Edge computing control system and method for total station

Abstract

The application belongs to the technical field of automatic control and edge computing control of measuring equipment, and particularly relates to an edge computing control system and method for a total station. The system comprises an edge computing control end, a servo pointing controller, a ranging trigger controller and an observation confirmation processor which interact with the edge computing control end. The edge computing control end establishes an observation confidence state space, receives a target locking state, angle encoding stability, servo turntable tracking deviation, ranging echo quality and measurement task geometric constraints, generates an observation confidence vector, and outputs at least one of fine adjustment search, ranging suppression, ranging trigger, repeated observation or recapture control instructions according to a physical constraint, confidence decay correction and control instruction mapping relationship. Thus, the aiming, ranging and observation confirmation form a closed-loop control locally in the total station, the mismatch between the angle state and the distance state caused by the fixed time sequence trigger is reduced, and the consistency and traceability of the automatic continuous observation data are improved.

Description

Technical Field

[0001] This invention belongs to the field of automatic control and edge computing control technology of measuring equipment, specifically an edge computing control system and method for total stations. Background Technology

[0002] Total stations typically employ a combination of functions including angle measurement, distance measurement, servo pointing, automatic aiming, and data processing to collaboratively observe target points. In conventional automatic measurement control, the task terminal sends the target point coordinates or observation sequence to the total station. The total station calculates the target azimuth based on the station orientation results, rotates to the vicinity of predetermined horizontal and vertical angles via a servo turntable, and then uses the automatic aiming function to capture the prism or reflecting target. Subsequently, after a preset waiting time, distance measurement is triggered, and the angle value, distance value, and observation time are written into the measurement record. Some devices perform data filtering based on target lock status, distance echo intensity, or repeated measurement difference, but this filtering usually occurs after distance measurement is completed, mainly used to mark abnormal data, prompt remeasurement, or delegate to backend software processing. Existing edge control schemes often use the total station's local calculations for task caching, data preprocessing, or simple threshold judgment. Aiming control, distance triggering, and observation confirmation still operate according to a fixed time sequence, failing to unify target lock status, angle encoding stability, servo turntable tracking deviation, distance echo quality, and measurement task geometric constraints into a unified real-time control basis.

[0003] The main technical problem with the aforementioned conventional techniques lies in the lack of a unified closed loop between observation quality judgment and total station control actions. This leads to potential time mismatches between angle acquisition, target locking, and distance measurement triggering states. Even when target reflection is unstable, the turntable is not fully stable, angle encoding exhibits short-term fluctuations, or distance measurement echo quality deteriorates, the total station may still trigger distance measurement according to a fixed waiting time. This results in distance values ​​corresponding to unreliable aiming states, or angle values ​​corresponding to distance measurement times that have deviated from the optimal locking state. Since abnormal states are typically identified only after observation results are generated, subsequent remediation can only be achieved by deleting data, manual remeasurement, or increasing the number of repeated measurements. It is impossible to select appropriate methods during the current observation based on the specific source of the anomaly, such as fine-tuning search, distance suppression, extending distance measurement, repeating observations, or re-acquiring. Consequently, although the total station's edge side possesses data processing capabilities, it fails to apply the observation confidence state inversely to the servo pointing and distance measurement triggering processes, making it difficult to guarantee that angle and distance data are in a consistent and stable observation state during automatic continuous observation. Summary of the Invention

[0004] The purpose of this invention is to provide an edge computing control system and method for total stations, which can solve the problems in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An edge computing control system for a total station includes an edge computing control terminal and a servo pointing controller, a distance triggering controller, and an observation confirmation processor that interact with the edge computing control terminal respectively. The edge computing control terminal is configured with an observation confidence state space, which receives target locking status, angle encoding stability, servo turntable tracking deviation, ranging echo quality and measurement task geometric constraints, and generates an observation confidence vector. The edge computing control terminal outputs at least one of the following mutually constrained control commands to the servo pointing controller, the ranging trigger controller, and the observation confirmation processor based on the observation confidence vector: fine-tuning search, ranging suppression, ranging trigger, repeated observation, or re-acquisition control command.

[0006] Preferably, the observation confidence state space includes time-aligned caching, feature normalization processing, and confidence vector generation processing; The time-aligned buffer establishes observation segments according to the aiming time, angle-encoded sampling time, and ranging echo sampling time of the same observation target; The feature normalization process maps the horizontal angle fluctuation, vertical angle fluctuation, servo turntable stability duration, target lock duration, range echo intensity variation, and continuous range dispersion to a unified dimension. The confidence vector generation process then uses the features under the unified dimension to form an observation confidence vector.

[0007] Preferably, the edge computing control terminal is provided with a physical constraint layer and a learning correction layer; The physical constraint layer stores the hard constraint conditions corresponding to the total station's angular resolution, distance measurement triggering conditions, servo turntable stability boundary, and task geometric boundary. The learning correction layer forms a confidence decay coefficient based on historical valid observation segments and historical invalid observation segments, and saves the confidence decay coefficient according to the target distance interval, horizontal turning angle interval and vertical angle interval; The edge computing control terminal performs joint discrimination on the observation confidence vector based on the hard constraint conditions and the confidence decay coefficient.

[0008] Preferably, the edge computing control terminal is equipped with a control command mapping table; The control command mapping table records the combination relationship between the angle stability component, target lock component, echo quality component and range dispersion component in the observation confidence vector, and maps the combination relationship into at least one control command among the following according to the observation stages before ranging, during ranging and after ranging: servo turntable damping wait, small angle micro-scan, extended ranging integration, ranging trigger permission, repeated observation confirmation and local re-acquisition. The control command mapping table also stores the command sequence for the same observation target in consecutive observation segments.

[0009] Preferably, the time alignment cache includes observation segment delimitation processing and segment validity marking processing; The observation segment demarcation process takes the moment when the servo turntable enters the target locking range as the segment start point and the moment when the ranging trigger is completed or the target locking is interrupted as the segment end point. The segment validity marking process generates sample markers based on the continuity of angle encoding, target locking, echo sampling integrity, and ranging trigger record within the observation segment. The confidence vector generation process selects observation segments to participate in the calculation based on the sample markers.

[0010] Preferably, the learning correction layer includes residual backfilling processing and partition decay update processing; The residual backfilling process associates the observation residuals confirmed by the observation confirmation processor with the target distance interval, horizontal angle interval, vertical angle interval, and echo quality interval. The partitioned attenuation update process updates the confidence attenuation coefficient within the range defined by the hard constraint condition, and configures update timing identifiers for different intervals. The edge computing control terminal corrects the ranging triggering condition of subsequent observed targets based on the updated confidence attenuation coefficient.

[0011] Preferably, the control command mapping table is configured with a conflict resolution sequence, which prioritizes outputting ranging suppression and servo turntable damping waiting commands when the angle stability component is below the angle stability threshold and the echo quality component meets the echo threshold. When the angle stability component meets the angle stability threshold and the echo quality component is lower than the echo threshold, a small-amplitude angle micro-scan or extended ranging integration command is output first. When both the target locking component and the rangefinding component are abnormal, a local re-acquisition command is output.

[0012] Preferably, the partition decay update process includes neighborhood inheritance processing; The neighborhood inheritance process allocates the confidence decay coefficients of confirmed observation segments in adjacent horizontal turning intervals, adjacent vertical angle intervals, and adjacent target distance intervals within the same station to the insufficient sample intervals according to the number of observation segments, residual stability, interval adjacency relationship, and update time sequence identifier. The insufficient sample interval is the interval in which the number of historical valid observation segments is less than the preset number of segments. The learning correction layer writes the allocated confidence decay coefficient into the observation confidence state space after verifying the consistency between the confidence decay coefficient and the hard constraint conditions.

[0013] Preferably, the edge computing control terminal is equipped with control trajectory recording processing; The control trajectory recording process records the observation confidence vector, the control command selected by the conflict resolution sequence, the triggering state of the ranging trigger controller, the execution state of the servo pointing controller, and the confirmation result of the observation confirmation processor according to the observation target identifier, and forms a continuous record according to the time sequence of the observation segments; The continuous records are bound to the same observation target identifier and stored, and serve as input conditions for subsequent local recapture, repeated observation confirmation, and ranging trigger permission.

[0014] The edge computing control method for total station includes: receiving target lock status, angle encoding stability, servo turntable tracking deviation, distance echo quality and measurement task geometric constraints from the edge computing control terminal; An observation confidence state space is established within the edge computing control terminal, and an observation confidence vector is generated based on the target locking state, the angle encoding stability, the servo turntable tracking deviation, the ranging echo quality, and the geometric constraints of the measurement task. Based on the observation confidence vector, generate at least one of the following mutually constrained control commands: fine-tuning search, ranging suppression, ranging triggering, repeated observation, or reacquisition. The control commands are sent to the servo pointing controller, the ranging trigger controller, and the observation confirmation processor, respectively.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By setting an observation confidence state space at the edge computing control terminal of the total station, the target lock state, angle encoding stability, servo turntable tracking deviation, distance echo quality, and geometric constraints of the measurement task are jointly used to generate an observation confidence vector. Based on this observation confidence vector, mutually constraining control commands are output to the servo pointing controller, distance trigger controller, and observation confirmation processor, transforming the observation quality judgment from post-selection to real-time control conditions before and during distance measurement. When the angle stability component is insufficient, the edge computing control terminal can suppress distance triggering and maintain servo turntable damping waiting; when the echo quality component is insufficient but the angle state meets the requirements, the edge computing control terminal can switch to small-amplitude angle micro-scanning or extend the distance integration; when both the target lock component and the distance dispersion component are abnormal, the edge computing control terminal can switch to local re-acquisition. After the observation confidence vector and control commands are mapped and coordinated, unreliable states are incorporated into the control chain before the final observation value is generated, reducing observation records where the source states of angle and distance values ​​are inconsistent.

[0016] 2. Through time alignment caching, physical constraint layer, learning correction layer, residual backfilling processing, partitioned attenuation update processing, and control trajectory recording processing, traceable observation segments are formed for the aiming time, angle encoding sampling time, and distance echo sampling time of the same observation target. The confidence attenuation coefficient is updated under the constraints of total station angular resolution, distance triggering conditions, servo turntable stability boundaries, and task geometry boundaries. The observation residuals confirmed by the observation confirmation processor are associated with the target distance interval, horizontal angle interval, vertical angle interval, and echo quality interval, allowing for the correction of distance triggering conditions under subsequent similar observation conditions. The control trajectory recording processing binds and stores the observation confidence vector, the control command selected for the conflict resolution sequence, the distance triggering state, the servo execution state, and the confirmation results in the order of the observation segments, providing continuous input conditions for repeated observation confirmation, local re-acquisition, and distance trigger permission. This improves the problems of control state breaks and insufficient basis for anomaly handling during long-term automatic observations by the total station. Attached Figure Description

[0017] Figure 1 This is the overall closed-loop control flowchart for the edge computing control system used in the total station; Figure 2 A flowchart for generating observation confidence vectors for the observation confidence state space; Figure 3 The flowchart shows the joint discrimination and residual backfilling update of the physical constraint layer and the learning correction layer. Figure 4 This is a flowchart for processing control command mapping tables, conflict resolution sequences, and control trajectory records. Detailed Implementation

[0018] In one embodiment, reference Figure 1This invention provides an edge computing control system for a total station, including an edge computing control terminal and a servo pointing controller, a distance triggering controller, and an observation confirmation processor that interact with the edge computing control terminal. The edge computing control terminal is configured with an observation confidence state space, which receives target lock status, angle encoding stability, servo turntable tracking deviation, distance echo quality, and geometric constraints of the measurement task, and generates an observation confidence vector. Based on the observation confidence vector, the edge computing control terminal outputs at least one of the following mutually constrained control commands to the servo pointing controller, the distance triggering controller, and the observation confirmation processor: fine-tuning search, distance suppression, distance triggering, repeated observation, or reacquisition. In specific implementation, the edge computing control terminal is located in the local control link of the total station. The target lock status is formed by the continuous target lock markers and lock deviation during automatic aiming. The angle encoding stability is formed by short-term fluctuations in the horizontal and vertical angle encoding sequences. The servo turntable tracking deviation is determined by the control target azimuth and the actual turntable feedback azimuth. The difference between the two is formed. The ranging echo quality is formed by the variation of ranging echo intensity and the consistency of continuous ranging. The geometric constraints of the measurement task are formed by the station orientation, the predicted azimuth of the target point, the range of the target point distance, and the geometric relationship allowed by the observation task. The edge computing control end aligns the above information within the same observation segment and converts it into an observation confidence vector. Before ranging is triggered, it determines whether the current observation state is suitable for forming a measurement value. When the angle encoding stability is insufficient, the ranging trigger controller does not execute the ranging action and the servo pointing controller maintains damping waiting. When the echo quality is insufficient but the angle state meets the requirements, the servo pointing controller performs fine-tuning search or the ranging trigger controller extends the integral trigger condition. When neither the locked state nor the ranging dispersion state meets the requirements, the observation confirmation processor marks the current observation segment as a re-capture segment and causes the edge computing control end to output a re-capture control command. The advantage of this embodiment is that aiming, ranging, and observation confirmation are in the same edge closed-loop control link, avoiding the observation record of inconsistent angle state and distance state caused by fixed timing trigger ranging.

[0019] In this embodiment, the observation confidence state space uses observation segments as the smallest processing object. Each observation segment corresponds to a continuous data interval of the same target from entering the locking range to the completion of ranging trigger or the interruption of locking. The edge computing control terminal merges the target locking status, angle encoding stability, servo turntable tracking deviation, ranging echo quality, and geometric constraints of the measurement task using time stamps. During merging, the last sampled value is not directly used as the judgment criterion. Instead, a multi-source sampling time difference is formed within the observation segment. The locking sampling time, angle encoding sampling time, echo sampling time, and ranging trigger time of the same target constitute a time consistency condition. Let the locking sampling time of the k-th observation segment be... The sampling time for the combined horizontal and vertical angles is... The echo sampling time is The ranging trigger time is The time mismatch amount is determined by the following formula: ; in, This represents the maximum time deviation of different observation sources relative to the target lock state within the k-th observation segment. , , and All use the same time base within the edge computing control terminal, if and similar, and Similar and If the delay in the locked state is too long, then If the value is small, the edge computing control terminal allows the observation segment to continue into the confidence vector generation process. If target locking has been interrupted and only occurs after echo sampling or ranging is triggered, then... As the size increases, the edge computing control terminal excludes the observation segment from the ranging trigger permission. The advantage of this embodiment is that it uses time consistency to constrain the locking, angle, echo and trigger data in the same observation state, providing a reliable input for subsequent closed-loop control.

[0020] Preferred, Reference Figure 2The observation confidence state space includes a time alignment buffer, feature normalization processing, and confidence vector generation processing. The time alignment buffer establishes observation segments according to the aiming time, angle encoding sampling time, and ranging echo sampling time of the same observation target. The feature normalization processing maps horizontal angle fluctuation, vertical angle fluctuation, servo turntable stability duration, target lock continuity, ranging echo intensity change, and continuous ranging dispersion to a unified dimension. The confidence vector generation processing composes the features under the unified dimension into an observation confidence vector. Specifically, the time alignment buffer receives turntable feedback data from the servo pointing controller, target lock data from the automatic aiming process, echo sampling data from the ranging trigger controller, and observation result tags from the observation confirmation processor, and writes data of the same target into the same buffer queue according to the target identifier. The feature normalization processing does not change the physical properties of the original sampled values. Instead of defining the meaning, this method normalizes each feature based on its directionality in total station measurement control. This results in higher values ​​for the angle stability component (smaller fluctuations), the target lock component (more complete lock-on duration), the echo quality component (smaller echo mutations), and the distance dispersion component (smaller continuous distance dispersion). The confidence vector generation process saves each component as a vector structure that can be read by the control command mapping table. The observation confidence vector includes at least the lock-on component, angle stability component, servo tracking component, echo quality component, distance dispersion component, and geometric constraint component. The value range of each component is unified to a normalized interval from low confidence to high confidence, facilitating comparison of the states of data from different sources within the same judgment plane at the edge computing control end. The advantage of this embodiment is that it avoids distortion of control criteria caused by direct comparison of data with different dimensions, allowing the total station to select control actions based on a unified confidence state. The spatial variable configuration for the observation confidence state is shown in Table 1 below.

[0021] Table 1. Observation Confidence State Space Variable Configuration Table

[0022] In a preferred embodiment, the variable configuration in Table 1 is used to support the edge computing control terminal in forming an executable observation confidence vector. The target locking continuity value in the table is used to determine whether the total station's automatic aiming process has achieved stable locking; the horizontal and vertical angle fluctuation values ​​are used to determine whether the angle encoding sequence is still in the convergence process; the servo turntable stability duration value is used to determine whether the turntable has achieved stable holding; the ranging echo intensity change value is used to determine whether the ranging optical path state has undergone a short-term abrupt change; the continuous ranging dispersion value is used to determine whether multiple ranging candidate values ​​have homology; and the geometric constraint conformance value is used to determine whether the current observed target matches the station and target geometric relationship given in the task. When reading the configuration in Table 1, the edge computing control terminal does not determine the final control command based on a single variable. Instead, it puts all variables into the same observation segment and makes judgments based on the three stages of pre-distance measurement, during distance measurement, and post-distance measurement. For example, in the pre-distance measurement stage, the angle fluctuation and turntable stability duration take priority in participating in the distance measurement trigger permission. In the during distance measurement stage, the echo intensity change and continuous distance dispersion take priority in participating in the integral extension or repeated observation confirmation. In the post-distance measurement stage, the target locking continuity and geometric constraint compliance take priority in participating in the observation segment validity marking. The advantage of this embodiment is that the observation confidence state space is no longer an abstract scoring, but an engineering processing structure that corresponds to the actual control stages of the total station.

[0023] Furthermore, the edge computing control terminal performs comprehensive calculations on the normalized components, assuming the target locking component of the k-th observation segment is... , Angle stability component is Servo tracking components are The echo quality component is The distance dispersion component is Geometric constraint components are The corresponding weights are respectively , , , , , Each weight takes a non-negative value and their sum is 1. The overall observation confidence value is determined according to the following formula: ; in, This represents the overall observation confidence value for the k-th observation segment. Indicates the degree of continuity of target lock within the observation segment. This indicates the stability of horizontal and vertical angles within the observed segment. This indicates the tracking status between the actual position of the servo turntable and the target position. Indicates the quality of the ranging echo. This indicates the consistency of consecutive ranging candidate values. Indicates the degree of agreement between the geometric relationships of the observation tasks. , , , , , This is used to reflect the intensity of participation of each component in the current observation stage. If a higher intensity is used in the pre-distance measurement stage... and The edge computing control unit will maintain ranging suppression for segments with unstable angles. If a higher [value] is used in the ranging stage, [this will help]. and The edge computing control unit will perform extended integration or repeated observations on segments with abrupt echo changes or discrete ranging candidate values. If all components are within the high confidence interval, then... The advantage of this embodiment is that it raises the range trigger controller to enter the range permission state, and uses the same calculation structure to transform the multi-source observation state into the control decision input, making the source of the control command clear and traceable.

[0024] In one embodiment, reference Figure 3 The edge computing control terminal is equipped with a physical constraint layer and a learning correction layer. The physical constraint layer stores hard constraints corresponding to the total station's angular resolution, distance measurement triggering conditions, servo turntable stability boundaries, and task geometric boundaries. The learning correction layer generates confidence attenuation coefficients based on historical valid and invalid observation segments and saves these coefficients according to target distance intervals, horizontal angle intervals, and vertical angle intervals. The edge computing control terminal performs joint discrimination on the observation confidence vector based on the hard constraints and the confidence attenuation coefficients. In specific implementation, the physical constraint layer ensures that the observation confidence vector does not exceed the basic boundaries of the total station's measurement control. For example, if the angle encoding continuity does not meet the angular resolution constraint, the learning correction layer must not adjust the observation segment to a distance measurement permissible state based on historical experience. If the distance measurement echo does not meet the triggering conditions, the confidence attenuation layer will be adjusted accordingly. The coefficient reduction can only decrease the trigger probability but cannot relax the hard constraints. The servo turntable stability boundary is used to constrain the turntable's transition from motion to observation. The task geometry boundary is used to exclude observation segments that are significantly inconsistent with the station orientation and target prediction azimuth. The learning correction layer is used to handle local differences within the hard constraint range. By comparing the differences between historical valid observation segments and historical invalid observation segments in terms of target distance, turning path, elevation direction, and echo state, confidence attenuation coefficients are formed for different zones. The edge computing control end uses the hard constraint conditions as the upper boundary and the confidence attenuation coefficients as local correction quantities, so that the observation confidence vector both obeys the instrument control mechanism and reflects the local observation patterns within the current station and the current task cycle. The advantage of this embodiment is that it avoids the simple learning correction from deviating from the physical constraints of the total station and also avoids the inability of fixed thresholds to adapt to local changes in different observation intervals.

[0025] Preferably, the time alignment cache includes observation segment demarcation processing and segment validity marking processing. The observation segment demarcation processing takes the moment when the servo turntable enters the target locking range as the segment start point and the moment when ranging triggering is completed or target locking is interrupted as the segment end point. The segment validity marking processing generates sample marks based on the angle encoding continuity, target locking continuity, echo sampling integrity, and ranging trigger record within the observation segment. The confidence vector generation processing selects observation segments to participate in the calculation based on the sample marks. Specifically, when the servo pointing controller drives the total station turntable to the vicinity of the target predicted azimuth and the automatic aiming process generates continuous locking marks, the observation segment demarcation processing establishes the segment start point, and then continuously writes the horizontal angle encoding sequence, vertical angle encoding sequence, and turntable follow-up... The tracking deviation, echo sampling sequence, and ranging trigger status are considered. When the ranging trigger controller completes a ranging measurement and returns a candidate ranging result, or when the automatic aiming process experiences a locking interruption, the observation segment delimitation process closes the current segment, and the segment validity marking process marks the closed segment. If there is a gap in the angle encoding sequence, the target locking record is discontinuous, the echo sampling is missing, or the trigger record is incomplete, the segment will not directly enter the comprehensive confidence value calculation, but will be used as an invalid observation segment for the learning correction layer to identify abnormal patterns. If the above records are complete and meet the hard constraint conditions of the physical constraint layer, the segment will enter the confidence vector generation process and generate an observation confidence vector. The advantage of this embodiment is that it divides the continuous measurement process into segments with clear boundaries, so that abnormal samples and usable samples can be distinguished before the confidence vector is generated.

[0026] In this embodiment, the stable angular component is determined by the short-time discreteness of the horizontal and vertical angle coding sequences. Let the stability statistic of the horizontal angle coding sequence within the k-th observation segment be... The stability statistics of the vertical angle coded sequence are The allowable boundary for angle stability is The angular stability component is determined by the following formula: ; in, This represents the stable angular component of the k-th observation segment. This represents the discreteness of the horizontal angle coded sequence relative to the stationary center within the observed segment. This represents the discreteness of the vertical angle coded sequence relative to the stationary center within the observed segment. This represents the allowable boundary for angular stability given by the physical constraint layer, when Less than hour, As the angle fluctuation decreases, it increases; when Reaching or exceeding hour, The range is suppressed to a low confidence level and triggers ranging suppression or damping wait. For example, in a certain observation segment, the horizontal and vertical angle encoding sequences still fluctuate as the turntable slowly converges, with the numerator approaching or exceeding the denominator. The edge computing control end determines that the segment is not stable in angle, and the ranging trigger controller does not generate permission. In another observation segment, both types of angle encoding sequences oscillate slightly around the stable center, with the numerator much smaller than the denominator. The edge computing control end allows the segment to continue to combine the echo mass component for ranging judgment. The advantage of this embodiment is that it converts the angle stability into a computable component, so that the servo turntable will not be mistakenly triggered for ranging when the fixed waiting time is reached before it is stable.

[0027] Furthermore, the learning correction layer includes residual backfilling and partitioned attenuation update processing. The residual backfilling process associates the observation residuals confirmed by the observation confirmation processor with the target distance interval, horizontal angle interval, vertical angle interval, and echo quality interval. The partitioned attenuation update processing updates the confidence attenuation coefficient within the range defined by the hard constraints and configures update timing identifiers for different intervals. The edge computing control terminal corrects the ranging triggering conditions for subsequent observation targets based on the updated confidence attenuation coefficients. Specifically, after obtaining the ranging candidate values ​​and angle candidate values, the observation confirmation processor generates confirmation results based on the consistency of repeated observations of the same target, the task geometry, and the fragment integrity. The residual backfilling process writes the confirmed observation residuals into the corresponding intervals. The target distance interval represents the distance between the target and the station. The distance range is defined as follows: the horizontal angle range represents the range of horizontal angles from the current direction to the target direction; the vertical angle range represents the range of target elevation direction; and the echo quality range represents the stable range of echo sampling records. The partitioned attenuation update process statistically analyzes the distribution of valid and invalid observation segments within the same interval. If the confirmed residuals within a certain interval are concentrated and the echo is stable, the confidence attenuation coefficient is adjusted towards a lower degree of conservatism. If segments within a certain interval frequently experience locking interruptions, distance measurement dispersion, or geometric inconsistencies, the confidence attenuation coefficient is adjusted towards a higher degree of conservatism. The update timing identifier is used to distinguish between recently observed segments and earlier observed segments, avoiding the long-term influence of old interval states on subsequent control. The advantage of this embodiment is that it enables the observation confirmation results to flow back to the distance measurement triggering conditions, forming a local closed loop between observation data and control conditions.

[0028] In a preferred embodiment, the confidence decay coefficient can be updated in a partitioned recursive manner, where the confidence decay coefficient of the q-th partition before the nth update is . The residual normalization value used in the nth update is The updated participation coefficient is The confidence decay coefficient after the (n+1)th update is determined by the following formula: ; in, This indicates that partition q is used as the decay coefficient to correct the observation confidence vector before the update. This represents the updated attenuation coefficient. This indicates the degree of deviation of the observation residuals after normalization. This indicates the percentage of the original partition status affected by this update. This indicates that the larger the residual, the lower the confidence level of the transformation relationship can be inherited. If the residual of a newly confirmed observation segment within a certain partition is small... The values ​​are close to low, and the fractional terms are close to high. (Updated) The tendency is to allow easier access to distance measurement clearance; if the residuals of newly confirmed observation segments within a certain zone are large, the fractional term decreases, and the updated... The advantage of this embodiment is that it publicly teaches how the correction layer changes the confidence state based on the residuals in a reproducible recursive manner, rather than simply describing historical data to participate in control, by reducing and making the ranging trigger conditions of subsequent similar partitions more conservative.

[0029] In one embodiment, reference Figure 4 The edge computing control terminal is equipped with a control command mapping table. This table records the combination relationships between the angle stability component, target lock component, echo quality component, and range dispersion component in the observation confidence vector. These combinations are mapped according to the observation stages before, during, and after ranging to at least one control command among servo turntable damping wait, small-amplitude angle micro-scan, extended ranging integration, ranging trigger permission, repeated observation confirmation, and local re-acquisition. The control command mapping table also stores the command sequence for the same observation target in consecutive observation segments. In specific implementation, the control command mapping table uses the observation stage as the upper-level index and the confidence component combination relationship as the lower-level index. The pre-ranging stage mainly reads the angle stability component, target lock component, and servo tracking component; the mid-ranging stage mainly reads... The echo quality component and range dispersion component are used in the post-range measurement stage. The main data reads include the range dispersion component, geometric constraint component, and observation confirmation results. When the same target transitions from damped waiting to range clearance in consecutive observation segments, the control command mapping table saves this connection sequence and the observation confirmation processor processes the two segments as part of the same acquisition process. When the same target fails to achieve echo quality improvement after continuous fine-tuning searches, the control command mapping table switches to local reacquisition instead of infinitely repeating fine-tuning. When range clearance has been issued but the echo quality changes abruptly during ranging, the ranging trigger controller cancels the current ranging result, and the observation confirmation processor marks it as a repeated observation segment. The advantage of this embodiment is that it establishes executable connections between different control commands, avoiding isolated commands output by the edge control end based on a single judgment. The staged configuration of the control command mapping table is shown in Table 2 below.

[0030] Table 2. Staged Configuration Table of Control Command Mapping

[0031] Preferably, the staged configuration in Table 2 is used to express the specific execution method of the control command mapping table. Before ranging, the edge computing control terminal reads the angle stability component and the servo tracking component. If the turntable has not yet entered a stable state but the target lock is still continuous, ranging is not triggered and damping is maintained. If the target lock component decreases but the turntable azimuth is close to the target prediction azimuth, a small angle micro-scan is used to try to restore the lock. If the angle stability component and the target lock component are both in the permissible range, a ranging trigger permission is issued to the ranging trigger controller. The echo quality component and the ranging dispersion component are read during ranging. If the echo quality component is insufficient but the target lock is still maintained, the ranging integration is extended. If the echo changes abruptly and the ranging candidate value is discrete, repeated observation confirmation is entered. After ranging, the target lock component and the ranging dispersion component are combined to determine whether to transfer the current target to local re-acquisition. The control command history is read in the continuous segment. If the same target has repeatedly exhibited abnormalities, re-acquisition takes priority over continuing repeated ranging. The advantage of this embodiment is that the total station control action changes with the observation stage, avoiding mechanically applying the same threshold judgment to the entire measurement process.

[0032] Furthermore, the control command mapping table is configured with a conflict resolution sequence. When the angular stability component is below the angular stability threshold and the echo quality component meets the echo threshold, the conflict resolution sequence prioritizes outputting ranging suppression and servo turntable damping waiting commands. When the angular stability component meets the angular stability threshold and the echo quality component is below the echo threshold, it prioritizes outputting small-amplitude angular micro-scan or extended ranging integration commands. When both the target locking component and the ranging dispersion component are abnormal, it outputs a local re-acquisition command. In specific implementation, the conflict resolution sequence is used to handle situations where different components in the observation confidence vector give opposite control tendencies. For example, when the echo quality has reached the permissible range but the angular stability component is still below the angular stability threshold, the ranging trigger controller will not automatically adjust its control based on the better echo status. Instead of immediately triggering ranging, the edge computing control terminal prioritizes maintaining ranging suppression and sends a damping wait command to the servo pointing controller to avoid unreliable distance records with unreliable angle states. For example, if the angle stable component meets the permissible conditions but the echo quality component is below the echo threshold, the edge computing control terminal does not require the turntable to perform a large-scale search again. Instead, it selects a small-amplitude angle micro-scan or extends the ranging integral while maintaining the lock, so that the control action is locally corrected around the current lock state. Furthermore, if the target lock component and the ranging dispersion component are abnormal at the same time, the conflict resolution sequence determines whether to continue ranging or if the local micro-scan cannot provide a usable segment, and directly outputs a local re-acquisition command. The advantage of this embodiment is that it transforms conflict states into clear control priorities, so that multi-source states will not produce mutually conflicting control outputs.

[0033] In a preferred embodiment, the partitioned attenuation update process includes a neighborhood inheritance process. This process allocates the confidence attenuation coefficients of confirmed observation segments within adjacent horizontal turning intervals, adjacent vertical turning intervals, and adjacent target distance intervals of the same station to insufficient sample intervals based on the number of observation segments, residual stability, interval adjacency, and update sequence identifier. The insufficient sample interval is defined as an interval where the number of historically valid observation segments is less than a preset number. The learning correction layer performs a consistency check between the allocated confidence attenuation coefficients and the hard constraint conditions before writing them into the observation confidence state space. In practice, adjacent azimuths and adjacent distances at the same station typically have partially similar aiming and ranging states. However, the insufficient sample interval cannot be directly replaced by the global default coefficient; otherwise, the ranging triggering conditions will become discontinuous between adjacent targets. Continuing the process, the neighborhood inheritance process selects the horizontal turning angle interval, vertical angle interval, and target distance interval adjacent to the insufficient sample interval. It reads the number of segments, residual stability, and most recent update time of these neighborhood intervals as confirmed by the observation confirmation processor. Neighborhood intervals with more segments and more stable residuals have higher allocation weights to the insufficient sample interval. Neighborhood intervals with more recent updates participate in the allocation first. After the allocation is completed, the learning correction layer compares the inherited confidence attenuation coefficient with the hard constraint conditions in the physical constraint layer. If the inheritance result would cause unstable angle segments or incomplete echo segments to enter the ranging permission, it is rejected. If the inheritance result is still within the hard constraint range, it is written into the observation confidence state space. The advantage of this embodiment is that the insufficient sample interval can also obtain control correction basis consistent with the adjacent observation conditions, while not breaking the total station measurement control boundary.

[0034] Furthermore, the neighborhood inheritance process can employ a weighted allocation method to determine the inheritance attenuation coefficient for the insufficient sample interval. Let the insufficient sample interval be q, and the set of neighborhood intervals participating in the inheritance be... The confidence decay coefficient of the neighborhood interval p is The weight of the number of observed segments is The residual stability weight is The interval adjacency weight is The confidence decay coefficient after inheritance in the insufficient sample interval is determined by the following formula: ; in, This represents the confidence decay coefficient inherited from the insufficient sample interval q. This represents the set of intervals adjacent to interval q that are allowed to participate in inheritance. This represents the existing confidence decay coefficient in the neighborhood interval p. This represents the weight corresponding to the number of available observation segments in the neighborhood interval p. This represents the weight corresponding to the stability of the residuals in the neighborhood interval p. This represents the adjacency weight between a neighborhood interval p and an insufficient sample interval q. If a neighborhood interval has more confirmed segments, smaller residual fluctuations, and is closer to the insufficient sample interval in terms of horizontal angle, vertical angle, and distance, then its weight is considered significant. The proportion of samples in the molecule is higher, and the sample-insufficient interval is closer to the decay state of the neighboring interval. If a certain neighboring interval has few samples or is updated earlier, its weight is reduced, and the inheritance result is not dominated by a single unstable neighboring interval. The advantage of this embodiment is that it discloses how the sample-insufficient interval obtains a calculable correction amount from the adjacent interval, so that the learning correction layer can still maintain a continuous control state when the target distribution is uneven.

[0035] In one embodiment, the edge computing control terminal is equipped with a control trajectory recording process. This process records the observation confidence vector, the control command selected by the conflict resolution sequence, the trigger status of the ranging trigger controller, the execution status of the servo pointing controller, and the confirmation result of the observation confirmation processor, according to the observation target identifier. These records are then arranged in chronological order to form a continuous record. This continuous record is bound to the same observation target identifier and stored, serving as input conditions for subsequent local recapture, repeated observation confirmation, and ranging trigger permission. Specifically, the control trajectory recording process establishes a target identifier for each observation target and writes the comprehensive observation confidence value, component vector, conflict resolution result, ranging trigger permission status, actual trigger status, servo pointing execution status, and observation confirmation result of each observation segment into the same... In the target recording chain, if a target obtains a ranging permission after a damped wait, the target recording chain retains the sequential relationship between the damped wait segment and the ranging permission segment. When confirming the results, the observation confirmation processor can identify whether the ranging value comes from the same locking process after a stable wait. If a target still produces a sudden change in echo after a fine-tuning search, the target recording chain saves the number of fine-tuning searches and the corresponding echo state. When the edge computing control terminal re-acquires the target, it will not repeat the same failed fine-tuning path, but will read the previous control state from the recording chain and directly adjust the acquisition entry. If a target obtains a consistent segment in repeated observation confirmation, the recording chain will use the consistent segment as the input condition for subsequent ranging trigger permission for the same target. The advantage of this embodiment is that it transforms the single control result into a continuous control basis, so that the total station will not lose the abnormal processing context during long-term automatic observation.

[0036] Preferably, a closed-loop reading relationship is formed between the control trajectory recording processing and the control command mapping table. Before generating a new control command for the same target, the edge computing control terminal first reads the most recent observation segment in the target recording chain, determines whether the previous control command has been executed by the servo pointing controller and the ranging trigger controller, and determines whether the observation confirmation processor has given a confirmation result. If the previous ranging trigger permission did not generate a complete echo sampling, the new control command does not directly enter the repeated observation confirmation, but returns to the echo quality component judgment in the ranging stage. If the previous local recapture has not generated a new target locking continuous quantity, the new control command does not allow ranging trigger permission. If the previous repeated observation confirmation obtained a consistent continuous ranging candidate value, the new control command can reduce the dependence on repeated confirmation in subsequent segments of the same target, but is still limited by the physical constraint layer. The control trajectory recording processing also writes the partition attenuation coefficient version of the same target in different observation periods into the recording chain to prevent the inability to distinguish between changes in control conditions and changes in target observation status during subsequent analysis. The advantage of this embodiment is that the generation of control commands depends on the historical state that has been executed and confirmed, avoiding repeated capture, false triggering or incorrect confirmation due to cache delay or state overwriting.

[0037] In a preferred embodiment, when the edge computing control terminal selects control instructions from the conflict resolution sequence, it can adopt a selection method that combines control benefits and control costs, assuming the candidate control instruction set is... The candidate control command is u, and the observation confidence vector of the k-th observation segment is... The control payoff function is The control cost function is The control command output for the current observation segment is determined by the following formula: ; in, This represents the control command that is finally output for the k-th observation segment. This represents the set of candidate control commands that are allowed to be selected in the current observation phase, as indicated by the control command mapping table. This represents the observation confidence vector composed of the locking component, angle stabilization component, servo tracking component, echo quality component, range divergence component, and geometric constraint component. This indicates the control benefit that can be obtained by executing candidate control instruction u under the current confidence state. This represents the control cost of executing the candidate control command u. If the angle stability component is low and the echo quality component is high, the control benefit of ranging triggering is offset by the cost of angle instability, and damping wait becomes the output command. If the angle stability component is high and the echo quality component is low, the benefits of small-amplitude angle micro-scanning and extending the ranging integral are higher than re-acquisition. The edge calculation control terminal selects local correction commands. If the locking component and the ranging dispersion component are both abnormal, the benefits of re-acquisition are higher than continuing ranging. The advantage of this embodiment is that it transforms the rules in the conflict resolution table into a computable command selection process, which facilitates the reuse of the same selection structure in different observation stages.

[0038] In one embodiment, an edge computing control method for a total station is provided, comprising: receiving target lock status, angle encoding stability, servo turntable tracking deviation, distance echo quality, and measurement task geometric constraints from an edge computing control terminal; establishing an observation confidence state space within the edge computing control terminal, and generating an observation confidence vector based on the target lock status, angle encoding stability, servo turntable tracking deviation, distance echo quality, and measurement task geometric constraints; generating at least one of mutually constrained fine-tuning search, distance suppression, distance triggering, repeated observation, or re-acquisition control commands based on the observation confidence vector; and sending the control commands to a servo pointing controller, a distance triggering controller, and an observation confirmation processor, respectively. Specifically, after the measurement task begins, the edge computing control terminal reads the station orientation result and the predicted target azimuth, receives the turntable tracking deviation returned by the servo pointing controller, establishes an observation segment after the target enters the lock range, and continuously receives the target lock status, angle encoding stability, servo turntable tracking deviation, distance echo quality, and measurement task geometric constraints. The system encodes horizontal and vertical angle sequences, measures the distance echo quality, and sets geometric constraints for the measurement task. Then, it performs time alignment, feature normalization, and confidence vector generation in the observation confidence state space. The edge computing control unit reads the physical constraint layer and learning correction layer based on the observation confidence vector to obtain a joint discrimination result. The control command is then selected using the control command mapping table and conflict resolution sequence. When the control command is sent to the servo pointing controller, it manifests as damped waiting, small-amplitude angle micro-scanning, or re-acquisition. When it is sent to the distance trigger controller, it manifests as distance suppression, extended distance integration, or distance trigger permission. When it is sent to the observation confirmation processor, it manifests as repeated observation confirmation, segment validity marking, or result confirmation. After the observation is completed, the control trajectory recording process records the entire observation segment and backfills the observation residuals to the learning correction layer. The advantage of this embodiment is that it fully implements the system's operational process within the edge-side method flow, ensuring that the total station can complete the closed-loop processing from receiving the observation state to the control command output and then backfilling the residuals locally.

[0039] Furthermore, in the edge computing control method for the total station, the edge computing control terminal does not wait for the remote platform to return control results when receiving data. Instead, it writes the local servo feedback, auto-aiming status, distance measurement echo sampling, and task geometric constraints of the total station into the same observation segment. If target locking is interrupted during the formation of the observation segment, the segment is immediately closed and handed over to the segment validity marking process to generate invalid sample markings. If the observation segment forms a complete angle encoding record, echo sampling record, and trigger record, it enters the confidence vector generation process. After the confidence vector is generated, the edge computing control terminal first uses the physical constraint layer to exclude segments that cannot trigger distance measurement, and then uses the learning correction layer to correct segments within the hard constraint range. The segments are partitioned and corrected. The corrected confidence state is entered into the control command mapping table. If there are multiple candidate commands in the mapping result, the conflict resolution sequence selects the control output according to the observation stage and component priority. The observation confirmation processor forms a confirmation result after ranging and binds the confirmation result, control command, trigger state and servo execution state to the target identifier. When the same target enters the locking range again, the edge computing control terminal reads the previous control trajectory from the target record chain and decides whether to directly enter the ranging permission, repeat the observation confirmation or locally recapture. The advantage of this embodiment is that it enables the method flow to have continuous memory capability and avoids the same target from repeatedly performing fixed measurement actions without basis during long-term automatic observation.

[0040] In this embodiment, the edge computing control system and method for the total station can be applied to automatic continuous observation under fixed stations, as well as to sequential observation of multiple targets within the same station. The edge computing control terminal performs the same confidence state generation, physical constraint discrimination, learning correction, command mapping, and trajectory recording processing in each target observation segment, without changing the basic control logic due to changes in the number of targets. When the distance interval, horizontal turning angle interval, and vertical angle interval between target points are different, the learning correction layer saves the confidence attenuation coefficient according to the partition, so that the control conditions under different observation conditions are updated respectively. When a target causes the echo quality component to repeatedly decrease due to unstable reflection state, the control command mapping table guides it to extend the distance integration, repeat the observation confirmation, or locally recapture, without affecting the distance trigger permission of other targets. When there are insufficient samples for a target in the adjacent interval, the neighborhood inheritance processing provides an attenuation coefficient verified by hard constraints, so that the distance trigger condition of the target remains continuous with the adjacent observation conditions. The advantage of this embodiment is that, without changing the hardware structure of the total station, the edge computing control logic combines observation confidence, servo control, distance triggering, and result confirmation into a repeatable automatic measurement control scheme.

Claims

1. An edge computing control system for a total station, characterized in that, It includes an edge computing control terminal and a servo pointing controller, a ranging trigger controller, and an observation confirmation processor that interact with the edge computing control terminal respectively; The edge computing control terminal is configured with an observation confidence state space, which receives target locking status, angle encoding stability, servo turntable tracking deviation, ranging echo quality and measurement task geometric constraints, and generates an observation confidence vector. The edge computing control terminal outputs at least one of the following mutually constrained control commands to the servo pointing controller, the ranging trigger controller, and the observation confirmation processor based on the observation confidence vector: fine-tuning search, ranging suppression, ranging trigger, repeated observation, or re-acquisition control command. The edge computing control terminal is equipped with a control command mapping table. The control command mapping table records the combination relationship between the angle stability component, target lock component, echo quality component, and range dispersion component in the observation confidence vector, and maps the combination relationship into at least one control command among servo turntable damping wait, small angle micro-scan, extended range integration, range trigger permission, repeated observation confirmation, and local re-acquisition according to the observation stages before, during, and after range measurement. The control command mapping table also stores the command sequence for the same observation target in consecutive observation segments. The control command mapping table is configured with a conflict resolution sequence. When the angle stable component is below the angle stability threshold and the echo quality component meets the echo threshold, the conflict resolution sequence prioritizes outputting ranging suppression and servo turntable damping waiting commands; when the angle stable component meets the angle stability threshold and the echo quality component is below the echo threshold, it prioritizes outputting small-amplitude angle micro-scan or extended ranging integration commands; when both the target lock component and the ranging dispersion component are abnormal, it outputs a local re-acquisition command.

2. The edge computing control system for a total station according to claim 1, characterized in that, The observation confidence state space includes time-aligned caching, feature normalization processing, and confidence vector generation processing; The time-aligned buffer establishes observation segments according to the aiming time, angle-encoded sampling time, and ranging echo sampling time of the same observation target; The feature normalization process maps the horizontal angle fluctuation, vertical angle fluctuation, servo turntable stability duration, target lock duration, range echo intensity variation, and continuous range dispersion to a unified dimension. The confidence vector generation process then uses the features under the unified dimension to form an observation confidence vector.

3. The edge computing control system for a total station according to claim 1, characterized in that, The edge computing control terminal is equipped with a physical constraint layer and a learning correction layer; The physical constraint layer stores the hard constraint conditions corresponding to the total station's angular resolution, distance measurement triggering conditions, servo turntable stability boundary, and task geometric boundary. The learning correction layer forms a confidence decay coefficient based on historical valid observation segments and historical invalid observation segments, and saves the confidence decay coefficient according to the target distance interval, horizontal turning angle interval and vertical angle interval; The edge computing control terminal performs joint discrimination on the observation confidence vector based on the hard constraint conditions and the confidence decay coefficient.

4. The edge computing control system for a total station according to claim 2, characterized in that, The time-aligned cache includes observation segment delimitation processing and segment validity marking processing; The observation segment demarcation process takes the moment when the servo turntable enters the target locking range as the segment start point and the moment when the ranging trigger is completed or the target locking is interrupted as the segment end point. The segment validity marking process generates sample markers based on the continuity of angle encoding, target locking, echo sampling integrity, and ranging trigger record within the observation segment. The confidence vector generation process selects observation segments to participate in the calculation based on the sample markers.

5. The edge computing control system for a total station according to claim 3, characterized in that, The learning correction layer includes residual backfilling and partition decay update processing; The residual backfilling process associates the observation residuals confirmed by the observation confirmation processor with the target distance interval, horizontal angle interval, vertical angle interval, and echo quality interval. The partitioned attenuation update process updates the confidence attenuation coefficient within the range defined by the hard constraint condition, and configures update timing identifiers for different intervals. The edge computing control terminal corrects the ranging triggering condition of subsequent observed targets based on the updated confidence attenuation coefficient.

6. The edge computing control system for a total station according to claim 5, characterized in that, The partition decay update process includes neighborhood inheritance processing; The neighborhood inheritance process allocates the confidence decay coefficients of confirmed observation segments in adjacent horizontal turning intervals, adjacent vertical angle intervals, and adjacent target distance intervals within the same station to the insufficient sample intervals according to the number of observation segments, residual stability, interval adjacency relationship, and update time sequence identifier. The insufficient sample interval is the interval in which the number of historical valid observation segments is less than the preset number of segments. The learning correction layer writes the allocated confidence decay coefficient into the observation confidence state space after verifying the consistency between the confidence decay coefficient and the hard constraint conditions.

7. The edge computing control system for a total station according to claim 6, characterized in that, The edge computing control terminal is equipped with control trajectory recording and processing; The control trajectory recording process records the observation confidence vector, the control command selected by the conflict resolution sequence, the triggering state of the ranging trigger controller, the execution state of the servo pointing controller, and the confirmation result of the observation confirmation processor according to the observation target identifier, and forms a continuous record according to the time sequence of the observation segments; The continuous records are bound to the same observation target identifier and stored, and serve as input conditions for subsequent local recapture, repeated observation confirmation, and ranging trigger permission.

8. An edge calculation control method for a total station, characterized in that, The edge computing control system for a total station as described in any one of claims 1-7 includes: receiving target lock status, angle encoding stability, servo turntable tracking deviation, distance echo quality, and measurement task geometric constraints from the edge computing control terminal; An observation confidence state space is established within the edge computing control terminal, and an observation confidence vector is generated based on the target locking state, the angle encoding stability, the servo turntable tracking deviation, the ranging echo quality, and the geometric constraints of the measurement task. Based on the observation confidence vector, generate at least one of the following mutually constrained control commands: fine-tuning search, ranging suppression, ranging triggering, repeated observation, or reacquisition. The control commands are sent to the servo pointing controller, the ranging trigger controller, and the observation confirmation processor, respectively.

Images