records in the same epoch: some records have the same version but too large a delay, and some records have too small a delay but missing fields; if there is no continuous quantization, the state tag will frequently flip with the arrival order of the records, and then the input fluctuations will be transmitted to the subsequent filtering and control link.

[0043] The mobile terminal takes the broadcast ephemeris version identifier in the broadcast ephemeris table as a reference to calculate the version difference of each correction record, and combines the field integrity flag to obtain the completeness, and then maps the delay amount to a monotonically decreasing weight, and finally forms a version consistency index for sorting. The version consistency index adopts the form of completeness x index attenuation x soft penalty, so that the version difference non-linearly suppresses the index, so that records that are not consistent in version but can still be used for transition judgment are left in the candidate set rather than being discarded directly, facilitating the generation of version transition state tags, wherein:

[0044] ;

[0045] In the formula, the version consistency index is a continuous quantity for sorting the confidence of the same type of correction records for the same satellite, ; the completeness is a continuous quantity mapped from the field integrity flag, , which takes a high value when the field is complete, and is reduced according to the missing category when the field is missing; the delay amount is the time offset between the epoch time and the correction record message time, and the value range is ;

[0046] The delay penalty coefficient is an index attenuation intensity parameter, and the value is , the larger the value, the faster the weight decreases with the same delay amount; the version difference is the difference measure between the broadcast ephemeris version identifier and the correction associated ephemeris version identifier, which is a non-negative integer, and a value of 0 indicates version consistency, and a value increase indicates an increase in version deviation; the version penalty coefficient is a soft penalty intensity parameter, and the value is ; the soft penalty mapping is defined as:

[0047] ;

[0048] The soft penalty mapping has a value range of , for mapping version difference to dampening factor; version penalty coefficient : value range is ; version difference : value range is non-negative integer set; boundary handling can be added in implementation: when exceeds preset upper bound , directly set , so as to avoid exponential overflow; preset upper bound is positive real number, which can be determined by terminal numerical range.

[0049] When sorting candidate set, the mobile terminal arranges the correction records of the same satellite and the same correction category according to the version consistency index from large to small, and encapsulates the sorting result together with the field integrity flag and the delay amount as the satellite-by-satellite correction candidate set.

[0050] In use, the version consistency index is calculated, the correction records are sorted and encapsulated, and the stable input of the satellite-by-satellite correction candidate set is formed; the candidate set sorting is driven by continuous quantity, which reduces the state label flip caused by the change of record arrival order; the version inconsistent records are retained in the candidate set but the weight is suppressed, which provides traceable basis for subsequent version transition determination.

[0051] After obtaining the satellite-by-satellite correction candidate set, the mobile terminal needs to compress the input state of each satellite at the current epoch into a single state label, and make the label maintain time continuity under weak network jitter. The difficulty lies in distinguishing between late arrival or out-of-order and real interruption: the former may still have records covering the current epoch in the correction record library or have history records that can be backfilled, and the latter has no available records in the window.

[0052] If all corrections not arriving at the current epoch are determined as interruptions, the late arrival backfill will be misjudged as an interruption, and the subsequent continuous correction generation will be forced to take the pure prediction path; on the contrary, if all delays are determined as backfillable, the expired records may be misused as current input. The mobile terminal first maps the delay amount to a late arrival availability coefficient, and then generates four types of state labels according to the sorting result of the version consistency index and the field integrity flag: version consistent availability, version transition, late arrival or out-of-order, and real interruption. The late arrival availability coefficient adopts a segmented function of full availability within the window, exponential decay outside the window, and zero when exceeding the upper limit, so that the late arrival backfill has clear availability semantics within a reasonable window and naturally fails when exceeding the upper limit; at the same time, a hysteresis determination rule is introduced, which requires consistent evidence for consecutive epochs when the state label switches from version consistent availability to late arrival or out-of-order or real interruption, so as to avoid link jitter causing label jitter, wherein:

[0053] ;

[0054] In the formula: late arrival availability coefficient : a continuous quantity used to characterize the backfill semantics of a late correction record at the current epoch, taking values in the range ; delay : the time offset between the epoch time and the correction record message time, taking values in the range , which is used as an argument in the function to determine the degree of lateness.

[0055] backfill upper limit : the time threshold for full availability of a late correction, taking values ; the late availability coefficient takes the value when the delay does not exceed the backfill upper limit : the time threshold for discarding of a late correction, taking values ; the late availability coefficient takes the value : the exponential decay strength parameter, taking values , which acts to continuously depress the late availability coefficient between the backfill upper limit and the discard upper limit ;

[0056] delay is taken as the absolute difference between the epoch time and the correction record message time :

[0057] ;

[0058] epoch time : the measurement time stamp from the receiver observation epoch; message time : the reference time from the correction text; if the text does not contain a reference time, the text arrival time is taken and the time source type is written into the record header field; meanwhile, the backfill threshold: discard upper limit is not greater than the window length of the correction record circular buffer , to ensure that the records allowed to be backfilled can be found in the cache.

[0059] In the state label generation, the mobile terminal takes the first record of the satellite-by-satellite correction candidate set as the current preferred record, and reads its version consistency index and late availability coefficient: when the version difference of the preferred record is 0 and the field integrity flag meets the requirements, it is marked as version consistent available; when the version difference is not 0 but the version consistency index is still non-zero and the late availability coefficient is not 0, it is marked as version transition; when the version difference is 0 but the late availability coefficient is significantly less than 1 and the preferred record message time is behind the epoch time, it is marked as late or out of order; when there is no record in the candidate set that meets the non-zero late availability coefficient and the field integrity flag meets the requirements, and this state lasts across the preset epoch count threshold, it is marked as real interruption. The preset epoch count threshold is used to realize hysteresis, so that the state label does not frequently flip due to the absence of a single epoch under link jitter.

[0060] Finally, the mobile terminal encapsulates and outputs the state label of each satellite together with the satellite-by-satellite correction candidate set, and the output object includes satellite identification, correction category, preferred record, alternative record, delay, field integrity flag, version consistency index, late availability coefficient, and state label body.

[0061] In use, the late availability coefficient is used to describe the backfill semantics, and the version consistency index and hysteresis are used to determine the generation of stable state labels and output the encapsulated object. Late and expiration are clearly distinguished by a segmented function, avoiding the misinclusion of expired records into the current epoch input semantics. The state label has hysteresis characteristics and remains continuous under link jitter, facilitating the triggering of a fixed path according to the label in subsequent steps.

[0062] Step two, establish a correction state filter for each satellite according to the satellite-by-satellite correction candidate set and the state label, and still form continuous equivalent correction and uncertainty description under the conditions of version transition and late or out of order, and then output continuous terminal precise satellite trajectory and clock difference products.

[0063] The object faced is the satellite-by-satellite correction candidate set and the state label output in step one. The mobile terminal no longer uses whether it has received a correction message as the only judgment, but uses the state label as the action switch of the correction state filter. To avoid the alternating appearance of fresh records and late records in the boundary epoch, resulting in the jumping of update intensity, a measurement injection coefficient is introduced and linked with the version consistency index and the late availability coefficient, so that the measurement update presents continuous intensity changes in time.

[0064] Link jitter in the field of agricultural operation will make the same satellite's correction records present different freshness and different version consistency in adjacent epochs. If the filter takes the same strength of measurement update for each record, it is easy to mix the version consistent but late record with the version inconsistent but transitional record, and further transfer the input fluctuation to the correction estimation fluctuation. Therefore, the version consistency index and the late available coefficient in step one need to be introduced into the update rhythm to form a reviewable injection rule.

[0065] The mobile terminal reads the state tag of each satellite respectively: when the state tag is version consistent available, the correction state filter performs measurement update; when the state tag is real interruption, the correction state filter only performs prediction and keeps measurement update closed; when the state tag is version transition or late or out of order, the correction state filter does not directly reject the candidate record, but scales the measurement information with the measurement injection coefficient, so that the update strength changes continuously with the version consistency index and the late available coefficient. Therefore, the measurement injection coefficient does not change the semantics of the state tag in step one, nor does it change the body of the candidate record, but changes the strength of the candidate record entering the filter.

[0066] The mobile terminal takes the preferred record from the satellite-by-satellite correction candidate set, and then reads the version consistency index and the late available coefficient corresponding to the preferred record , and combines them into the measurement injection coefficient , so that the injection strength and the degree of version consistency and freshness change in the same direction. The specific form of the measurement injection coefficient adopts a logarithmic compression form to avoid causing the injection strength to change too steeply at close to one, and to keep the filter update rhythm continuous:

[0067] ;

[0068] In the formula: the measurement injection coefficient : a continuous coefficient for scaling the injection strength of the measurement information, with a value range of ; for scaling the measurement information increment; or, equivalent increment measurement noise, for soft injection; the version consistency index : a continuous quantity for measuring the matching degree of the correction record and the broadcast ephemeris version, the field completeness and the freshness, with a value range of ;

[0069] The late available coefficient : a continuous quantity for measuring whether the correction record still has backfill semantics under a given delay amount, with a value range of ; the delay amount : the time offset between the epoch time and the correction record message time, with a value range of ; after the injection coefficient is formed, the mobile terminal enters the high interval in the version consistent available state and takes the preferred record as the measurement update input, takes 0 and closes the measurement update in the real interruption state, and scales the measurement information in the version transition and late or out-of-order state, so that the strength of the measurement update continuously changes with the quality of the candidate record.

[0070] To reduce numerical instability, the correction state filter can use information form Kalman filter or square root information filter, and the measurement injection coefficient directly acts on the scaling of the information increment, so that the condition number of the matrix decomposition in the update process is kept controllable.

[0071] In use, the state label determines the opening and closing of the measurement update, and the measurement injection coefficient introduces the version consistency index and the late available coefficient into the measurement information injection strength, so that a continuous update rhythm is formed. The four types of semantics of the state label are reserved as action trigger entrances, and the filter update path remains consistent between epochs, reducing the action swing introduced by link jitter; the measurement injection coefficient makes the strength of the version consistent but late and the inconsistent version in transition entering the filter have continuous changes, avoiding the update mutation caused by forced full injection or full rejection in a single epoch.

[0072] For satellites in the state label version transition, the mobile terminal needs to continuously generate available correction input during the broadcast ephemeris version switching and correction of unsynchronized stage, so that the terminal precise satellite trajectory and clock error product remain continuous at the version boundary. Instead of directly equating version transition to real interruption, the precise satellite state anchor point is used as a reference to construct a cross-version bridging correction and inject the correction state filter.

[0073] Version transition is essentially an input semantic misplacement of reference model switching preceding correction switching, and its mechanism is as follows: the broadcast ephemeris version identifier of the same satellite changes in adjacent epochs, while the available correction record in the per-satellite correction candidate set still points to the old version identifier. If the old correction is still superimposed on the new ephemeris at this time, a systematic bias will be introduced and the filter innovation will be abnormal; if the correction is directly rejected and the pure prediction is entered, the version transition will be mistaken for a link interruption, and then unnecessary prediction drift will be caused.

[0074] Therefore, the terminal precise satellite trajectory and clock error product formed in the last epoch is needed as an anchor point, and the anchor point is re-expressed under the framework of the new version ephemeris, so as to form an injectable bridging correction.

[0075] The mobile terminal fixes the terminal precise satellite trajectory and clock error product used for output in each epoch as a precise satellite state anchor point at the end of each epoch, which contains the position component and clock error component of the satellite in the earth-fixed coordinate system; when the next epoch enters and the state label is determined to be version transition, the mobile terminal solves the broadcast satellite state of the same satellite , and then a bridging correction vector is generated by the difference between the two The bridging correction vector is written back to the first position of the satellite-by-satellite correction candidate set and participates in the correction state filter update as the first record under the version transition condition.

[0076] Since the bridging correction comes from the difference construction rather than the original correction message, the injection strength is required to be controlled by the measurement injection coefficient, and the uncertainty description quantity is monotonically tightened with the version transition duration, avoiding the backend from over-relying on the bridging correction during the transition period.

[0077] The bridging correction vector adopts a difference construction form, which is specifically represented as:

[0078] ;

[0079] In the formula, the bridging correction vector is a correction vector used to express the difference between the anchor point and the new broadcast ephemeris during the version transition, including a position difference component and a clock difference component, and the value range is constrained by the satellite orbit scale and the clock difference scale; the epoch time is a time parameter used to identify the current processing time, and the value range covers the terminal continuous operation period; the precise satellite state anchor point is a satellite state vector obtained by solidifying the terminal output at the last epoch, including at least a satellite position component and a clock component;

[0080] The broadcast satellite state is a satellite state vector obtained by solving the current broadcast ephemeris table at the epoch time, including at least a satellite position component and a clock component;

[0081] The precise satellite state anchor point is obtained by solidifying the terminal precise satellite trajectory and clock difference product output externally at the last epoch, including at least a satellite geodetic coordinate position component and a clock component. When the system is started and there is no anchor point, the anchor point is initialized by the satellite state solved by the broadcast ephemeris, and the anchor point source flag is set to broadcast initialization; when the first version consistent correction update occurs, the anchor point source is switched to the correction superposition output. In this way, the cross-version bridging has an anchor point source before the first version transition arrives.

[0082] The epoch time is taken from the system time of the receiver observation epoch (such as the measurement time mark), and the message time is taken from the reference time carried by the correction message; when the correction message does not provide the reference time, the message time can be replaced by the message arrival time, and the time source type is marked in the correction record, so that different processing paths are selected in the subsequent steps.

[0083] After the differential construction, the mobile terminal adds the bridging correction vector as a preferred record into the per-satellite correction candidate set, and sets the version consistency index of the record to a transitional value distinguishable from the version-consistent available value, so that the measurement injection coefficient is in the soft injection interval during the transition period. The broadcast satellite state solution uses the standard solution process of the broadcast ephemeris to obtain the satellite position and clock error, and the precise satellite state anchor point is derived from the terminal output at the previous epoch without the need for a measurement station. The position component can be smoothed by cubic Lagrange interpolation to align the values at the epoch, and the clock error component can be interpolated to the same time point, so that the differential construction time is consistent.

[0084] For example, when a tractor working in the orchard rows, the on-board terminal is congested in the cellular link when turning at the end of the row, and then the broadcast ephemeris completes the version modification, while the network correction is still the old version. The terminal marks the satellite as a transitional version in the screen log, and takes the precise satellite state anchor point from the output at the previous epoch, and solves the broadcast satellite state with the new broadcast ephemeris to form a bridging correction vector which is written into the candidate set. The driver does not find the trajectory suddenly jumping, and the controller turns along the established row line. After several epochs, the correction messages are synchronized, the terminal restores the state label to version-consistent available, and the bridging correction record naturally exits the preferred position.

[0085] The terminal outputs the solidified precise satellite state anchor point , and solves the broadcast satellite state with the new broadcast ephemeris , and constructs a bridging correction vector through differential construction and writes it back to the per-satellite correction candidate set to participate in filter update in a soft injection manner. The version transition is expressed as an injectable bridging correction rather than a direct interruption, and the terminal precise satellite trajectory and clock error product maintain continuous input semantics at the version boundary. The bridging correction vector is constructed with reference to the anchor point and controlled by the measurement injection coefficient, so that the injection strength during the transition period is controllable, and the differential construction quantity is avoided to be used as high-trust original correction.

[0086] For satellites with delayed or out-of-order state labels, the mobile terminal needs to convert the delayed arrival correction record into continuous correction estimation that is still meaningful at the current time, rather than simply superimposing the delayed record directly on the current broadcast ephemeris.

[0087] Through historical filter state reservation, delayed epoch backfill update, and forward repropagation after backfilling to form a time sequence closed loop, the correction estimation sequence is self-consistent on the time axis.

[0088] Weak network backfilling is usually represented as multiple epoch correction records arriving in a set. If only the latest arriving record is considered as the current measurement, the record with an earlier timestamp will be mistakenly used for the current epoch, which will destroy the time consistency of the correction. If all the late records are discarded, the effective information that can be backfilled will be lost, and the correction state filter will stay in the pure prediction mode for a long time, and the uncertainty description will gradually become more severe. In order to balance the time consistency and information utilization, the backfilling update needs to be completed at the message time of the late record, and then the update influence is propagated to the current epoch along the time axis.

[0089] The mobile terminal saves the historical state of the correction state filter at each epoch. The historical state includes the correction estimate value, the uncertainty description, and the trigger path label of the epoch. When receiving the late record, the terminal locates the corresponding epoch in the historical state sequence according to the message time of the late record, and updates the input with the late record as the measurement, and scales the update strength with the measurement injection coefficient. After the update is completed, the terminal performs prediction and necessary bridge injection for each subsequent epoch in chronological order until it is pushed to the current epoch, so that the current output reflects the information increment brought by the late record.

[0090] The step one late available coefficient limits the backfilling window. If it exceeds the upper limit, it will not enter the backfilling chain, which ensures the boundary of the time closed loop from the source. The mobile terminal implements a ring sequence of satellite-independent historical states, and the index is determined by the message time in the step one correction record library. The positioning process locates the historical epoch position through the time index structure, and then performs backfilling update. The forward re-propagation after backfilling update does not save all the observation information of each epoch, but saves the trigger input summary, including the state label, the preferred record identifier, and the measurement injection coefficient of the epoch. The re-propagation uses the trigger semantics at that time without relying on re-diagnosis.

[0091] The state transition during re-propagation can use the constant speed discretization form (use constant or constant drift discrete model, and the discrete interval is given by the epoch interval). The discretization coefficient is realized by matrix exponentiation or first-order Taylor expansion. The engineering implementation of matrix exponentiation can use Padé approximation and square power process, so that the terminal can still be executed under the condition of limited computing power.

[0092] In use, the historical filter state sequence independent of satellites is saved. After the late record arrives, the backfilling epoch is located according to the message time and the measurement update is performed. Then the late information is propagated forward to the current time along the epoch sequence, so that the late information enters the current correction output in a time-consistent manner. The late record is not directly applied to the current epoch. The backfilling update occurs at the message time, which ensures the time consistency and traceability of the correction estimate sequence. The forward re-propagation reuses the state label and the measurement injection coefficient at that time, so that the late backfilling will not introduce an update path inconsistent with the semantics at that time, and the output will not appear abrupt jump.

[0093] After the cross-version bridging and late backfilling are completed, the output of the correction state filter is organized as a fixed time-lag smoothing output, and is superimposed with the broadcast ephemeris solution result to encapsulate the terminal precise satellite trajectory and clock error product. The fixed time-lag smoothing output does not change the timing update inside the filter, but only changes the output time and output method, so that the control closed loop obtains more stable input semantics.

[0094] Agricultural automatic operation control usually relies on continuous trajectory input. If the current epoch correction estimate value is directly output externally, the information supplement at the time of late backfilling may form a significant update amplitude change at the current epoch, which in turn manifests as a transient change in the lateral command at the control side. The fixed time-lag output moves the external output point to the past by a certain time lag, and performs kernel smoothing on the correction estimate sequence within the time lag window, so that the output is more in line with the requirements of the control system for smooth input, while retaining the information increment brought by late backfilling.

[0095] The mobile terminal generates a correction estimate sequence at each epoch, and sets the external output time as the current epoch minus a fixed time lag, which is determined by the sampling period of the operation control. Near the output time, the mobile terminal takes out the correction estimate values of several adjacent epochs from the historical state sequence, and uses exponential kernel weights for smoothing and synthesis, with the weight monotonically decaying with time distance, so that the estimate values closer to the output time have a larger proportion. After smoothing, the terminal uses the broadcast ephemeris account solution output to output the broadcast satellite state at the output time, and then superimposes the smoothed correction to obtain the terminal precise satellite trajectory and clock error product, and outputs the uncertainty description quantity along with it.

[0096] The fixed time-lag smoothing output adopts a discrete exponential kernel form, and the smoothing output is given by the weighted sum of the correction estimate sequence:

[0097] ;

[0098] In the formula: smoothing output : the smoothed correction vector corresponding to the fixed time-lag output time, including the three-direction correction components of the orbit, the clock error correction components, and the necessary bias correction components; correction estimate sequence : the correction estimate vector output by the correction state filter at the historical epoch; epoch subscript : represents the historical offset relative to the current epoch, with a value range of an integer set ; sample number : the number of historical samples included in the smoothing window, a positive integer; decay coefficient : the decay strength parameter of the exponential kernel, which is a positive real number. The larger the value, the faster the weight decays with the historical offset;

[0099] Further, the output time of the fixed time delay can be directly indexed by the historical state sequence. If the output time falls between two epochs, the components of the output can be interpolated on the time axis using a cubic Hermite interpolation. The first derivative required for interpolation can be given by the drift term inside the correction state filter or the difference between adjacent samples to ensure the feasibility of the interpolation process.

[0100] After completing the smoothing output, the mobile terminal encapsulates it with the broadcast satellite state as the terminal precise satellite trajectory and clock difference product, and writes the state tag source information into the output header field, so that step three can identify that the output comes from the consistent version available, version transition or late backfill path.

[0101] Specifically, the output time is determined based on the fixed time delay, and the correction estimation sequence is smoothed and synthesized to form a smoothed output using an exponential kernel weight , and then encapsulated with the broadcast ephemeris solution result as the terminal precise satellite trajectory and clock difference product and carrying the uncertainty description quantity.

[0102] In use, the output point is shifted from the current epoch to the fixed time delay epoch and is smoothed and synthesized using the kernel, so that the information supplement caused by late backfill is presented in a smooth form, reducing the transient input change on the control side. The output encapsulation carries the path source semantics and uncertainty description quantity, so that step three can directly map the work criteria, work status and action instructions based on the same semantic system for work index mapping and degradation recovery judgment.

[0103] Step three, the continuous equivalent correction and uncertainty description quantity output by step two are converted into work criteria, work status and action instructions that can be directly used for agricultural work control, and the uncertainty description quantity is kept conservative and traceable under the condition that multiple source corrections coexist and the correlation is unknown.

[0104] In the presence of on-board broadcast correction paths and network differential correction paths in the work site, the terminal precise satellite trajectory and clock difference products of different sources are first unified to the same data semantics, and then the correction inputs for each satellite are screened for compatibility to avoid sending obviously incompatible inputs directly to the fusion calculation, causing work side state jitter, and forming a reviewable screening evidence chain.

[0105] The cellular link in the agricultural work area may be jittered by the terrain, inter-row shading and equipment antenna swing. The correction records of the same satellite by the network differential correction path and the on-board broadcast correction path may be misaligned on the time axis. At the same time, the cross-version bridging path of step two may generate a bridging correction vector during the version transition period and soft inject it with a measurement injection coefficient. If the work side does not distinguish between real measurement update output and bridging soft injection output, it may mix the transition period output with the normal period output, causing confusion in the fusion inlet semantics. Therefore, the same marking and compatibility screening need to be done first to allow the subsequent conservative fusion to be performed between comparable objects.

[0106] The mobile terminal takes one terminal precise satellite orbit and clock error product record from different sources for each satellite, and reads the source identification, corresponding state label, corresponding measurement injection coefficient from the record header field, and whether the record is formed by the participation of the bridging correction vector The terminal takes the same ephemeris output time as the alignment reference, aligns the smooth outputs of different sources according to the reference, and if the alignment time falls between two ephemerides, the time alignment of each component is carried out along the interpolation process of step two to ensure that the compatibility comparison occurs under the same time semantics.

[0107] After alignment, the terminal filters according to the state label priority: when the state label of a source is real interruption, the source does not enter fusion at the current ephemeris; when the state label of a source is version transition and its measurement injection coefficient is in the soft injection interval, the source only participates in compatibility comparison as an alternative input but not as a preferred fusion input; when the state label of a source is version consistent available or late or out of order and is in the backfill window, the source enters the preferred candidate set. Then, the terminal performs compatibility filtering in the preferred candidate set, and the filtering basis is whether the terminal precise satellite orbit difference and clock error difference of different sources of the same satellite present an explainable continuous change, which is determined by whether the difference sequence of adjacent ephemerides is continuous and the magnitude relationship between the difference sequence and the uncertainty description quantity of step two.

[0108] The compatibility filtering takes the terminal precise satellite orbit difference and clock error difference of different sources of the same satellite at the same output time as the discriminant; the discriminant forms a difference sequence over several ephemerides, and if the difference sequence appears sign reversal or amplitude mutation without the support of version transition / late backfill state label, the source is downgraded to an alternative or excluded at the satellite and ephemeris. The difference sequence buffer length is consistent with the fixed time lag window to ensure semantic alignment.

[0109] To make the compatibility filtering implementable and reviewable, the mobile terminal adopts a two-stage filtering chain: first, field-level filtering is used to exclude records with missing fields, unalignable time, and real interruption state label; then, sequence-level filtering is used to exclude records with reverse mutation in a ephemeris and lack of state label support. During sequence-level filtering, the terminal retains the difference sequence buffer of the recent several ephemerides, which is established in satellite granularity and indexed with the same satellite identification as the step one circular buffer to avoid index drift.

[0110] Meanwhile, the field-level filtering can replace the source identification with the transmission link identification, and the sequence-level filtering can replace the difference sequence continuity with the difference sequence monotonicity test, as long as the filtering input is still based on the state label, measurement injection coefficient, and bridge participation identification, and the output is still the candidate source set that can enter fusion, the terminology and object can remain unchanged.

[0111] In use, first, the same label and time alignment are completed according to the source identification, state label, measurement injection coefficient and bridge participation identification, then, the candidate source set that can enter fusion is output according to the two-stage screening chain of field level and sequence level. The fusion entry object is unified in time semantics and source semantics, and the subsequent fusion calculation faces the comparable smooth output , avoiding mixing unalignable records into fusion. The state label and measurement injection coefficient are used as screening evidence chain, so that the version transition and real interruption are presented as different input qualifications on the job side, reducing the misuse of transition output on the job side.

[0112] On the basis of the candidate source set, the multi-source smooth output of the same satellite is conservatively fused, and the uncertainty description quantity is also fused at the same time, avoiding underestimating the uncertainty description quantity when the source correlation is unknown. Unlike the position result layer switching, the fusion is always completed in the correction layer, and the output is still the equivalent correction which can be superimposed on the broadcast ephemeris, and carries the equivalent uncertainty description quantity, which is directly used for subsequent job criterion construction.

[0113] The correction information of different sources may share part of the upstream processing chain or share part of the observation network, and there is an unobservable correlation between their error terms. If the correlation is unknown, the uncertainty description quantity may be compressed to an unreasonable level when fused by conventional weighted summation, which may lead to an optimistic evaluation of the risk by the job criterion.

[0114] In order to make the downgrade and recovery trigger of step three not dependent on the premise that the correlation is known, the covariance intersection idea is used for conservative fusion, and the weight solving is disclosed as a terminal executable single-variable search process, avoiding conceptual description.

[0115] The mobile terminal selects two candidate source records for each satellite as input for fusion, including source one smooth output vector and source one uncertainty description matrix, as well as source two smooth output vector and source two uncertainty description matrix. If the candidate source set is more than two, the terminal fuses them two by two in turn according to the compatibility screening score, forming an iterative fusion chain, so that each step still satisfies the conservative semantics of covariance intersection.

[0116] During the fusion process, the mobile terminal first fuses the uncertainty description matrix to form an equivalent uncertainty description matrix, and then performs information domain synthesis on the smooth output vector with the equivalent uncertainty description matrix, finally obtaining an equivalent correction vector. The covariance intersection type conservative fusion uses weight coefficients to control the proportion of information from two sources, first obtaining the equivalent uncertainty description matrix :

[0117] ;

[0118] where: equivalent uncertainty description matrix : symmetric positive definite matrix used to describe the equivalent correction uncertainty after conservative fusion, taking value from the set of symmetric positive definite matrix; source one uncertainty description matrix : uncertainty description matrix carried by source one record, taking value from the set of symmetric positive definite matrix; source two uncertainty description matrix : uncertainty description matrix carried by source two record, taking value from the set of symmetric positive definite matrix;

[0119] : conservative fusion weight coefficient : scalar coefficient used to control the proportion of two sources information, taking value from the interval , which is used to adjust the dependence of fusion result on two sources under the premise of conservatism;

[0120] After obtaining, the mobile terminal composes the equivalent correction vector in the information domain :

[0121] ;

[0122] In the formula: equivalent correction vector : correction vector output after conservative fusion, containing orbit three-dimensional correction component, clock error correction component and bias correction component included in fusion; source one correction vector : correction vector recorded by source one, formed by extracting correction component from the smoothing output of source one, taking value constrained by satellite orbit scale and clock error scale;

[0123] source two correction vector : correction vector recorded by source two, formed by extracting correction component from the smoothing output of source two, taking value constrained by satellite orbit scale and clock error scale; equivalent uncertainty description matrix : matrix obtained from the previous formula, which is used to map the information domain composition result back to the correction vector domain in this formula.

[0124] source one uncertainty description matrix : same as above, which is used to provide the weight of source one information domain; source two uncertainty description matrix : same as above, which is used to provide the weight of source two information domain; conservative fusion weight coefficient : same as above, which is used to adjust the proportion of two sources information.

[0125] The specific form of weight objective function is:

[0126] ;

[0127] objective function : used to select conservative fusion weight coefficient a scalar function with a value range of real number set, which is used to measure the volume scale of the equivalent uncertainty description matrix; conservative fusion weight coefficient : a value range of ; equivalent uncertainty description matrix : a symmetric positive definite matrix calculated by the covariance intersection conservative fusion formula; The function is the logarithm of the determinant, which improves stability by first calculating the logarithm of the determinant and then taking the exponential.

[0128] The solution of the conservative fusion weight coefficient adopts a one-dimensional search process: the mobile terminal selects a candidate weight sequence in the interval, calculates the logarithm value of the determinant of the equivalent uncertainty description matrix for each candidate weight, and selects the candidate weight that makes the logarithm value minimum, so as to select the fusion result with more concentrated information under the constraint of conservatism. The solving process can use the golden section method or grid search, and the terminal only needs to realize the single variable function evaluation and interval contraction. As an equivalent implementation, the weight target function can be replaced by the trace function, as long as the weight solving is still limited to and the fusion form remains the covariance intersection structure, the conservatism semantics can remain consistent.

[0129] In use, the covariance intersection conservative fusion is used in the candidate source set, the equivalent uncertainty description matrix is obtained by fusion first, and then the equivalent correction vector is synthesized in the information domain, and the is solved by a one-dimensional search process; under the condition that the source correlation is unknown, the equivalent uncertainty description matrix maintains the conservatism semantics, providing an input basis that is not easily underestimated for subsequent job criterion construction; the fusion occurs in the correction layer rather than the result layer, and the job side always receives continuous input of the equivalent correction vector, reducing the semantic break caused by the switching of the result layer.

[0130] The output equivalent correction vector and the equivalent uncertainty description matrix are further mapped to the criterion quantity that can be inspected by the agricultural job, so that the job control does not depend on the fixed interruption time threshold, but depends on the judgment chain whether the predicted upper bound meets the job threshold. The key is to disclose the mapping relationship and the construction method of the upper bound scale, so that the criterion can be reproduced by those skilled in the art.

[0131] The agricultural job closed loop is really concerned about the control risk corresponding to the lateral offset, the heading offset, the trajectory spacing deviation, the upper line distance, and the anti-disturbance endurance time, and these quantities are not directly equivalent to the uncertainty description quantity of the single satellite correction. If the size of the correction uncertainty description quantity is directly used as the degradation trigger basis, the job risk may change due to the change of satellite geometry, but the correction uncertainty description quantity remains unchanged, and the criterion is out of touch with the job risk. Therefore, it is necessary to map the correction uncertainty description quantity to the job index space and construct a comparable predicted upper bound scale.

[0132] The mobile terminal first determines a job index vector, which contains at least three components of lateral deviation, heading deviation, and track spacing deviation, and links these components with the components of the equivalent correction vector through a linearization relationship of an observation equation. The linearization relationship is given by a sensitivity mapping matrix, each column of which corresponds to the influence of a small perturbation of one correction component on each component of the job index vector.

[0133] The sensitivity mapping matrix can be obtained by analytical derivation or numerical differentiation: the terminal applies a perturbation step to each component and recalculates the job index vector, the perturbation step being controlled by a perturbation step coefficient, the value of which is in the range of positive real numbers and is scaled with the dimension of the component to ensure that the numerical differentiation is not dominated by the dimension. After obtaining the sensitivity mapping matrix, the terminal maps the equivalent uncertainty description matrix to a job index uncertainty description matrix, and then constructs a prediction upper bound scale from the matrix and compares it with the job threshold.

[0134] The mapping of the job index uncertainty description matrix adopts a matrix congruence transformation form:

[0135] ;

[0136] In the formula, the job index uncertainty description matrix : a symmetric positive semi-definite matrix used to describe the uncertainty semantics of the job index vector, the value of which is a symmetric positive semi-definite matrix set; the sensitivity mapping matrix : a linearization mapping matrix connecting the equivalent correction vector and the job index vector, the value of which is a real matrix set; the sensitivity mapping matrix is the Jacobian matrix of the job index vector with respect to the equivalent correction vector :

[0137] ;

[0138] The job index vector : contains at least one or more components of lateral deviation, heading deviation, and track spacing deviation; the equivalent correction vector : the correction layer unified input output in step three.

[0139] The equivalent uncertainty description matrix : a symmetric positive definite matrix, the value range of which is a symmetric positive definite matrix set, which provides the correction layer uncertainty semantics as input in the formula.

[0140] After obtaining the job index uncertainty description matrix, the terminal constructs a prediction upper bound scale as a single scalar criterion, and the prediction upper bound scale adopts the maximum eigenvalue operator to take the value:

[0141] ;

[0142] wherein: the predicted upper bound scale : a non-negative scalar used to characterize the scale of the uncertainty envelope of the job metric, taking values is to compress the matrix criterion into a single value criterion that can be directly compared with the job threshold; the maximum eigenvalue operator : the maximum eigenvalue extraction operator for symmetric matrices, inputting a symmetric positive semi-definite matrix and outputting a non-negative real number; the job metric uncertainty description matrix : as above, inputting the maximum eigenvalue operator;

[0143] After the predicted upper bound scale is constructed, the mobile terminal takes the job threshold coefficient as the comparison object, which is formed by the job type configuration: the corresponding threshold is formed for seeding, fertilizing and spraying jobs, and is consistent with the dimension of the job metric vector component.

[0144] The input source of the job metric: the job line (such as a straight job line or a curved job line) is set by the operator or imported from the field job task file; the track spacing parameter is obtained from the machine width / row spacing configuration; the current speed and heading of the vehicle can be provided by the wheel speed sensor and the heading sensor (or the inertial measurement unit); if no external sensor is configured, the speed and heading output by the positioning solution are used as a substitute, and the source type is marked in the job record.

[0145] The terminal further calculates the predicted sustainable job time, which is calculated by discretely extrapolating the kinematics of the vehicle and the geometric changes of the observation within the future prediction window, gradually updating the sensitivity mapping matrix and recalculating the predicted upper bound scale to find the first discrete time when the job threshold coefficient is exceeded. The solution of this first exceeding time can use linear search or bisection method, and the terminal only needs to implement the time stepping-criterion calculation-threshold comparison closed loop.

[0146] As an example: when a spraying machine drives along the long side of the field, the terminal receives the output of the satellite broadcasted and network differential corrected paths at the same time, then the network link appears jitter, and the state label appears late or out of order in adjacent epochs. The terminal first filters the candidate source set that can be fused, then calculates the equivalent correction vector and equivalent uncertainty description matrix , and then calculates the job metric uncertainty description matrix and the predicted upper bound scale . The terminal keeps the job state information continuously updated on the screen, and gives a textual prompt of the sustainable job time within the prediction window, so that the driver can judge whether to continue spraying along the current row or prepare to adjust the job mode at the head.

[0147] The sensitivity mapping matrix is obtained in the following manner: analytical derivation path: partial derivation of satellite position / clock error correction error based on pseudo-range observation equation to form a linear relationship between correction error and ranging error; the ranging error is then mapped to terminal position error through the geometric matrix of positioning solution; finally, the position error is projected to the work line coordinate system to obtain the lateral deviation, heading deviation and other work indicators. Numerical differentiation path: a perturbation step is applied to each component of the equivalent correction vector, and the work indicator vector is calculated repeatedly. The difference between the two is divided by the perturbation step to obtain the corresponding column of the sensitivity mapping matrix. The perturbation step is bound to the component dimension (for example, the position component uses a position perturbation step, and the clock difference component uses a clock difference perturbation step), and the perturbation step is required to still be able to complete a complete work indicator calculation.

[0148] When used, the equivalent uncertainty description matrix is mapped to the work indicator uncertainty description matrix using the sensitivity mapping matrix The maximum eigenvalue operator is used to obtain the predicted upper bound scale, which is compared with the work threshold coefficient to form an executable calculation chain of the predicted sustainable work time.

[0149] The minimum prediction model of the predicted sustainable work time: the vehicle motion in the future prediction window adopts constant speed along the work line or constant speed and constant heading model for discrete extrapolation; when the satellite geometry and the sensitivity mapping matrix are updated over time, the current ephemeris extrapolation of the satellite and the correction uncertainty description matrix growth rule can be used.

[0150] The work criterion is derived from the correction layer uncertainty description quantity. The dimension relationship between the criterion and the lateral deviation, heading deviation and trajectory spacing deviation is explicitly disclosed, which is convenient for reproduction and parameter adjustment. The predicted upper bound scale is compressed into a single value criterion by the matrix criterion, which can be directly used for threshold comparison and time step solution on the work side, avoiding exposing the complex matrix criterion directly to the control side. The work threshold coefficient and the recovery threshold coefficient are given by the work type configuration file, and the recovery threshold coefficient is strictly less than the work threshold coefficient to form a hysteresis. This configuration file can be pre-installed or selected by the operator to load the work mode.

[0151] Based on the predicted upper bound scale and the predicted sustainable work time, an action chain of adaptive degradation and smooth recovery is formed, and the trigger reason, trigger time and corresponding input semantics of the action chain are written into the work record for subsequent verification and delivery. It is emphasized that the action chain is executable, the trigger chain is traceable, and the recovery chain is hysteresis, so as to avoid frequent switching during the work process, which causes difficulty for the operator to understand.

[0152] The agricultural operation control involves human-machine cooperation, and the operator needs to see clear state changes and clear triggering reasons; if the system switches back and forth near the threshold value, the operator will have difficulty in determining when to take over and when to resume, and the controller may repeatedly receive different levels of trajectory input in a short time, causing steering wheel shaking. Therefore, hysteresis logic is needed to make the degradation trigger and the recovery trigger have different criterion boundaries, and the triggering reasons and input semantics are recorded together to facilitate review.

[0153] The mobile terminal describes the operation level with an operation state machine, which at least includes four states: operation maintaining state, degradation preparation state, degradation execution state, and recovery preparation state. The state transition is jointly driven by the predicted upper bound scale , the predicted sustainable operation time, and the state label semantics: when the operation threshold coefficient will be exceeded within the prediction window, the state machine enters the degradation preparation state, the terminal outputs a deceleration instruction or an increased overlap amount instruction to the controller, and prompts the operator on the interface; when the operation threshold coefficient has been exceeded or the predicted sustainable operation time is zero, the state machine enters the degradation execution state, the terminal outputs a low-level navigation trajectory or prompts manual takeover to the controller, and writes the state label, the measurement injection coefficient , whether there is a bridging correction vector, and a summary of the equivalent uncertainty description matrix at this time into the operation record; when the recovery threshold coefficient is returned within the subsequent several epochs and the state label returns to the consistent available version or is late or out of order but within the backfill window, the state machine enters the recovery preparation state, the terminal uses a fixed time lag smooth output as the trajectory input base and gradually removes the degradation action, and finally returns to the operation maintaining state. The recovery threshold coefficient and the operation threshold coefficient take different configuration values, so that the degradation and recovery form a hysteresis boundary to avoid repeated switching near the threshold value.

[0154] The mobile terminal decomposes the action chain into two parallel chains: control side action output and human-machine interface alarm output. The control side action output includes: switching the control target from the operation trajectory to the low-level navigation trajectory, limiting the maximum steering angle change rate, triggering the deceleration gear, and triggering the overlap control strategy of the spray valve; the human-machine interface alarm output includes: displaying the current operation level, displaying the predicted sustainable operation time, and displaying the degradation reason field. The operation record is encapsulated in a structured field writing manner, and the field at least includes: time stamp, satellite identification set summary, state label statistical summary, whether in version transition, whether late backfill occurs, equivalent correction vector summary, equivalent uncertainty description matrix summary, predicted upper bound scale , and trigger action type. As an equivalent implementation, the control side action output can be replaced by an angle limiting instruction, and the human-machine interface alarm output can be replaced by a text prompt, as long as the state machine transition is still driven by , the expected sustainable operation time and the state label semantics are driven, and the operation record field still contains the trigger cause and the input semantics, so that the executability and traceability of the action chain can be maintained.

[0155] As an example: when a rice transplanter enters a soft mud area near a ridge, the body posture change causes the antenna obstruction to increase, the network difference correction path appears to be late and is supplemented, the terminal screen displays the operation state from the operation maintenance state to the degraded preparation state, and prompts the operator to reduce the vehicle speed. When the rice transplanter continues to travel along the edge of the field, the predicted upper bound scale continues to rise, the terminal switches the operation state to the degraded execution state, the controller changes the steering control target to a low-level navigation track and limits the steering wheel change speed, and the operator can see the degradation reason field on the interface and select to continue to slowly advance. Subsequently, the rice transplanter returns to the unobstructed area, the state label returns to the version consistent and available, the predicted upper bound scale decreases to within the recovery threshold coefficient, the terminal enters the recovery preparation state and gradually releases the angle limiting strategy, the interface prompts the operation state to return to the operation maintenance state, and the operation record saves the trigger chain of this degradation and recovery, facilitating post-shift review.

[0156] When used, the predicted upper bound scale , the expected sustainable operation time and the state label semantics drive the hysteresis migration of the operation state machine, and the control side actions and human-machine interface alarms are output in parallel, while the structured field is written into the operation record to solidify the trigger cause and the input semantics. Degradation and recovery are controlled by the hysteresis boundary, the state migration has continuity, and the operator can see the state change and the reason field consistent with the trigger chain from the interface. The operation record solidifies the trigger action and the input semantics, so that subsequent review can determine whether the degradation reason comes from version transition, late backfill or real interruption, providing a basis for delivery and maintenance.

[0157] Those skilled in the art can appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0158] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0159] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the division of the above-described device embodiment is only a logical function division, and there can be another division manner for actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, or the among different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0160] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0161] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A high-precision satellite trajectory processing method for a mobile terminal based on a Kalman filtering algorithm, characterized in that: The application relates to a mobile terminal high-precision satellite trajectory processing method. The method comprises the following steps: receiving broadcast ephemeris and high-precision correction information, writing broadcast ephemeris version identification, high-precision correction information message time stamp, associated version identification and field integrity flag into a correction record database and buffering, forming a satellite-by-satellite correction candidate set and a state label; According to the state label, a correction state filter is established for each satellite, Kalman filtering prediction and updating are adopted, a precise satellite state anchor point and a broadcast ephemeris difference are used to generate a bridging correction vector during version transition, backfilling and updating and forward propagation are performed during delay or disorder, and a fixed time lag smoothing equivalent correction vector and an equivalent uncertainty description matrix are outputted; The equivalent correction vector and the equivalent uncertainty description matrix are used to generate terminal precise satellite trajectory and clock difference products, which are mapped into operation index vectors and prediction upper bounds, compared with operation threshold coefficients and recovery threshold coefficients, and operation state machines are driven to output operation levels and degradation or recovery instructions.

2. The mobile terminal high-precision satellite trajectory processing method according to claim 1, wherein: The correction record database establishes a circular buffer according to the satellite number and a time index according to the message time stamp, inserts the corresponding position according to the time index when receiving the high-precision correction information, writes the broadcast ephemeris version identification, the associated version identification, the field integrity flag, the delay time length and the correction category into the corresponding correction record, and eliminates the correction records exceeding the buffer window according to time.

3. The mobile terminal high-precision satellite trajectory processing method according to claim 2, wherein: At each epoch, a correction record is selected from the satellite-by-satellite correction candidate set for each satellite and a state label is generated, wherein the version consistency is that the associated version identification is consistent with the broadcast ephemeris version identification and the field is complete, the version transition is that the broadcast ephemeris version identification has been updated and the correction record corresponds to the old version, the delay or disorder is that the message time stamp is behind the current epoch and is within the backfilling window, and the real interruption is that there is no available correction record for a plurality of continuous epochs.

4. The mobile terminal high-precision satellite trajectory processing method according to claim 3, wherein: The judgment of the version transition is based on the version difference, and the version difference is determined according to the minimum difference of the broadcast ephemeris version identification and the associated version identification; when the associated version identification is missing, the corresponding correction record is marked as version identification missing and the priority of the correction record in the satellite-by-satellite correction candidate set is reduced.

5. The mobile terminal high-precision satellite trajectory processing method according to claim 4, wherein: The state quantity of the correction state filter comprises orbit three-way correction, clock correction and its drift term, the correction record is used as the measurement input to perform updating when the version consistency is available, only prediction is performed and the equivalent uncertainty description matrix is expanded when the real interruption occurs, and the measurement injection coefficient is introduced in the updating process according to the state label to adjust the measurement updating strength.

6. The mobile terminal high-precision satellite trajectory processing method according to claim 5, wherein: The precise satellite state anchor point is determined by the terminal precise satellite trajectory and clock difference product at the last epoch, the bridging correction vector is formed by the difference between the satellite position and clock difference obtained by the precise satellite state anchor point and the broadcast ephemeris propagation, and the position difference is projected into radial, tangential and normal unit vectors to generate orbit three-way bridging correction.

7. The mobile terminal high-precision satellite trajectory processing method according to claim 6, characterized in that: The correction state filter history state is saved in the cache window of the correction record library when late or out-of-order, and the history epoch is located according to the message timestamp to perform backfilling update after receiving the late correction record, and is propagated forward to the current epoch according to the recorded state label.

8. The mobile terminal high-precision satellite trajectory processing method according to claim 7, characterized in that: When the terminal receives multi-source corrections, the terminal first performs compatibility screening based on the difference sequence of the terminal precise satellite trajectory and clock error product, then uses covariance intersection to obtain the fusion result of the equivalent correction vector and equivalent uncertainty description matrix that pass the screening, and determines the intersection weight by one-dimensional search and performs positive definite processing on the fusion matrix.

9. The mobile terminal high-precision satellite trajectory processing method according to claim 8, characterized in that: The operation index vector is obtained based on the terminal precise satellite trajectory and clock error product in the operation line coordinate system, the predicted upper bound is determined by the operation uncertainty obtained by sensitivity mapping of the equivalent uncertainty description matrix, and the maximum eigenvalue of the operation uncertainty is obtained by power iteration as the predicted upper bound, wherein the sensitivity mapping is obtained by perturbing each component of the equivalent correction vector by a preset perturbation step and recalculating the operation index vector.

10. The mobile terminal high-precision satellite trajectory processing method according to claim 9, characterized in that: The operation state machine sets four states: keep, downgrade preparation, downgrade execution, and recovery preparation, the state transition is triggered by comparing the predicted upper bound with the operation threshold coefficient and the recovery threshold coefficient, and the transition is executed after a continuous number of epochs satisfy the same triggering condition; The operation state machine outputs the operation level, the predicted sustainable operation time, and the instructions corresponding to the current state, the instructions include downgrade instructions and recovery instructions and are written into the operation log.

Images