Large aperture optical system automatic alignment system and method based on real-time wavefront feedback

Abstract

The present application relates to the technical field of optical alignment, and discloses a large-aperture optical system automatic alignment system and method based on real-time wavefront feedback, which comprises the following steps: collecting real-time wavefront aberration data of the current iteration round to solve a target adjustment vector, and calculating a dead zone crossing ratio by dividing the target adjustment vector by a hysteresis dead zone radius; when the dead zone crossing ratio is not greater than a crossing threshold, a three-section broken line compensation path comprising a pre-advance vector, a lateral bias vector and a regression vector is constructed, and displacement instructions corresponding to the pre-advance and the lateral bias are sequentially issued; then, the regression vector is extended to an overshoot point coordinate and an extended displacement instruction is issued, and immediately after reaching the overshoot point, a reverse displacement instruction is issued, and a dynamic zero position coordinate is calculated and locked; the present application avoids the command failure and overshoot rebound of the adjustment table caused by the hysteresis dead zone due to the small displacement instruction, and improves the convergence lower limit and positioning accuracy of the closed-loop iteration alignment.

Description

Technical Field

[0001] This invention relates to the field of optical assembly and adjustment technology, and more specifically, to an automatic assembly and adjustment system and method for large-aperture optical systems based on real-time wavefront feedback. Background Technology

[0002] Optical assembly and adjustment is a crucial step in the manufacturing process of large-aperture optical systems (aperture ≥ 500 mm), directly determining the final imaging quality and wavefront accuracy. With the continuous increase in optical system aperture and the rising demands for resolution, the precision requirements for the pose and assembly of each optical component have reached the micrometer or even sub-micrometer level. Traditional manual assembly and adjustment methods are time-consuming and struggle to meet high consistency requirements. Therefore, automated assembly and adjustment technology based on feedback evaluation mechanisms has gradually become a research focus in this field.

[0003] In existing technologies, for example, patent CN112558318B provides an auxiliary assembly and adjustment method for large-aperture optical elements. This method selects curvature sensing for wavefront correction and feedback, combines it with the multi-degree-of-freedom spatial motion function of a robotic arm, and uses the mapping relationship between aberration space and robotic arm joint space, combined with a sensitivity matrix, to calculate the amount of movement of the robotic arm, thereby achieving a large dynamic range and high precision assembly and adjustment of large-aperture optical elements. Furthermore, patent CN112946880B discloses a model-based automatic assembly and adjustment method for wavefront-sensorless optical systems. This method applies wavefront-sensorless adaptive optics technology to optical system assembly and adjustment, using the linear relationship between the image evaluation function and the degrees of freedom of the misaligned mirror to solve for the system's misalignment. It does not require an additional wavefront sensor, and offers fast assembly and adjustment speed and high accuracy.

[0004] However, in actual closed-loop automated assembly and adjustment production lines for large-aperture optical systems, operators often encounter a specific phenomenon of assembly and adjustment stagnation and uncontrolled jumps. When the closed-loop iterative assembly and adjustment enters its later stages (e.g., around the 10th iteration), the system-calculated misalignment compensation is extremely small (e.g., the required translation is only 0.015 μm). After the master control unit issues a displacement command based on this extremely small amount, the operator observes no change in the wavefront data, and the adjustment stage physically appears to be in a "step-holding" state without moving. When the system determines that the position is not in place and continues to issue superimposed, tiny displacement commands, accumulating to a certain point, the adjustment stage will suddenly experience a "jump" movement far exceeding the command requirement (e.g., a sudden change of 0.05 μm), directly crossing the theoretical zero-point of convergence. Such command failures and sudden jumps within extremely small displacement ranges not only cause the assembly and adjustment process to oscillate repeatedly at the convergence edge, failing to reach the final design accuracy target, but also easily cause irreversible damage to the optical film or substrate deformation of expensive large-aperture optical components due to instantaneous overshoot collisions or stress mutations, resulting in the direct scrapping of high-value components and delaying the overall development cycle. Summary of the Invention

[0005] The failure of micro-adjustment commands and sudden overshoot in the later stages of closed-loop iteration stem from the inherent nonlinear physical characteristics of the underlying mechanical transmission chain of the high-precision six-degree-of-freedom electric adjustment stage. Regardless of the precision of the lead screw or gear structure used inside the adjustment stage, microscopic mechanical friction, backlash, and hysteresis loops of piezoelectric ceramics are unavoidable. When the magnitude of the linear displacement vector output by the algorithm is smaller than the inherent hysteresis dead zone radius of the adjustment stage hardware, the driving force output by the actuator cannot overcome the static friction between the transmission contact surfaces. The input electrical signal is converted into the elastic potential energy of the mechanical structure rather than effective displacement, manifesting as macroscopic command "step loss". As the main control system continues to issue superimposed commands, the accumulated elastic deformation inside the mechanical transmission chain gradually increases. When the accumulated elastic stress finally exceeds the critical threshold of static friction, the static friction state collapses instantaneously and transforms into the dynamic friction state. The previously accumulated elastic potential energy is released with extremely high acceleration, causing the actual displacement of the adjustment stage to be much greater than the single set micro-step amount, exhibiting an overshoot rebound that crosses the convergence zero point. The physical barrier formed by this nonlinear geometric dead zone makes the conventional path planning logic, which directly points to the target location in a single direction, inevitably fail within a small displacement range.

[0006] To overcome the aforementioned deficiencies of existing technologies, this invention provides an automatic assembly and adjustment system and method for large-aperture optical systems based on real-time wavefront feedback. When adjustment demands trigger mechanical hysteresis dead-zone limitations, a three-segment piecewise linear compensation path, including forward thrust, lateral offset, and regression vectors, is constructed. This transforms the original minimal unidirectional displacement into a spatial piecewise linear sequence where each segment's modulus is within the stable range of displacement transmissibility. Combined with the endpoint overshoot and reverse retracement mechanism introduced in the regression phase to capture dynamic zero-position coordinates, this scheme avoids command failure and stress accumulation overshoot / rebound caused by static friction and hysteresis characteristics of the adjustment stage under minimal displacement commands, thereby improving the final convergence lower limit and positioning reliability of the optical system's closed-loop iterative assembly and adjustment.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] Automatic assembly and adjustment methods for large-aperture optical systems based on real-time wavefront feedback include:

[0009] Collect real-time wavefront aberration data for the current iteration and calculate the target adjustment vector; calculate the dead zone crossing ratio based on the target adjustment vector; when the dead zone crossing ratio is greater than the preset crossing threshold, directly send the target adjustment vector to the adjustment station for execution; when the dead zone crossing ratio is less than or equal to the crossing threshold, construct a three-segment piecewise linear compensation path containing the forward thrust vector, the lateral offset vector, and the regression vector, and send the first segment displacement command corresponding to the forward thrust vector and the second segment displacement command corresponding to the lateral offset vector to the adjustment station along the three-segment piecewise linear compensation path in sequence;

[0010] After the second displacement command is executed, the endpoint of the regression vector is extended along the direction of the regression vector to generate an extended regression vector. The endpoint coordinates of the extended regression vector are marked as the overshoot coordinates. The extended third displacement command corresponding to the extended regression vector is sent to the adjustment platform. When the adjustment platform reaches the overshoot coordinates, the reverse displacement command is immediately sent, and the dynamic zero coordinates are calculated to lock the adjustment platform at the dynamic zero coordinates.

[0011] The method for calculating the target adjustment vector includes:

[0012] The wavefront residual is obtained by subtracting the real-time wavefront aberration data from the preset target wavefront data, and the target adjustment vector is calculated based on the wavefront residual.

[0013] The starting point of the target adjustment vector is the current position coordinate of the adjustment platform, and the ending point is the target position coordinate.

[0014] The method for calculating the dead zone pass-through ratio includes:

[0015] Read the value of the hysteresis dead zone radius, calculate the ratio of the magnitude of the target adjustment vector to the hysteresis dead zone radius, and define it as the dead zone crossing ratio.

[0016] The method for constructing the forward thrust vector is as follows:

[0017] An orthogonal auxiliary coordinate system is established with the current position coordinates of the adjustment platform as the origin, the direction of the target adjustment vector as the positive direction of the first coordinate axis, and the unit normal vector orthogonal to the target adjustment vector as the direction of the second coordinate axis; the second coordinate axis has two directions, positive and negative.

[0018] In the orthogonal auxiliary coordinate system, a forward thrust vector is generated along the positive direction of the first coordinate axis. The starting point of the forward thrust vector is the current position coordinate of the adjustment platform, and the ending point of the forward thrust vector is recorded as the coordinate of the first auxiliary point. The magnitude of the forward thrust vector is set to the first preset multiple of the hysteresis dead zone radius.

[0019] The method for constructing the lateral bias vector includes:

[0020] A lateral bias vector is generated along the second coordinate axis. The starting point of the lateral bias vector is the coordinate of the first auxiliary point, and the ending point of the lateral bias vector is the coordinate of the second auxiliary point. The magnitude of the lateral bias vector is set to the second preset multiple of the hysteresis dead zone radius.

[0021] The method for constructing the lateral bias vector also includes:

[0022] Read the upper and lower limits of the travel of each degree of freedom of the adjustment stage to form the travel range of the adjustment stage. Select the direction that ensures the coordinates of the second auxiliary point do not exceed the travel range of the adjustment stage as the direction of the lateral offset vector. When neither the positive nor negative direction will cause the coordinates of the second auxiliary point to exceed the travel range of the adjustment stage, select the direction that makes the coordinates of the second auxiliary point farther from the travel boundary. When both the positive and negative directions will cause the coordinates of the second auxiliary point to exceed the travel range of the adjustment stage, reduce the magnitude of the lateral offset vector. This magnitude is the maximum value that ensures the coordinates of the second auxiliary point do not exceed the travel range of the adjustment stage.

[0023] The regression vector points from the coordinates of the second auxiliary point to the coordinates of the target position; the regression vector is calculated based on the difference between the coordinates of the second auxiliary point and the coordinates of the target position.

[0024] The method for calculating the dynamic zero-position coordinates includes:

[0025] During the execution of the reverse displacement command, the real-time position reading of the position encoder is read periodically, the position deviation signal between the real-time position reading and the target position coordinate is calculated, the zero-crossing time is determined based on the position deviation signal, the position encoder readings in the sampling periods adjacent to the zero-crossing time are linearly interpolated, and the interpolated position coordinates corresponding to the position deviation signal being equal to zero are taken as the dynamic zero coordinates.

[0026] The method for determining the zero-crossing time based on the position deviation signal is as follows:

[0027] The zero-crossing moment is determined when the sign of the position deviation signal reverses between two adjacent sampling periods.

[0028] An automated assembly and adjustment system for large-aperture optical systems based on real-time wavefront feedback is provided to implement the aforementioned automated assembly and adjustment method for large-aperture optical systems based on real-time wavefront feedback. The system includes:

[0029] Wavefront calculation and compensation module: used to collect real-time wavefront aberration data of the current iteration, calculate the target adjustment vector; calculate the dead zone crossing ratio based on the target adjustment vector; when the dead zone crossing ratio is greater than the preset crossing threshold, the target adjustment vector is directly sent to the adjustment station for execution; when the dead zone crossing ratio is less than or equal to the crossing threshold, a three-segment piecewise linear compensation path containing the forward thrust vector, the lateral offset vector, and the regression vector is constructed, and the first segment displacement command corresponding to the forward thrust vector and the second segment displacement command corresponding to the lateral offset vector are sequentially sent to the adjustment station along the three-segment piecewise linear compensation path.

[0030] Dynamic zero-position locking module: After the second displacement command is executed, the endpoint of the regression vector is extended along the direction of the regression vector to generate an extended regression vector. The endpoint coordinates of the extended regression vector are marked as the overshoot coordinates. The extended third displacement command corresponding to the extended regression vector is sent to the adjustment table. When the adjustment table reaches the overshoot coordinates, the reverse displacement command is immediately sent, and the dynamic zero coordinates are calculated to lock the adjustment table at the dynamic zero coordinates.

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0032] To address the problem of traditional linear displacement commands failing and stress accumulation leading to overshoot caused by the mechanical friction and hysteresis dead zone of the adjustment stage in the later stages of closed-loop iteration in large-aperture optical systems when the adjustment vector of a small target falls into the dead zone, this invention quantifies the issue by calculating the dead zone crossing ratio. When the dead zone limitation is triggered by a small adjustment requirement, a three-segment broken-line compensation path is constructed, including forward thrust, lateral offset, and regression vectors. This path transforms the original unidirectional small displacement, which is smaller than the dead zone radius, into a spatial broken-line sequence where the modulus of each segment is larger than the hysteresis dead zone radius. This ensures that the driving force of the adjustment stage in each segment can overcome the static friction force, keeping the displacement transfer rate of the adjustment stage within a stable range and avoiding the "step loss" phenomenon caused by the nonlinear geometric dead zone. Simultaneously, this invention introduces an endpoint overtaking and reverse retracement mechanism in the regression stage, transforming the target position from the traditional deceleration braking endpoint into a crossing point in the continuous motion trajectory. This eliminates the nondeterministic positioning error caused by the switching between dynamic and static friction states near the target point by the servo controller. Combined with the calculation and locking of dynamic zero coordinates, high-precision zero-crossing capture in motion is achieved. The overall solution eliminates the physical blockage of minimal displacement commands by the underlying hardware mechanical lag through software path reconstruction, avoids stagnation and overshoot rebound in the later stages of assembly and adjustment, and improves the final convergence accuracy and positioning reliability of the optical system's closed-loop assembly and adjustment. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a flowchart of an automatic assembly and adjustment method for a large-aperture optical system based on real-time wavefront feedback, provided in an embodiment of the present invention.

[0035] Figure 2 This is a schematic diagram of wavefront sensor spot sampling provided in an embodiment of the present invention;

[0036] Figure 3 This is a schematic diagram of the deep neural network structure of the aberration-misalignment mapping model provided in an embodiment of the present invention.

[0037] Figure 4 This is a schematic diagram of the direction selection for verifying the travel range of the lateral offset vector provided in an embodiment of the present invention;

[0038] Figure 5 This is a flowchart of the segmented displacement command execution and positioning determination provided in an embodiment of the present invention;

[0039] Figure 6 This is a schematic diagram illustrating the determination of the zero-crossing moment for the sign reversal of the position deviation signal provided in an embodiment of the present invention;

[0040] Figure 7 This is a functional block diagram of an automatic assembly and adjustment system for a large-aperture optical system based on real-time wavefront feedback, provided in an embodiment of the present invention. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] Example 1:

[0043] Please see Figure 1 As shown, this embodiment provides an automatic assembly and adjustment method for a large-aperture optical system based on real-time wavefront feedback, including:

[0044] Step S10: Collect real-time wavefront aberration data for the current iteration and calculate the target adjustment vector; calculate the dead zone crossing ratio based on the target adjustment vector; when the dead zone crossing ratio is greater than the preset crossing threshold, directly send the target adjustment vector to the adjustment station for execution; when the dead zone crossing ratio is less than or equal to the crossing threshold, construct a three-segment piecewise linear compensation path containing the forward thrust vector, the lateral offset vector, and the regression vector, and send the first segment displacement command corresponding to the forward thrust vector and the second segment displacement command corresponding to the lateral offset vector to the adjustment station along the three-segment piecewise linear compensation path.

[0045] Specifically, in each closed-loop iteration of the large-aperture optical system closed-loop automatic assembly method, step S10 undertakes the complete signal conversion link from real-time wavefront aberration data acquisition to the generation of adjustment stage displacement commands. It gradually maps the wavefront deviation in the optical domain, characterized by Zernike polynomial coefficients, into the displacement control quantity that the adjustment stage can execute in the mechanical domain. Furthermore, it introduces a piecewise linear compensation path mechanism to address the inherent hysteresis dead zone characteristics of the adjustment stage, so that the displacement commands issued in each iteration can be converted into the actual mechanical displacement of the adjustment stage.

[0046] Further, step S10 includes:

[0047] Step S11: Collect real-time wavefront aberration data for the current iteration, subtract the real-time wavefront aberration data from the preset target wavefront data to obtain the wavefront residual, and calculate the target adjustment vector based on the wavefront residual; the starting point of the target adjustment vector is the current position coordinate of the adjustment platform, and the ending point is the target position coordinate.

[0048] In step S11, the collimating light source module outputs wavelength-stable monochromatic collimated light to illuminate the large-aperture optical system under test. The Shack-Hartmann wavefront sensor in the wavefront detection module performs wavefront sampling on the outgoing beam after passing through the large-aperture optical system under test. The working principle of the Shack-Hartmann wavefront sensor is as follows: See Figure 2 This is a schematic diagram of the Shack-Hartmann wavefront sensor spot sampling provided in the embodiments of this application. Figure 2The diagram illustrates the incident wavefront, microlens array, sub-aperture regions, area array detector, spot centroid, and reference centroid. The microlens array at the sensor front end divides the incident wavefront into several sub-aperture regions. The beam corresponding to each sub-aperture region is focused by the microlens and forms a spot on the area array detector. The offset of the spot centroid relative to the reference centroid is proportional to the local wavefront slope at the corresponding sub-aperture. The area array detector collects the spot offsets of all sub-apertures and transmits them to the main control and calculation module. Based on the wavefront slope data of all sub-apertures, the main control and calculation module reconstructs the complete outgoing wavefront shape using a mode-based wavefront reconstruction algorithm. This algorithm performs least-squares fitting of the local wavefront slope data of each sub-aperture with the slope basis functions corresponding to each order of the Zernike polynomial, solving for the Zernike polynomial coefficients that minimize the sum of squared residuals, thus obtaining the complete outgoing wavefront shape expressed in Zernike polynomial expansion. The wavefront shape refers to the shape of the equiphase surface on the exit pupil plane after the light wave passes through the large-aperture optical system under test. The degree to which the wavefront shape deviates from the ideal plane wave directly reflects the influence of the pose deviation and spacing deviation of each optical element in the large-aperture optical system under test on the distortion of the imaging beam. The reconstructed wavefront shape is represented by a Zernike polynomial expansion. The Zernike polynomial is a set of orthogonal complete basis functions defined on the unit circle domain. Each order coefficient corresponds to different types of aberration components. Lower-order coefficients correspond to large-scale wavefront distortions such as defocus, coma, and astigmatism, while higher-order coefficients correspond to spherical aberration and more refined wavefront structural features. For example, the first 37 Zernike polynomial coefficients are taken as the order of the wavefront representation. The 37 orders cover the main aberration types that need to be considered during the assembly and adjustment of the large-aperture optical system, while achieving a balance between representation accuracy and computational cost. These 37 coefficients together constitute the real-time wavefront aberration data of the current iteration. The preset target wavefront data is the Zernike polynomial coefficient set corresponding to the theoretical output wavefront of the large-aperture optical system under test when the pose and spacing of each optical element meet the design values. This data is generated by optical design software using ray tracing calculations based on the radius of curvature of each optical element, the design spacing, and the material refractive index. It is then written into the storage unit of the main control and calculation module during the initialization phase of the assembly and adjustment system. The Zernike polynomial coefficients of each order in the real-time wavefront aberration data are subtracted from the corresponding Zernike polynomial coefficients in the target wavefront data. The multidimensional vector formed by these subtractions is the wavefront residual. The value of each component in the wavefront residual reflects the magnitude of the deviation of the corresponding order aberration from the design state, and the sign of the component reflects the direction of the deviation. The wavefront residual, as a whole, completely represents the deviation between the current assembly and adjustment state and the designed assembly and adjustment state of the large-aperture optical system under test in vector form.

[0049] The main control and computation module inputs the wavefront residuals into a pre-trained aberration-misalignment mapping model. See also... Figure 3This is a schematic diagram of the deep neural network structure of the aberration-misalignment mapping model provided in the embodiments of this application. Figure 3 The diagram illustrates the input layer, hidden layers, output layer, and activation function. The input layer corresponds to the wavefront residual, and the output layer corresponds to the offset. The input layer contains Z1, Z2, Z3, up to Z... 37The network consists of 37 nodes, each corresponding to one of the first 37 Zernike polynomial coefficients, forming the input feature vector of the wavefront residual. The network comprises three fully connected hidden layers, all using the ReLU activation function. The output layer contains three nodes: eccentricity, tilt, and spacing, corresponding to the predicted misalignment values ​​of each optical element in its respective degree of freedom direction. The degree of freedom direction refers to the spatial direction in which the adjustment stage can independently perform displacement or rotational motion. For example, for a six-degree-of-freedom adjustment stage, this includes three translational degrees of freedom along the X, Y, and Z axes and three rotational degrees of freedom around the X, Y, and Z axes. The aberration-misalignment mapping model uses a deep neural network architecture, employing the Zernike polynomial coefficients of each order in the wavefront residual as features of the network input layer. Each node in the network output layer corresponds to the eccentricity, tilt, and spacing errors of each optical element in its respective degree of freedom direction. The reason for choosing a deep neural network instead of the traditional linear sensitivity matrix method is that: large-aperture optical systems usually contain multiple optical elements. The influence of the misalignment of each optical element in each degree of freedom direction on the coefficients of each Zernike polynomial is not a simple linear superposition relationship, but rather a nonlinear coupling. The coma caused by the eccentricity of one optical element and the coma caused by the tilt of another optical element are nonlinearly superimposed in the wavefront residual. The inversion accuracy of the linear sensitivity matrix will decrease under the complex conditions of multiple elements and multiple degrees of freedom being misaligned at the same time. Deep neural networks have the ability to extract this nonlinear coupling feature through multiple fully connected layers and nonlinear activation operations, and maintain higher inversion accuracy under the same complex conditions. The training process of the aberration-misalignment mapping model is based on a sample dataset generated from optical simulation: In the optical design model of the large-aperture optical system under test, eccentricity, tilt, and spacing deviations of different magnitudes and combinations of directions are randomly applied to each optical element in each degree of freedom direction. Ray tracing is used to calculate the Zernike polynomial coefficients of the outgoing wavefront corresponding to each combination of deviations. The misalignment combinations are used as labels, and the Zernike polynomial coefficients are used as input features, thus forming paired training samples. The number of training samples needs to cover the magnitude range and direction combinations of eccentricity, tilt, and spacing errors in each degree of freedom direction; for example, there should be no fewer than 1000 training samples. The deep neural network includes an input layer, multiple hidden layers, and an output layer. For example, the number of hidden layers is 3 to 6, the number of nodes in each hidden layer is 128 to 512, and the activation function of each hidden layer is the ReLU function. The output layer does not have an activation function and directly outputs the predicted values ​​of the misalignment in each degree of freedom direction. The training of deep neural networks uses the mean squared error between the output layer prediction error and the label error as the loss function. The network parameters are updated iteratively through backpropagation and gradient descent until the loss function converges to below a preset threshold. After training, the network parameters are stored in the main control and computing module.The aberration-misalignment mapping model, after receiving the wavefront residual, outputs the eccentricity, tilt, and spacing errors of each optical element in each degree of freedom direction. These are the compensation displacement and compensation rotation quantities that each adjustment stage needs to perform. These compensation quantities determine the target adjustment vector: the starting point of the target adjustment vector is the current position coordinates of the adjustment stage obtained in real time from the position encoder, and the ending point is the target position coordinates obtained by superimposing the compensation quantities on the current position coordinates. The direction of the target adjustment vector represents the direction in which the adjustment stage needs to move, and the magnitude of the target adjustment vector represents the distance in which the adjustment stage needs to move. The position encoder is a high-resolution position sensor installed in each degree of freedom direction of the adjustment stage. It converts the mechanical displacement of the adjustment stage into an electrical signal output through the principle of grating diffraction. The output signal is processed by the subdivision circuit to form a digital position reading proportional to the mechanical displacement. The reading resolution of the position encoder is determined by the grating pitch and the electronic subdivision factor.

[0050] By employing wavefront residual-driven target adjustment vector calculation instead of relying on indirect evaluation quantities such as image sharpness, the adjustment direction and distance in each round of closed-loop iteration are directly determined by the core performance indicators of the optical system. Indirect evaluation quantities such as image sharpness have a many-to-one mapping relationship with the misalignment of each optical element; the same sharpness value may correspond to multiple different combinations of misalignment, leading to ambiguity in the adjustment direction. The selection of the adjustment direction in closed-loop iteration requires additional search or trial-and-error processes, increasing the number of iterations and setup time. In contrast, the mapping relationship between the Zernike polynomial coefficients of each order of wavefront residual and the misalignment of each degree of freedom of each optical element, established through an aberration-misalignment mapping model, has a one-to-one correspondence. The adjustment direction in each round of iteration uniquely points in the direction of reducing wavefront aberration, eliminating the need for search or trial-and-error, making the convergence path of the closed-loop iteration more direct and requiring fewer convergence rounds. If the acquisition of wavefront residuals and the calculation of target adjustment vector in step S11 are missing, the piecewise linear compensation path and segmented instruction execution mechanism constructed in subsequent steps S12 to S17 will lack driving input, and the entire closed-loop iterative link will break at the beginning.

[0051] Step S12: Read the hysteresis dead zone radius value from the factory calibration parameters of the adjustment table, calculate the ratio of the magnitude of the target adjustment vector to the hysteresis dead zone radius, and define it as the dead zone crossing ratio; compare the dead zone crossing ratio with the preset crossing threshold to determine whether to start the polyline compensation path construction process.

[0052] When the dead zone crossing ratio is greater than the crossing threshold, there is no need to start the polyline compensation path construction process. The target adjustment vector is directly sent to the adjustment platform for execution. When the dead zone crossing ratio is less than or equal to the crossing threshold, the process proceeds to step S13 and starts the polyline compensation path construction process.

[0053] In step S12, the hysteresis dead zone is an inherent mechanical characteristic of the adjustment platform. It is a blind spot in the displacement response caused by nonlinear factors such as friction, bearing preload, and the hysteresis effect of the piezoelectric ceramic material in the mechanical transmission chain of the adjustment platform. Specifically, when the displacement command issued by the main control and calculation module to the adjustment platform corresponds to a displacement amount less than a critical value, the resistance or hysteresis loop generated by the aforementioned nonlinear factors is sufficient to completely offset the displacement output of the driving force at the end of the transmission. The actual displacement of the adjustment platform is zero or much smaller than the commanded displacement, and the displacement command cannot be faithfully executed. This critical value is the hysteresis dead zone radius. The hysteresis dead zone radius is determined during the factory calibration phase of the adjustment table: The manufacturer applies a series of incrementally increasing small command displacements to the adjustment table in each degree of freedom direction, simultaneously recording the actual displacement (measured by a high-resolution external displacement sensor) corresponding to each command displacement. The ratio of the actual displacement to the command displacement is defined as the displacement transmissibility. The command displacement corresponding to the jump from near zero to near 1 is determined as the hysteresis dead zone radius, and the value of the hysteresis dead zone radius is recorded in the factory calibration parameter file of the adjustment table. During the system initialization phase, the main control and calculation module reads the hysteresis dead zone radius values ​​of each adjustment table in each degree of freedom direction from the factory calibration parameter file.

[0054] The dead-zone pass-through ratio is defined as the dimensionless ratio obtained by dividing the magnitude of the target adjustment vector by the radius of the hysteresis dead zone. The dead-zone pass-through ratio quantifies the sufficiency of the target adjustment vector's magnitude relative to the hysteresis dead zone radius. A dead-zone pass-through ratio much greater than 1 indicates that the magnitude of the target adjustment vector far exceeds the hysteresis dead zone radius, and the control station can faithfully execute the command displacement. A dead-zone pass-through ratio close to or less than 1 indicates that the magnitude of the target adjustment vector is near or within the hysteresis dead zone radius, and directly issuing the target adjustment vector will face the risk that the control station will not be able to generate effective actual displacement. The pass-through threshold is a preset judgment boundary value used to delineate the boundary between the above two situations. The pass-through threshold is set to a value greater than 1, based on the following: In the transition range near the hysteresis dead zone radius, the displacement transfer rate transitions rapidly from near zero to near 1, but the displacement transfer rate in the transition range fluctuates greatly with the number of repetitions, and the repetition consistency is low. The pass-through threshold needs to be set to a value greater than 1 to ensure that the target adjustment vector is only allowed to be directly executed when the displacement transfer rate has entered the stable range (close to 1 and high repetition consistency). For example, the crossing threshold is a value between 1.5 and 3. The specific value is determined based on the width and fluctuation range of the displacement transmissibility transition range in the factory calibration data. The wider the transition range and the greater the fluctuation, the larger the crossing threshold value.

[0055] When the dead-zone pass-through ratio is greater than the pass-through threshold, the magnitude of the target adjustment vector has sufficient margin relative to the hysteresis dead-zone radius. The main control and calculation module directly sends the target adjustment vector as a displacement command to the closed-loop drive module. The closed-loop drive module drives the adjustment table to perform displacement according to the direction and magnitude of the target adjustment vector, without the need to construct a piecewise linear compensation path, resulting in a simplified execution process. The closed-loop drive module is the drive execution unit between the main control and calculation module and the adjustment table. It receives the displacement command issued by the main control and calculation module, converts the command displacement amount in the displacement command into a drive signal acceptable to the adjustment table servo controller, and drives the adjustment table to perform mechanical displacement along the direction and magnitude specified by the displacement command. The closed-loop drive module internally includes a digital-to-analog conversion circuit and a power amplifier circuit. The digital-to-analog conversion circuit converts the digital displacement command output by the main control and calculation module into an analog drive voltage, and the power amplifier circuit amplifies the analog drive voltage before outputting it to the servo controller of the adjustment table. When the dead-zone pass-through ratio is less than or equal to the pass-through threshold, the magnitude of the target adjustment vector is insufficient to ensure that the displacement transfer rate of the adjustment table is within the stable range. The main control and calculation module then proceeds to step S13 to initiate the piecewise linear compensation path construction process. This branching decision mechanism based on dead-zone crossing ratio allows the system to dynamically select the execution strategy in each iteration, initiating the construction of the polyline compensation path only when absolutely necessary. This avoids the additional path detours and time overhead caused by uniformly using the polyline path in all iterations. The limitation of the hysteresis dead zone on the convergence of closed-loop iterations is particularly prominent in the later stages of assembly and adjustment: as the iteration progresses, the wavefront residual shrinks round by round, and the magnitude of the target adjustment vector decreases accordingly. When the magnitude approaches or enters the hysteresis dead zone range, without compensation measures, the adjustment station will lose its effective correction capability for the remaining small wavefront residuals, and the closed-loop iteration will stagnate at a wavefront residual level higher than the design accuracy requirement and will no longer converge. Step S12 identifies this risk in each iteration through quantitative judgment and triggers the construction of the compensation path in a timely manner, so that the closed-loop iteration still has the ability to continuously correct even after the magnitude of the target adjustment vector falls into the hysteresis dead zone range. If the dead zone traversal ratio determination in step S12 is missing, the broken line compensation mechanism in steps S13 to S17 will not be triggered as needed: either it will never be triggered, causing the later iterations of assembly and adjustment to stagnate, or it will be triggered in every round, causing unnecessary path detours in the early stages of assembly and adjustment.

[0056] Step S13: Establish an orthogonal auxiliary coordinate system with the current position coordinates of the adjustment platform as the origin, the direction of the target adjustment vector as the positive direction of the first coordinate axis, and the unit normal vector orthogonal to the target adjustment vector as the direction of the second coordinate axis. When the dimension of the adjustment platform's degree of freedom space is greater than two, select the normalized vector of the projection direction of the degree of freedom direction with the largest distance from the travel boundary margin onto the plane orthogonal to the target adjustment vector as the unit normal vector of the second coordinate axis direction. The second coordinate axis has two directions, positive and negative.

[0057] Step S13 establishes a geometric reference frame for constructing the polyline compensation path. The orthogonal auxiliary coordinate system is a locally orthogonal coordinate system established with the current position coordinates of the adjustment platform as the origin, the direction of the target adjustment vector as the positive direction of the first coordinate axis, and the unit normal vector orthogonal to the target adjustment vector as the direction of the second coordinate axis. When the adjustment platform's degree of freedom space is two-dimensional, the direction orthogonal to the target adjustment vector is uniquely determined; when the adjustment platform's degree of freedom space is three-dimensional or higher, there are infinitely many directions orthogonal to the target adjustment vector. In this case, the projection direction of the degree of freedom direction with the largest distance from the current position coordinates of the adjustment platform to the travel boundary margin onto the plane orthogonal to the target adjustment vector is selected as the initial candidate direction of the second coordinate axis. After normalizing this projection direction, the unit normal vector of the second coordinate axis direction is obtained. The second coordinate axis has two selectable directions, positive and negative (the two directions are symmetrical about the first coordinate axis), and the specific selection is determined by the travel range verification result in subsequent step S15. By setting the origin of the coordinate system at the current position of the adjustment platform, the starting and ending coordinates of each segment vector of the polyline path are referenced to the current position as zero. The displacement of subsequent displacement commands can be directly obtained from the components of each segment vector without the need for coordinate system transformation, thus eliminating floating-point truncation errors that may be introduced during coordinate transformation. Aligning the first coordinate axis with the target adjustment vector, the displacement component along the first coordinate axis directly corresponds to the "displacement towards the target position," and the displacement component along the second coordinate axis directly corresponds to the "lateral displacement away from the target direction." The construction process of the polyline path in the orthogonal auxiliary coordinate system is transformed into a parameter setting problem of determining the displacement magnitude of each segment in two orthogonal directions. The original path planning problem in any direction in three-dimensional space is reduced to two independent one-dimensional parameter determination operations, thereby reducing the computational load and the complexity of the construction logic in the path construction process. Without the orthogonal auxiliary coordinate system constructed in step S13, the determination of the directions of the forward thrust vector and the lateral offset vector in step S14 will lack a unified geometric reference, the orthogonal constraint relationship between the vectors of each path segment cannot be expressed concisely, and the calculation of the regression vector and the magnitude verification in step S16 will also increase the computational complexity due to the lack of a coordinate framework.

[0058] Step S14: Generate a forward thrust vector along the positive direction of the first coordinate axis in the orthogonal auxiliary coordinate system. The starting point of the forward thrust vector is the current position coordinate of the adjustment platform, and the ending point of the forward thrust vector is recorded as the coordinate of the first auxiliary point. Generate a lateral offset vector along the direction of the second coordinate axis. The starting point of the lateral offset vector is the coordinate of the first auxiliary point, and the ending point of the lateral offset vector is recorded as the coordinate of the second auxiliary point. The magnitude of the forward thrust vector is set to a first preset multiple of the hysteresis dead zone radius, and the magnitude of the lateral offset vector is set to a second preset multiple of the hysteresis dead zone radius.

[0059] Step S14 generates the first two segments of the polyline compensation path in the orthogonal auxiliary coordinate system. The forward thrust vector extends from the origin along the positive direction of the first coordinate axis, and the endpoint of the forward thrust vector is recorded as the coordinates of the first auxiliary point. The magnitude of the forward thrust vector is equal to the hysteresis dead zone radius multiplied by a first preset multiple. The lateral offset vector extends from the coordinates of the first auxiliary point along the direction of the second coordinate axis, and the endpoint of the lateral offset vector is recorded as the coordinates of the second auxiliary point. The magnitude of the lateral offset vector is equal to the hysteresis dead zone radius multiplied by a second preset multiple. Both the first and second preset multiples are values ​​greater than 1. The setting of the first preset multiple must simultaneously satisfy two constraints: the lower constraint is to ensure that the magnitude of the forward thrust vector is greater than the hysteresis dead zone radius, thereby ensuring that the displacement transfer rate is in a stable range when the adjustment platform executes the displacement command corresponding to the forward thrust vector; the upper constraint is to prevent the magnitude of the forward thrust vector from being too large. An excessively large forward thrust vector magnitude means that the adjustment platform generates an overshoot far exceeding the magnitude of the target adjustment vector in the target adjustment direction, increasing the path length and execution time of the subsequent regression vector. For example, the first preset multiple is in the range of 1.5 to 3. The setting of the second preset multiple requires ensuring that the magnitude of the lateral offset vector is greater than the hysteresis dead zone radius, and also that the magnitude of the regression vector (the vector pointing from the second auxiliary point coordinates to the target position coordinates) in step S16 is greater than the hysteresis dead zone radius; for example, the second preset multiple is in the range of 1.2 to 2.5. The forward thrust vector pushes the adjustment platform forward along the target direction, exceeding the target position coordinates. The lateral offset vector pushes the adjustment platform away from the first coordinate axis where the target position coordinates are located in the horizontal direction orthogonal to the target direction. After the execution of these two vectors, the second auxiliary point coordinates reached by the adjustment platform are located "behind" the target position coordinates in space, exceeding the target position coordinates along the first coordinate axis and deviating from the target position coordinates along the second coordinate axis. This spatial geometric configuration ensures that the Euclidean distance from the second auxiliary point coordinates to the target position coordinates (i.e., the magnitude of the regression vector) is jointly determined by the overshoot of the forward thrust vector and the lateral offset of the lateral bias vector. By reasonably setting the first and second preset multiples, it can be guaranteed that the magnitude of the regression vector is greater than the hysteresis dead zone radius, thus ensuring that all three vector segments of the polyline compensation path meet the requirement that the magnitude is greater than the hysteresis dead zone radius. The orthogonal auxiliary coordinate system established in step S13 eliminates the need for additional direction calculations when selecting the directions of the forward thrust vector and the lateral bias vector in step S14. The direction of the forward thrust vector is directly taken as the positive direction of the first coordinate axis, and the direction of the lateral bias vector is directly taken as the direction of the second coordinate axis. The direction determination is completely decoupled from the magnitude determination, and the two parameters (the first and second preset multiples) can completely describe the geometric properties of the two vector segments.

[0060] Step S15: Read the upper and lower limits of the travel of each degree of freedom of the adjustment table to form the travel range of the adjustment table. Select the direction that ensures the coordinates of the second auxiliary point do not exceed the travel range of the adjustment table as the direction of the lateral offset vector. When neither the positive nor negative direction causes the coordinates of the second auxiliary point to exceed the travel range of the adjustment table, select the direction that makes the coordinates of the second auxiliary point farther from the travel boundary. When both the positive and negative directions cause the coordinates of the second auxiliary point to exceed the travel range of the adjustment table, reduce the magnitude of the lateral offset vector. This magnitude is the maximum value that satisfies the condition that the coordinates of the second auxiliary point do not exceed the travel range of the adjustment table.

[0061] Step S15 performs a travel range check on the direction of the lateral offset vector. See also Figure 4 This is a schematic diagram of the direction selection for lateral bias vector travel range verification provided in the embodiments of this application. Figure 4 The diagram illustrates the adjustment table's travel range, upper and lower limits, first auxiliary point, positive and negative candidate points, and travel margin comparison, clarifying the verification principle of selecting the direction farther from the travel boundary. The upper and lower limits of the travel in each degree of freedom direction of the adjustment table are determined by the table's mechanical structure and recorded in the factory calibration parameters. The closed interval between the upper and lower limits constitutes the adjustment table's travel range. Step S15 calculates the coordinates of the two second auxiliary points corresponding to the lateral offset vector extending along the positive and negative directions of the second coordinate axis, and determines whether each second auxiliary point coordinate falls within the adjustment table's travel range. When only one direction places the second auxiliary point coordinate within the adjustment table's travel range, the direction not exceeding the travel range is selected as the direction of the lateral offset vector. When the second auxiliary point coordinates corresponding to both directions are within the adjustment table's travel range, the distances from each second auxiliary point coordinate to their nearest travel boundary are calculated, and the direction with the larger distance is selected to retain more travel margin for use in the subsequent regression vector execution stage. Figure 4Taking the example shown, the first auxiliary point extends along the positive direction of the second coordinate axis to obtain a positive direction candidate point, and extends along the negative direction of the second coordinate axis to obtain a negative direction candidate point. The margin from the positive direction candidate point to the upper limit of the travel is greater than the margin from the negative direction candidate point to the lower limit of the travel. Therefore, the positive direction is selected as the direction of the lateral offset vector. When the coordinates of the second auxiliary point corresponding to both directions exceed the travel range of the adjustment stage, the maximum allowable modulus of the lateral offset vector under the condition of not exceeding the travel range is calculated based on the distance between the travel boundary and the coordinates of the first auxiliary point in the direction of the second coordinate axis. This maximum allowable modulus is used as the actual modulus of the lateral offset vector. The travel range verification prevents the adjustment stage from triggering the overtravel protection mechanism or colliding with the mechanical limit structure due to receiving displacement commands that exceed the physical reach. In the assembly and adjustment scenario of a large-aperture optical system, the carrier platform of each optical component may be installed in a position area close to the travel boundary of the adjustment stage. The travel verification in step S15 provides a strict guarantee for the physical executability of the path when the broken line compensation path introduces additional lateral displacement. If the travel range verification in step S15 is missing, the lateral offset vector generated in step S14 may drive the adjustment table beyond its travel limit, causing the adjustment table to enter a locked state after the overtravel protection is triggered, interrupting the assembly and adjustment process, or causing mechanical damage in the direction of freedom without a protection mechanism. When step S15 reduces the magnitude of the lateral offset vector due to travel range limitations, if step S16 verifies that the magnitude of the regression vector is less than or equal to the hysteresis dead zone radius and needs to return to step S14 to increase by a second preset multiple, and the increased magnitude of the lateral offset vector again exceeds the travel range, forming a cyclic conflict, the main control and calculation module will lock the magnitude of the lateral offset vector to the maximum allowable value of the reduced travel range in step S15, and instead only increase the first preset multiple to increase the overshoot of the forward thrust vector. By increasing the overshoot of the forward thrust vector, the component of the regression vector in the first coordinate axis direction is lengthened, so that the total magnitude of the regression vector exceeds the hysteresis dead zone radius, thereby resolving the cyclic conflict.

[0062] Step S16: Calculate the regression vector from the second auxiliary point coordinates to the target position coordinates based on the difference between the second auxiliary point coordinates and the target position coordinates. Concatenate the forward thrust vector, lateral offset vector, and regression vector in sequence to form a three-segment broken line compensation path. Verify whether the magnitude of each segment vector in the three-segment broken line compensation path is greater than the hysteresis dead zone radius. If the magnitude of any segment vector is less than or equal to the hysteresis dead zone radius, return to step S14, increase the magnitude of the forward thrust vector and lateral offset vector, and recalculate.

[0063] Step S16 calculates the regression vector based on the coordinates of the second auxiliary point and the target position coordinates, and assembles and verifies the three-segment polyline compensation path. The starting point of the regression vector is the coordinates of the second auxiliary point, the ending point is the target position coordinates, and the direction is from the second auxiliary point coordinates to the target position coordinates. The magnitude of the regression vector is the Euclidean distance between the second auxiliary point coordinates and the target position coordinates. The forward thrust vector, lateral offset vector, and regression vector are connected end-to-end in the execution order to form a three-segment polyline compensation path starting from the current position coordinates of the adjustment platform, passing through the coordinates of the first and second auxiliary points, and reaching the target position coordinates. After the path is assembled, the magnitude of each vector segment is verified: the magnitudes of the forward thrust vector, lateral offset vector, and regression vector are compared with the hysteresis dead zone radius. The magnitudes of the forward thrust vector and lateral offset vector are directly set by the first and second preset multiples and are necessarily greater than the hysteresis dead zone radius under normal parameter ranges. The magnitude of the regression vector depends on the geometric relationship between the overshoot of the forward vector, the lateral offset of the lateral bias vector, and the magnitude of the target adjustment vector. When the magnitude of the target adjustment vector is extremely small and the first and second preset multiples are set too low, the endpoint (target position coordinates) of the regression vector is close to the coordinates of the second auxiliary point, and the magnitude of the regression vector may be less than or equal to the hysteresis dead zone radius. When the verification finds that the magnitude of any vector segment is less than or equal to the hysteresis dead zone radius, return to step S14, increase the first and second preset multiples by a preset increment, and recalculate the forward vector and the lateral bias vector. The preset increment is determined so that the expected increment of the regression vector magnitude after the increase is not less than the difference between the hysteresis dead zone radius and the current regression vector magnitude. For example, the preset increment is a value between 0.3 and 0.8. After increasing the first and second preset multiples, re-execute the direction verification in step S15 and the magnitude verification in step S16, repeating the cycle until the magnitudes of all three vector segments are greater than the hysteresis dead zone radius. This verification-recalculation loop enables the piecewise linear compensation path to have adaptive adjustment capabilities. Regardless of how small the magnitude of the target adjustment vector is, the magnitudes of all three vector segments can be increased by the first and second preset factors to ensure they all exceed the hysteresis dead zone range. The piecewise linear compensation path's ability to cope with the hysteresis dead zone is not constrained by the lower limit of the target adjustment vector's magnitude. When the travel constraint and the hysteresis dead zone constraint cannot be satisfied simultaneously, i.e., the magnitude of the lateral bias vector still exceeds the travel range after increasing the second preset factor, the magnitude of the lateral bias vector is fixed to the maximum allowable travel value determined in step S15. Only the first preset factor is increased to increase the overshoot of the forward vector, thereby increasing the magnitude of the regression vector to meet the requirement of being greater than the hysteresis dead zone radius. If increasing the first preset factor alone still cannot ensure that the magnitudes of all three vector segments meet the constraints, the main control and calculation module issues an insufficient travel alarm and records the current iteration status.The verification-recalculation loop between steps S16 and S14 simultaneously considers the hysteresis dead zone constraint and the travel constraint already included in step S15, ensuring that the final generated polyline compensation path satisfies both types of physical constraints.

[0064] Step S17: Along the three-segment broken line compensation path, the first segment displacement command corresponding to the forward thrust vector and the second segment displacement command corresponding to the lateral offset vector are sequentially sent to the adjustment table. Each segment displacement command contains the command displacement amount of the corresponding segment. When executing each segment displacement command, the actual displacement amount of the adjustment table is read in real time, and the actual displacement amount is compared with the command displacement amount. When the absolute value of the difference between the actual displacement amount and the command displacement amount is less than the preset single segment positioning accuracy threshold, it is determined that the current segment displacement command has been completed, and the next segment displacement command is sent. After the second segment displacement command is completed, proceed to step S20.

[0065] See Figure 5 After the polyline compensation path is constructed, step S17 sends displacement commands to the adjustment platform segment by segment according to the execution order of the three-segment polyline compensation path. The main control and calculation module encapsulates the command displacement corresponding to the forward thrust vector into the first segment displacement command and sends it to the closed-loop drive module. The closed-loop drive module drives the adjustment platform to perform displacement along the forward thrust vector direction. During the execution of each segment displacement command, the main control and calculation module reads the actual displacement of the adjustment platform periodically, using the sampling period of the adjustment platform position encoder as the time interval. The actual displacement is subtracted from the command displacement of the corresponding segment, and the absolute value is taken. When the absolute value is less than the preset single-segment positioning accuracy threshold, the displacement command for this segment is determined to be completed. If the determination condition is not met, the actual displacement of the adjustment platform is continuously read in real time until the determination condition is met. The single-segment positioning accuracy threshold is set based on the resolution of the position encoder on the adjustment table and the overall positioning accuracy requirements of the assembly and adjustment system. The single-segment positioning accuracy threshold must be greater than the position encoder resolution (otherwise the encoder cannot distinguish whether the position is correct or not), and smaller than the hysteresis dead zone radius (otherwise the positioning determination would be so lenient as to be meaningless). For example, the single-segment positioning accuracy threshold can be 2 to 5 times the position encoder resolution. After the first displacement command is executed, the main control and calculation module immediately issues the second displacement command corresponding to the lateral offset vector. The issuance, execution, and positioning determination process of the second displacement command is the same as that of the first displacement command. Figure 5After the second displacement command is issued, the process of real-time displacement reading and positioning determination is repeated. After the second displacement command is executed, the adjustment table reaches the coordinate position of the second auxiliary point. The main control and calculation module then proceeds to step S20 to execute the third displacement corresponding to the regression vector and the subsequent dynamic zero-position locking process. The method of executing and determining positioning segment by segment ensures that each displacement segment of the polyline compensation path is completed independently along the set path direction and set modulus. If the endpoint coordinates corresponding to the three vectors in the three-segment polyline compensation path are issued to the servo controller of the adjustment table at once, the servo controller's built-in path planning logic may choose a straight path to reach the endpoint, causing the actual displacement component at a certain stage of the movement to fall into the hysteresis dead zone range and generate cumulative positioning deviation. Issuing the command segment by segment and issuing the next segment only after confirming positioning forces the adjustment table to move strictly along the direction of each segment of the polyline path. The direction and modulus of each segment are independently controlled by the main control and calculation module, avoiding interference from the servo controller's path planning logic on the polyline path.

[0066] Step S10 elevates the final achievable assembly accuracy of the closed-loop iteration from a level constrained by the hysteresis dead zone radius to a level constrained by the wavefront sensor measurement accuracy and the inversion accuracy of the aberration-misalignment mapping model. Without the introduction of the piecewise linear compensation path, the closed-loop iteration stalls in the later stages of assembly due to the target adjustment vector magnitude entering the hysteresis dead zone, and the final wavefront residual remains at the level of the wavefront deviation corresponding to the hysteresis dead zone radius. With the introduction of the piecewise linear compensation path, any tiny target adjustment vector can be reliably executed by the adjustment stage via the piecewise linear path. The lower limit of convergence of the closed-loop iteration is no longer limited by the hysteresis dead zone radius, but depends only on the measurement resolution of the wavefront sensor and the inversion residual of the aberration-misalignment mapping model. In actual assembly scenarios of large-aperture optical systems, the wavefront deviation magnitude corresponding to the measurement resolution of the wavefront sensor and the inversion residual of the mapping model is usually much smaller than the wavefront deviation magnitude corresponding to the hysteresis dead zone radius. Therefore, the piecewise linear compensation path mechanism in step S10 substantially improves the final achievable assembly accuracy of the closed-loop iteration by a significant order of magnitude. Meanwhile, the forward thrust vector and lateral offset vector cause actual mechanical displacement of the adjustment platform in both the target adjustment direction and two dimensions orthogonal to the target adjustment direction. This multi-directional reciprocating motion produces a pre-sliding activation effect on the friction contact surfaces in the mechanical transmission chain of the adjustment platform. In tribology, the static friction coefficient will temporarily transition to the dynamic friction coefficient after micro-displacement activation, which is equivalent to a temporary reduction in the hysteresis dead zone radius. This means that in the immediate iteration after the completion of the piecewise linear compensation path, the hysteresis dead zone radius of the adjustment platform is temporarily lower than the factory calibration value, and the dead zone crossing ratio increases relatively under the condition that the magnitude of the target adjustment vector remains unchanged. This helps subsequent iterations return from the piecewise linear compensation mode to the direct execution mode, reduces the frequency of use of the piecewise linear compensation path in the overall closed-loop iteration process, and shortens the total time and total path travel required for the closed-loop iteration to reach convergence. After step S17 completes the segmented execution of the forward thrust vector and the lateral offset vector, the adjustment platform reaches the coordinate position of the second auxiliary point. Step S20 takes over the execution authority of the regression vector and further implements endpoint overtaking and reverse backscanning during the regression process to accurately lock the dynamic zero position. The reliable execution of the first two segments of the polyline compensation path in step S10 provides an accurate starting position (coordinates of the second auxiliary point) for the execution of the regression vector in step S20. The connection between the two steps ensures the complete closure of the polyline compensation path from the starting point to the end point.

[0067] Step S20: After the second displacement command is executed, the endpoint of the regression vector is extended along the direction of the regression vector to generate an extended regression vector. The endpoint coordinates of the extended regression vector are marked as the overshoot coordinates. The extended third displacement command corresponding to the extended regression vector is sent to the adjustment platform. When the adjustment platform reaches the overshoot coordinates, the reverse displacement command is immediately sent, and the dynamic zero coordinates are calculated to lock the adjustment platform at the dynamic zero coordinates.

[0068] Specifically, after completing the segmented execution of the forward thrust vector and lateral offset vector in step S17, step S20 takes over the final execution authority of the broken line compensation path, undertaking the complete control link of guiding the adjustment platform from the second auxiliary point coordinate position to the vicinity of the target position coordinate position via the regression vector, and implementing endpoint locking at the target position coordinate position. When step S17 is completed, the adjustment platform has reached the second auxiliary point coordinate position. The second auxiliary point coordinate position is located in space in front of the target position coordinate position along the positive direction of the first coordinate axis and deviates from the straight line of the target position coordinate position along the direction of the second coordinate axis. The adjustment platform needs to move from the second auxiliary point coordinate position to the target position coordinate position along the direction of the regression vector to complete one round of closed-loop iterative pose correction. The regression vector is the spatial vector from the second auxiliary point coordinate position to the target position coordinate position, calculated in step S16 based on the difference between the second auxiliary point coordinate position and the target position coordinate position. The magnitude of the regression vector has been confirmed by step S16 to be greater than the hysteresis dead zone radius, satisfying the condition that the displacement transfer rate of the adjustment platform is in the stable range. In step S20, when executing the displacement command corresponding to the regression vector, the adjustment stage is not directly stopped at the target position coordinates. Instead, an endpoint overtaking and reverse retracement mechanism is introduced. This allows the adjustment stage to first cross the target position coordinates along the regression vector direction to reach the overtaking point coordinates, and then execute the reverse displacement command along the opposite direction of the regression vector. During the reverse movement, the zero-crossing moment is captured when crossing the target position coordinates, and the dynamic zero coordinates are calculated by linear interpolation based on the position encoder readings before and after the zero-crossing moment, locking the adjustment stage at the dynamic zero coordinates. The reason for introducing the endpoint overtaking and reverse retracement mechanism is that when the servo controller of the adjustment stage executes a unidirectional displacement command and decelerates to the endpoint position, the sudden change in braking force caused by the transition of friction from dynamic friction to static friction, as well as the creep effect of the piezoelectric ceramic drive element in the voltage holding state, will cause the actual stopping position of the adjustment stage to deviate from the command endpoint coordinates. The direction and magnitude of the deviation fluctuate with the movement direction, movement speed, and ambient temperature before stopping, which is a non-deterministic positioning error. If the endpoint coordinates of the regression vector (i.e., the target position coordinates) are directly used as the endpoint coordinates of the third displacement command, the actual position of the adjusting table after deceleration and stopping will deviate from the target position coordinates by an unpredictable positioning deviation. The correction effect of the wavefront residual in the closed-loop iteration is constrained by the positioning deviation, making it impossible for the assembly and adjustment accuracy to exceed the wavefront deviation level corresponding to the single-direction positioning deviation of the servo controller. The endpoint overtaking and reverse retracement mechanism, by having the adjusting table cross the target position coordinates in two opposite directions, uses the zero-crossing event of the position deviation signal during the reverse movement to calibrate the precise alignment time between the adjusting table and the target position coordinates. This improves the positioning accuracy from the single-direction stopping accuracy of the servo controller to the reading resolution level of the position encoder. The magnitude of the positioning accuracy is changed from being determined by the braking characteristics of the servo controller to being determined by the resolution of the position sensor.

[0069] Further, step S20 includes:

[0070] Step S21: Extend the endpoint coordinates of the regression vector backward along the direction of the regression vector by the endpoint overshoot amount, which is the third preset multiple of the hysteresis dead zone radius; the endpoint coordinates of the extended regression vector are marked as the overshoot point coordinates, and the endpoint coordinates of the third displacement command are replaced by the overshoot point coordinates instead of the target position coordinates.

[0071] Step S21 extends the endpoint coordinates of the regression vector backward by an endpoint overshoot amount along the direction of the regression vector. The endpoint overshoot amount is the product of the hysteresis dead zone radius and a third preset multiple. The third preset multiple is a value greater than 1, and its setting is based on two aspects: First, the endpoint overshoot amount must be greater than the hysteresis dead zone radius, so that the adjustment platform maintains sufficient motion inertia during its movement to the overshoot point coordinate position after passing the target position coordinates, preventing the adjustment platform from stopping prematurely near the target position coordinates due to the remaining displacement entering the hysteresis dead zone range; second, the endpoint overshoot amount must provide sufficient buffer margin for the braking distance of the servo controller in the regression direction, so that the adjustment platform can reliably stop when it reaches the overshoot point coordinate position instead of continuing to slide due to inertia. For example, the third preset multiple is in the range of 1.2 to 2.0, with the upper limit of the range used when the braking distance of the adjustment platform is large relative to the hysteresis dead zone radius, and the lower limit of the range used when the braking distance is small. The starting point of the extended regression vector remains the coordinates of the second auxiliary point, and its direction is consistent with the direction of the regression vector. The ending point coordinates are the spatial positions reached after the target position coordinates are extended along the direction of the regression vector by the endpoint, and are denoted as the overtaking point coordinates. In step S21, the ending point coordinates of the third segment displacement command are replaced from the target position coordinates to the overtaking point coordinates. The replaced third segment displacement command drives the adjustment platform to move from the second auxiliary point coordinates along the direction of the regression vector. It passes the target position coordinates but does not stop, and continues to move until it reaches the overtaking point coordinates. When the adjustment platform passes the target position coordinates, it is in a state of uniform motion or deceleration, rather than being in a stopped state. In the motion state, the friction force in the mechanical transmission chain of the adjustment platform is maintained at the dynamic friction level. There is no braking positioning deviation caused by the sudden change of static friction force. The target position coordinates are only an intermediate passing point on the motion trajectory, not a stopping point. The spatial relationship between the adjustment platform and the target position coordinates is continuously recorded by the position encoder at the passing time, rather than determined by the stopping position of the servo controller. The regression vectors in steps S21 and S16 maintain the same direction, with only the endpoint overshoot added to the magnitude. The magnitude of the extended regression vector equals the magnitude of the regression vector plus the endpoint overshoot. Since the magnitude of the regression vector has been verified in step S16 to be greater than the hysteresis dead zone radius, and the endpoint overshoot itself is also greater than the hysteresis dead zone radius, the magnitude of the extended regression vector must be greater than the hysteresis dead zone radius. Therefore, the displacement transfer rate of the adjustment table when executing the extended third displacement command is within the stable range. Step S21 transforms the target position coordinates from the endpoint of the displacement command to the intermediate point of the displacement trajectory. This transformation eliminates the need for the servo controller to perform deceleration and stop operations at the target position coordinates, changing the motion state of the adjustment table near the target position coordinates from "deceleration-stop" to "uniform speed passage" or "slow deceleration passage," fundamentally avoiding the positioning deviation caused by the sudden change in static friction and the creep effect of piezoelectric ceramics during deceleration and stop.

[0072] Step S22: Execute the extended third segment displacement command corresponding to the extended regression vector. When the adjustment platform reaches the overshoot point coordinates, immediately issue a reverse displacement command. The direction of the reverse displacement command is opposite to the direction of the regression vector. The magnitude of the reverse displacement command is set to the sum of the endpoint overshoot amount and the hysteresis dead zone radius.

[0073] Step S22: Immediately after the adjustment platform reaches the overshoot point coordinates, a reverse displacement command is issued. The direction of the reverse displacement command is opposite to the direction of the regression vector, i.e., from the overshoot point coordinates to the second auxiliary point coordinates. The magnitude of the reverse displacement command is set as the sum of the endpoint overshoot amount and the hysteresis dead zone radius. The basis for this setting is: the execution of the reverse displacement command causes the adjustment platform to move from the overshoot point coordinates in the opposite direction of the regression vector. During the movement, the adjustment platform will cross the target position coordinates again. The magnitude of the reverse displacement command must ensure that the adjustment platform still has sufficient remaining displacement to continue moving a certain distance after crossing the target position coordinates, so that the adjustment platform will not enter the deceleration phase prematurely due to approaching the endpoint of the reverse displacement command during the crossing process. The target position coordinates are located at a distance equal to the endpoint overshoot amount in the opposite direction of the overshoot point coordinates. The magnitude of the reverse displacement command is the sum of the endpoint overshoot amount and the hysteresis dead zone radius. When the adjustment platform crosses the target position coordinates, there is still a distance equal to the hysteresis dead zone radius remaining from the endpoint of the reverse displacement command. The adjustment platform is still in a stable motion state rather than a decelerated and stopped state when crossing the target position coordinates. The adjustment platform first crosses the target position coordinates along the regression vector direction to reach the overshoot point coordinates, and then crosses the target position coordinates in the opposite direction of the regression vector. The two crossing directions are opposite, and the movement trajectory of the adjustment platform forms a complete round trip near the target position coordinates. Since the adjustment platform is in motion during both crossings of the target position coordinates, the readings of the position encoder are continuous and stable. In subsequent step S23, the zero-crossing moment can be captured by the deviation signal between the position encoder reading and the target position coordinates during the reverse crossing. The timing of issuing the reverse displacement command in step S22 is immediately after the adjustment platform reaches the overshoot point coordinates. This means that the main control and calculation module determines that the third segment of the extended displacement command has been executed. That is, it reads the actual displacement of the adjustment platform cycle by cycle with the sampling period of the position encoder as the time interval. When the absolute value of the difference between the actual displacement and the command displacement of the third segment of the extended displacement command is less than the single-segment positioning accuracy threshold, it determines that the reverse displacement command is issued in the next control cycle after the execution is completed. No additional waiting time is inserted between the two displacement commands. The immediate issuance of the signal serves two purposes: the shorter the dwell time of the adjustment platform at the overtaking point coordinates, the shorter the time for static friction to transition from dynamic friction back to static friction. This means the static friction that the mechanical transmission chain of the adjustment platform needs to overcome during reverse startup is closer to the level of dynamic friction, resulting in a higher displacement transfer rate and a faster startup response during reverse startup, which is beneficial for the displacement tracking accuracy of the adjustment platform during the reverse motion phase. Steps S22 and S21 logically constitute a continuous reciprocating motion sequence: step S21 causes the adjustment platform to pass the target position coordinates in the positive direction along the regression vector, and step S22 causes the adjustment platform to pass back through the target position coordinates in the opposite direction along the regression vector. These two steps work together to create a change in motion direction from positive to negative near the target position coordinates, establishing the kinematic prerequisite for step S23 to capture the precise moment the adjustment platform passes through the target position coordinates during the reverse motion.

[0074] Step S23: During the execution of the reverse displacement command, read the real-time position reading of the position encoder periodically, calculate the position deviation signal between the real-time position reading and the target position coordinates, and determine the zero crossing moment when the sign of the position deviation signal reverses between two adjacent sampling periods.

[0075] Step S23 involves reading the real-time position reading of the position encoder periodically during the execution of the reverse displacement command, calculating the position deviation signal between the real-time position reading and the target position coordinates, and monitoring the sign change of the position deviation signal to determine the zero-crossing moment. (See also...) Figure 6 This is a schematic diagram illustrating the zero-crossing moment determination of the position deviation signal sign reversal provided in the embodiments of this application. Figure 6 The diagram uses the sampling time as the horizontal axis and the position deviation signal as the vertical axis, illustrating the continuous change trend of the position deviation signal with the sampling time, the zero-crossing time interval, and the reversal process of the position deviation signal from positive to negative. The main control and calculation module uses the sampling period of the position encoder as the time interval. During each sampling period of the reverse displacement command execution, it reads the real-time position reading output by the position encoder and subtracts the target position coordinates from the real-time position reading to obtain the position deviation signal. The position deviation signal is a signed scalar value. A positive value indicates that the adjusting platform is located on the side of the target position coordinate along the positive direction of the regression vector at the current sampling time (i.e., it has not yet crossed the target position coordinate). A negative value indicates that the adjusting platform has crossed the target position coordinate and is located on the side of the target position coordinate along the opposite direction of the regression vector. A zero value indicates that the adjusting platform is exactly at the target position coordinate. During the reverse movement, as the adjusting platform moves from the overshoot point coordinate in the opposite direction of the regression vector and crosses the target position coordinate, the position deviation signal continuously changes from positive to negative. The sign of the position deviation signal reverses from positive to negative at the moment the adjusting platform crosses the target position coordinate. Figure 6Taking the example shown, the position deviation signal exhibits a monotonically decreasing trend with the sampling time. When the adjustment platform passes the target position coordinates, the position deviation signal crosses the zero point of the horizontal axis, and its sign changes from positive to negative. The zero-crossing moment occurs between two adjacent sampling periods where the sign reverses. The main control and calculation module compares the position deviation signal sign of the current sampling period with that of the immediately preceding sampling period within each sampling period. When the position deviation signal signs of two adjacent sampling periods are different (one is positive and the other is negative, or one is positive and the other is zero, or one is zero and the other is negative), it is determined that the zero-crossing moment occurs within the time interval between these two adjacent sampling periods. The physical meaning of the zero-crossing moment is the moment when the adjustment platform precisely passes the target position coordinates on the reverse motion trajectory. The position of the adjustment platform corresponding to the zero-crossing moment is the crossing position of the target position coordinates on the reverse motion trajectory. Step S23 uses the sign reversal of the position deviation signal as the zero-crossing criterion instead of the absolute value of the position deviation signal being less than a certain threshold. The reason is that the sign reversal criterion only relies on discrete events where the position deviation signal crosses from a positive value to a negative value between adjacent sampling periods. It is not affected by the magnitude of the absolute value of the position deviation signal. It can be reliably triggered when the adjustment table moves at any constant or variable speed across the target position coordinates. It will not miss the detection due to the position deviation signal absolute value jumping too much in a single sampling period because the adjustment table moves too fast. On the other hand, the absolute value threshold criterion requires a threshold parameter to be set in advance. If the threshold is too large, the judgment interval will cover a wide spatial range on both sides of the target position coordinates, thereby reducing the positioning accuracy. If the threshold is too small, the crossing event will be missed when the adjustment table moves too fast because the position jump between adjacent sampling periods exceeds the threshold width. Step S23 captures the zero-crossing moment during the reverse motion rather than the forward motion, which works in conjunction with the reverse displacement command design in step S22: During the forward motion (step S21) phase, the adjustment platform moves from the second auxiliary point coordinates to the overtaking point coordinates and passes the target position coordinates for the first time. However, the main task of the forward motion phase is to push the adjustment platform to the overtaking point coordinates. When passing the target position coordinates in the forward direction, the adjustment platform may be in the acceleration or constant speed phase, with a relatively high speed and a large position increment within a single sampling period, resulting in a relatively low time resolution for zero-crossing determination. During the reverse motion (step S22) phase, the adjustment platform moves from the overtaking point coordinates in the opposite direction to the regression vector and crosses the target position coordinates again. After the reverse motion is started, the speed of the adjustment platform gradually increases from zero. The speed when crossing the target position coordinates is lower than the speed when passing in the forward direction, resulting in a smaller position increment within a single sampling period, higher time resolution for zero-crossing determination, and correspondingly higher position interpolation accuracy.

[0076] Step S24: Linearly interpolate the position encoder readings within the sampling periods adjacent to the zero crossing time, and use the interpolated position coordinates corresponding to the position deviation signal being equal to zero as the dynamic zero coordinates. Send a positioning lock command to the adjustment station with the dynamic zero coordinates as the set value to lock the adjustment station at the dynamic zero coordinates. In the locked state of the adjustment station, new real-time wavefront aberration data is collected. Based on the new real-time wavefront aberration data and the preset target wavefront data, it is determined whether the closed-loop iteration has converged. If it has not converged, the new real-time wavefront aberration data is used as input, and the process returns to step S11 to recalculate the target adjustment vector and determine whether to restart the piecewise linear compensation path construction process.

[0077] Step S24 performs linear interpolation on the position encoder readings within the sampling periods adjacent to the zero-crossing moment, and uses the interpolated position coordinates corresponding to the zero-position signal as the dynamic zero-position coordinates. The zero-crossing moment determined in step S23 is located between two adjacent sampling periods. The precise moment when the adjustment platform passes the target position coordinates does not fall exactly on the sampling moment of a single sampling period, but rather within the time interval between two adjacent sampling moments. Therefore, the precise position reading of the position encoder at the zero-crossing moment cannot be directly obtained from the reading of any single sampling period. Step S24 takes the position encoder reading from the sampling period before the zero-crossing moment (denoted as the positive side reading, with the position deviation signal being positive) and the position encoder reading from the sampling period after the zero-crossing moment (denoted as the negative side reading, with the position deviation signal being negative). Using the position deviation signal as the dependent variable and the position encoder reading as the independent variable, the linear interpolation method is used to calculate the position encoder reading corresponding to the zero-position deviation signal. The linear interpolation is based on the assumption that the movement speed of the adjustment platform between two adjacent sampling periods is approximately constant, the position encoder reading changes approximately linearly with time, and the position deviation signal also changes approximately linearly with the position encoder reading. Under the condition that the sampling period of the position encoder is short enough (compared to the movement speed of the adjustment stage), the displacement of the adjustment stage between two adjacent sampling periods is much smaller than the total movement stroke of the adjustment stage. The error of constant speed approximation can be ignored. The calculation result of linear interpolation is the approximate value of the position encoder reading when the adjustment stage passes through the target position coordinates. The approximation accuracy is determined by the position nonlinear component caused by the acceleration of the adjustment stage within the sampling period. The specific calculation process of linear interpolation adopts the well-known one-dimensional linear interpolation formula in this field: using the position deviation signal value at the positive side reading and the position deviation signal value at the negative side reading as two known interpolation nodes, and taking the position deviation signal value equal to zero as the interpolation condition, the corresponding position encoder reading value is solved. The position encoder reading value obtained by interpolation calculation is the dynamic zero coordinate. The physical meaning of the dynamic zero coordinate is: the actual spatial position of the adjustment stage recorded by the position encoder when the adjustment stage passes through the target position coordinates on the reverse movement trajectory. There may be a slight difference between the dynamic zero coordinate and the target position coordinate. This difference arises from the limitations of the position encoder's reading resolution and the approximate error of linear interpolation. The magnitude of this difference is the same as the position encoder's reading resolution, and is much smaller than the magnitude of the servo controller's single-direction stop positioning deviation.

[0078] After calculating the dynamic zero-position coordinates, the main control and calculation module sends a positioning lock command to the closed-loop drive module, using the dynamic zero-position coordinates as the setpoint. The execution logic of the positioning lock command is as follows: the closed-loop drive module switches the servo controller of the adjustment table to position holding mode. In position holding mode, the servo controller uses the dynamic zero-position coordinates as the target setpoint of the position loop, reads the position feedback from the position encoder in real time, and continuously outputs drive signals through the proportional-integral-derivative control law of the position loop to maintain the actual position of the adjustment table near the dynamic zero-position coordinates, thus offsetting the influence of external disturbances (vibration, thermal drift, etc.) on the position of the adjustment table. The position holding accuracy of the adjustment table in the positioning lock state depends on the gain parameter of the position loop control law and the feedback resolution of the position encoder; the order of magnitude of the holding accuracy is at the same level as the reading resolution of the position encoder. After the adjustment platform enters the positioning lock state, the main control and calculation module activates the wavefront detection module to collect new real-time wavefront aberration data. It subtracts the Zernike polynomial coefficients of each order in the new real-time wavefront aberration data from the corresponding order Zernike polynomial coefficients in the preset target wavefront data term by term to obtain the new wavefront residual. The root mean square (RMS) value of the new wavefront residual is calculated and compared with a preset assembly convergence threshold. The assembly convergence threshold is the upper limit value of the wavefront residual pre-written into the main control and calculation module during the initialization phase of the assembly system based on the design specifications of the large-aperture optical system under test. When the RMS value of the new wavefront residual is less than or equal to the assembly convergence threshold, and the fluctuation is less than or equal to the preset fluctuation tolerance, the closed-loop iterative convergence is determined, the entire closed-loop iterative assembly process ends, and the process proceeds to the assembly locking and verification steps. The fluctuation is the difference between the maximum and minimum RMS values ​​of the wavefront residuals in a consecutive preset number of rounds (for example, 3 consecutive rounds). The fluctuation tolerance is set as a certain percentage of the assembly and adjustment convergence threshold. For example, the fluctuation tolerance is 10% to 20% of the assembly and adjustment convergence threshold to ensure that the wavefront residual has stabilized rather than still oscillating when convergence is determined. When the root mean square value of the new wavefront residual is greater than the assembly and adjustment convergence threshold, or when the fluctuation between the root mean square values ​​of the wavefront residual in consecutive preset rounds is greater than the fluctuation tolerance, it is determined that the closed-loop iteration has not yet converged. The main control and calculation module returns the new real-time wavefront aberration data as input to step S11 to recalculate the target adjustment vector, and decides whether to directly execute the target adjustment vector or restart the piecewise linear compensation path construction process based on the dead zone crossover ratio determination result in step S12, and enter the next round of closed-loop iteration.

[0079] The endpoint overtaking and reverse retracement mechanism in step S20 and the piecewise linear compensation path mechanism in step S10 form a progressive relationship in improving assembly and adjustment accuracy. The piecewise linear compensation path in step S10 advances the lower limit of convergence for closed-loop iteration from the wavefront deviation magnitude corresponding to the hysteresis dead zone radius to the wavefront deviation magnitude corresponding to the stopping positioning accuracy of the adjustment stage servo controller, thus eliminating the hysteresis dead zone as a bottleneck to assembly and adjustment accuracy. The endpoint overtaking and reverse retracement mechanism in step S20 further advances the lower limit of convergence for closed-loop iteration from the wavefront deviation magnitude corresponding to the stopping positioning accuracy of the servo controller to the wavefront deviation magnitude corresponding to the position encoder reading resolution, thus eliminating the stopping positioning error of the servo controller as a bottleneck to assembly and adjustment accuracy. These two mechanisms successively eliminate two independent accuracy limiting factors on the closed-loop iteration convergence path, ensuring that the final achievable assembly and adjustment accuracy depends only on the measurement resolution of the wavefront sensor and the inversion residual of the aberration-misalignment mapping model. Both of these are accuracy indicators at the signal processing level and are not constrained by the non-ideal characteristics of the adjustment stage's mechanical transmission chain. The calculation and positioning locking process of the dynamic zero coordinate in step S20 provides a stable pose reference for the closed-loop iteration: at the end of each iteration, the adjustment stage is locked at the dynamic zero coordinate, and subsequent wavefront data acquisition is carried out under the state of stable adjustment stage position. The wavefront aberration data acquired by the wavefront sensor reflects the wavefront state corresponding to the pose of each optical element at the dynamic zero coordinate of the adjustment stage, rather than the transient wavefront state during the motion process. The acquisition reference of wavefront data is strictly corresponding to the position reference of the adjustment stage, avoiding data association mismatch caused by time asynchrony between wavefront data and position data under motion state. In step S20, the pre-sliding activation effect of the reverse motion process on the mechanical transmission chain of the adjustment platform is superimposed with the pre-sliding activation effect of the forward thrust vector and the lateral offset vector in step S10. After the adjustment platform completes a full adjustment process including a piecewise linear compensation path and an end-point overshoot and backsweep, each friction contact surface in the mechanical transmission chain undergoes micro-displacement activation in three dimensions: forward, lateral, and reverse. The transition from static friction coefficient to dynamic friction coefficient is more complete than when only the forward and lateral dimensions are activated. The temporary reduction of the hysteresis dead zone radius is greater. In subsequent iterations, the magnitude of the target adjustment vector can be directly executed within a larger range without activating the piecewise linear compensation path. The total number of iterations and the total time required for the closed-loop iteration to reach the convergence condition are further reduced.

[0080] Example 2:

[0081] This embodiment, based on Embodiment 1, provides an automated assembly and adjustment system for large-aperture optical systems based on real-time wavefront feedback, such as... Figure 7 As shown, it includes:

[0082] Wavefront calculation and compensation module: used to collect real-time wavefront aberration data of the current iteration, calculate the target adjustment vector; calculate the dead zone crossing ratio based on the target adjustment vector; when the dead zone crossing ratio is greater than the preset crossing threshold, the target adjustment vector is directly sent to the adjustment station for execution; when the dead zone crossing ratio is less than or equal to the crossing threshold, a three-segment piecewise linear compensation path containing the forward thrust vector, the lateral offset vector, and the regression vector is constructed, and the first segment displacement command corresponding to the forward thrust vector and the second segment displacement command corresponding to the lateral offset vector are sequentially sent to the adjustment station along the three-segment piecewise linear compensation path.

[0083] Dynamic zero-position locking module: After the second displacement command is executed, the endpoint of the regression vector is extended along the direction of the regression vector to generate an extended regression vector. The endpoint coordinates of the extended regression vector are marked as the overshoot coordinates. The extended third displacement command corresponding to the extended regression vector is sent to the adjustment table. When the adjustment table reaches the overshoot coordinates, the reverse displacement command is immediately sent, and the dynamic zero coordinates are calculated to lock the adjustment table at the dynamic zero coordinates.

[0084] Furthermore, in the wavefront compensation module, the method for calculating the target adjustment vector includes:

[0085] The wavefront residual is obtained by subtracting the real-time wavefront aberration data from the preset target wavefront data, and the target adjustment vector is calculated based on the wavefront residual.

[0086] The starting point of the target adjustment vector is the current position coordinate of the adjustment platform, and the ending point is the target position coordinate.

[0087] The method for calculating the dead zone pass-through ratio includes:

[0088] Read the value of the hysteresis dead zone radius, calculate the ratio of the magnitude of the target adjustment vector to the hysteresis dead zone radius, and define it as the dead zone crossing ratio.

[0089] The method for constructing the forward thrust vector is as follows:

[0090] An orthogonal auxiliary coordinate system is established with the current position coordinates of the adjustment platform as the origin, the direction of the target adjustment vector as the positive direction of the first coordinate axis, and the unit normal vector orthogonal to the target adjustment vector as the direction of the second coordinate axis; the second coordinate axis has two directions, positive and negative.

[0091] In the orthogonal auxiliary coordinate system, a forward thrust vector is generated along the positive direction of the first coordinate axis. The starting point of the forward thrust vector is the current position coordinate of the adjustment platform, and the ending point of the forward thrust vector is recorded as the coordinate of the first auxiliary point. The magnitude of the forward thrust vector is set to the first preset multiple of the hysteresis dead zone radius.

[0092] The method for constructing the lateral bias vector includes:

[0093] A lateral bias vector is generated along the second coordinate axis. The starting point of the lateral bias vector is the coordinate of the first auxiliary point, and the ending point of the lateral bias vector is the coordinate of the second auxiliary point. The magnitude of the lateral bias vector is set to the second preset multiple of the hysteresis dead zone radius.

[0094] The method for constructing the lateral bias vector also includes:

[0095] Read the upper and lower limits of the travel of each degree of freedom of the adjustment stage to form the travel range of the adjustment stage. Select the direction that ensures the coordinates of the second auxiliary point do not exceed the travel range of the adjustment stage as the direction of the lateral offset vector. When neither the positive nor negative direction will cause the coordinates of the second auxiliary point to exceed the travel range of the adjustment stage, select the direction that makes the coordinates of the second auxiliary point farther from the travel boundary. When both the positive and negative directions will cause the coordinates of the second auxiliary point to exceed the travel range of the adjustment stage, reduce the magnitude of the lateral offset vector. This magnitude is the maximum value that ensures the coordinates of the second auxiliary point do not exceed the travel range of the adjustment stage.

[0096] The methods and systems of this application may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the method is for illustrative purposes only, and the steps of the method of this application are not limited to the order specifically described above, unless otherwise specifically stated.

[0097] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0098] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An automatic assembly and adjustment method for large-aperture optical systems based on real-time wavefront feedback, characterized in that, The method includes: Collect real-time wavefront aberration data for the current iteration and calculate the target adjustment vector; read the hysteresis dead zone radius value and calculate the ratio of the magnitude of the target adjustment vector to the hysteresis dead zone radius, which is defined as the dead zone crossing ratio; when the dead zone crossing ratio is greater than the preset crossing threshold, the target adjustment vector is directly sent to the adjustment station for execution; when the dead zone crossing ratio is less than or equal to the crossing threshold, a three-segment piecewise linear compensation path containing the forward thrust vector, the lateral offset vector, and the regression vector is constructed, and the first segment displacement command corresponding to the forward thrust vector and the second segment displacement command corresponding to the lateral offset vector are sequentially sent to the adjustment station along the three-segment piecewise linear compensation path. After the second displacement command is executed, the endpoint of the regression vector is extended along the direction of the regression vector to generate an extended regression vector. The endpoint coordinates of the extended regression vector are marked as the overshoot coordinates. The extended third displacement command corresponding to the extended regression vector is sent to the adjustment platform. When the adjustment platform reaches the overshoot coordinates, the reverse displacement command is immediately sent, and the dynamic zero coordinates are calculated to lock the adjustment platform at the dynamic zero coordinates.

2. The automatic assembly and adjustment method for a large-aperture optical system based on real-time wavefront feedback according to claim 1, characterized in that, The method for calculating the target adjustment vector includes: The wavefront residual is obtained by subtracting the real-time wavefront aberration data from the preset target wavefront data, and the target adjustment vector is calculated based on the wavefront residual. The starting point of the target adjustment vector is the current position coordinate of the adjustment platform, and the ending point is the target position coordinate.

3. The automatic assembly and adjustment method for a large-aperture optical system based on real-time wavefront feedback according to claim 2, characterized in that, The method for constructing the forward thrust vector is as follows: An orthogonal auxiliary coordinate system is established with the current position coordinates of the adjustment platform as the origin, the direction of the target adjustment vector as the positive direction of the first coordinate axis, and the unit normal vector orthogonal to the target adjustment vector as the direction of the second coordinate axis; the second coordinate axis has two directions, positive and negative. In the orthogonal auxiliary coordinate system, a forward thrust vector is generated along the positive direction of the first coordinate axis. The starting point of the forward thrust vector is the current position coordinate of the adjustment platform, and the ending point of the forward thrust vector is recorded as the coordinate of the first auxiliary point. The magnitude of the forward thrust vector is set to the first preset multiple of the hysteresis dead zone radius.

4. The automatic assembly and adjustment method for a large-aperture optical system based on real-time wavefront feedback according to claim 3, characterized in that, The method for constructing the lateral bias vector includes: A lateral bias vector is generated along the second coordinate axis. The starting point of the lateral bias vector is the coordinate of the first auxiliary point, and the ending point of the lateral bias vector is the coordinate of the second auxiliary point. The magnitude of the lateral bias vector is set to the second preset multiple of the hysteresis dead zone radius.

5. The automatic assembly and adjustment method for a large-aperture optical system based on real-time wavefront feedback according to claim 4, characterized in that, The method for constructing the lateral bias vector also includes: Read the upper and lower limits of the travel of each degree of freedom of the adjustment stage to form the travel range of the adjustment stage. Select the direction that ensures the coordinates of the second auxiliary point do not exceed the travel range of the adjustment stage as the direction of the lateral offset vector. When neither the positive nor negative direction will cause the coordinates of the second auxiliary point to exceed the travel range of the adjustment stage, select the direction that makes the coordinates of the second auxiliary point farther from the travel boundary. When both the positive and negative directions will cause the coordinates of the second auxiliary point to exceed the travel range of the adjustment stage, reduce the magnitude of the lateral offset vector. This magnitude is the maximum value that ensures the coordinates of the second auxiliary point do not exceed the travel range of the adjustment stage.

6. The automatic assembly and adjustment method for a large-aperture optical system based on real-time wavefront feedback according to claim 4, characterized in that, The regression vector points from the coordinates of the second auxiliary point to the coordinates of the target position; the regression vector is calculated based on the difference between the coordinates of the second auxiliary point and the coordinates of the target position.

7. The automatic assembly and adjustment method for a large-aperture optical system based on real-time wavefront feedback according to claim 2, characterized in that, The method for calculating the dynamic zero-position coordinates includes: During the execution of the reverse displacement command, the real-time position reading of the position encoder is read periodically, the position deviation signal between the real-time position reading and the target position coordinate is calculated, the zero-crossing time is determined based on the position deviation signal, the position encoder readings in the sampling periods adjacent to the zero-crossing time are linearly interpolated, and the interpolated position coordinates corresponding to the position deviation signal being equal to zero are taken as the dynamic zero coordinates.

8. The automatic assembly and adjustment method for a large-aperture optical system based on real-time wavefront feedback according to claim 7, characterized in that, The method for determining the zero-crossing time based on the position deviation signal is as follows: The zero-crossing moment is determined when the sign of the position deviation signal reverses between two adjacent sampling periods.

9. An automatic assembly and adjustment system for a large-aperture optical system based on real-time wavefront feedback, used to implement the automatic assembly and adjustment method for a large-aperture optical system based on real-time wavefront feedback as described in any one of claims 1-8, characterized in that, The system includes: Wavefront calculation and compensation module: used to collect real-time wavefront aberration data of the current iteration, calculate the target adjustment vector; calculate the dead zone crossing ratio based on the target adjustment vector; when the dead zone crossing ratio is greater than the preset crossing threshold, the target adjustment vector is directly sent to the adjustment station for execution; when the dead zone crossing ratio is less than or equal to the crossing threshold, a three-segment piecewise linear compensation path containing the forward thrust vector, the lateral offset vector, and the regression vector is constructed, and the first segment displacement command corresponding to the forward thrust vector and the second segment displacement command corresponding to the lateral offset vector are sequentially sent to the adjustment station along the three-segment piecewise linear compensation path. Dynamic zero-position locking module: After the second displacement command is executed, the endpoint of the regression vector is extended along the direction of the regression vector to generate an extended regression vector. The endpoint coordinates of the extended regression vector are marked as the overshoot coordinates. The extended third displacement command corresponding to the extended regression vector is sent to the adjustment table. When the adjustment table reaches the overshoot coordinates, the reverse displacement command is immediately sent, and the dynamic zero coordinates are calculated to lock the adjustment table at the dynamic zero coordinates.

Images