A paper self-adaptive based calligraphy robot writing control method and system

Abstract

The present application relates to the technical field of calligraphy robot, and particularly relates to a calligraphy robot writing control method and system based on paper self-adaptation. The method comprises a paper ink absorption model construction step, a control mechanical arm driving brush pen to perform stepwise pressure pressing in the blank area of the paper, collecting the pressure value, pressing duration and final stable ink width of each pressing, fitting to obtain the ink absorption coefficient and constructing the paper ink absorption model; a writing control step, based on the model to perform writing control: force sense stable pressure control dynamically adjusts the tip pressure according to the preset target pressure and the actual pressure; visual calibration control dynamically corrects the preset target pressure according to the actual stroke width, and uses the corrected target pressure as the input of the force sense stable pressure control; model prediction control predicts the ink diffusion width based on the paper ink absorption model, and actively adjusts the tip pressure in advance when the prediction exceeds the preset width boundary. The setting can automatically adapt to paper with different ink absorption characteristics and realize accurate writing.

Description

Technical Field

[0001] This invention relates to the field of calligraphy robot technology, and in particular to a paper-adaptive calligraphy robot writing control method and system. Background Technology

[0002] As an intangible cultural heritage, Chinese calligraphy's artistic value lies not only in the precision of its strokes but also in the dynamic control of brush tip pressure to achieve variations in ink application—the thickness of the brush tip and the density of the ink both depend on the writer's real-time control of the pressure applied between the brush tip and the paper. One of the core challenges for calligraphy robots to replicate this artistic effect is adapting to the ink absorption characteristics of different papers: raw Xuan paper has loose fibers, absorbs ink quickly, and spreads significantly, while cardboard has dense fibers, absorbs ink slowly, and spreads very little. Under the same pressure, the ink diffusion effect of different papers can differ by several times. Most existing calligraphy robots rely on a fixed pressure parameter library that is manually adjusted. After changing paper, the parameters need to be manually recalibrated, which is cumbersome and inefficient, failing to meet the needs of flexible creation.

[0003] Furthermore, the control logic of existing calligraphy robots generally suffers from lag and fragmentation. Most systems rely solely on preset force control curves or simple post-visual feedback, resulting in delayed control response when ink is about to smudge, making it difficult to avoid bloated and distorted strokes. At the same time, there is a lack of effective coordination between visual detection and force control, leading to insufficient pressure control precision and a mechanical and stiff writing effect that is far removed from the artistic effect of human writing. Summary of the Invention

[0004] To address at least one deficiency in the existing calligraphy robot writing control technology, this invention provides a paper-adaptive calligraphy robot writing control method, comprising the following steps:

[0005] The paper ink absorption model construction steps involve controlling a robotic arm to drive a brush to perform stepped pressure pressing on a blank area of ​​the paper, collecting and recording the pressure value, pressing duration, and final stable ink width data for each pressing; based on the collected multiple sets of pressure values, pressing duration, and ink width data, an ink absorption coefficient reflecting the paper's ink absorption characteristics is fitted to construct a paper ink absorption model characterizing the relationship between pressure, time, and ink width.

[0006] The writing control step involves performing writing control based on the paper ink absorption model, and the writing control includes:

[0007] Force-sensing pressure control dynamically adjusts the pressure at the tip of the brush based on the preset target pressure and the actual pressure fed back in real time by the force sensing module, in order to maintain stable pressure.

[0008] Visual calibration control extracts the actual stroke width from the stroke image acquired in real time by the visual perception module, compares it with the preset target width, dynamically corrects the preset target pressure based on the comparison result, and uses the corrected preset target pressure as the input of the force-sensing pressure stabilization control.

[0009] The model predictive control, based on the paper ink absorption model, predicts the ink spread width within a preset time period according to the current actual pressure. When the predicted ink spread width exceeds the preset width boundary, the pen tip pressure is proactively adjusted in advance to intervene before ink smudging occurs.

[0010] This invention also provides a paper-adaptive calligraphy robot writing control system, comprising:

[0011] A robotic arm used to drive a calligraphy brush to perform writing actions;

[0012] The force sensing module is used to collect the pressure value between the pen tip and the paper in real time;

[0013] The visual perception module is used to acquire images of ink stains on paper in real time;

[0014] The main control unit is connected to the robotic arm, the force sensing module, and the vision sensing module, respectively. The main control unit includes:

[0015] The paper ink absorption model construction module is used to control the robotic arm to drive the brush to perform stepped pressure pressing on the blank area of ​​the paper, collect and record the pressure value, pressing duration and final stable ink width data of each pressing; based on the collected multiple sets of pressure values, pressing duration and ink width data, the ink absorption coefficient reflecting the ink absorption characteristics of the paper is fitted to construct a paper ink absorption model characterizing the relationship between pressure, time and ink width.

[0016] The force-sensing pressure control module is used to dynamically adjust the robotic arm to stabilize the pen tip pressure based on the preset target pressure and the actual pressure fed back by the force-sensing module.

[0017] The visual calibration control module is used to extract the actual stroke width from the stroke image acquired in real time by the visual perception module, compare it with the preset target width, dynamically correct the preset target pressure according to the comparison result, and use the corrected preset target pressure as the input of the force-sensing pressure stabilization control.

[0018] The model prediction control module is used to predict the ink spread width within a preset time period based on the paper ink absorption model and the current actual pressure. When the predicted ink spread width exceeds the preset width boundary, the pen tip pressure is proactively adjusted in advance to intervene before the ink smudges occur.

[0019] Based on the above, the paper-adaptive calligraphy robot writing control method and system provided by this invention, compared with the prior art, automatically identifies the ink absorption characteristics of paper and establishes an ink absorption model through the paper ink absorption model construction step, realizing adaptive adaptation to different papers without manual intervention; through the dual closed-loop structure composed of force-sensory pressure stabilization control and visual calibration control, it realizes the coordinated correction of pressure stability and visual feedback, ensuring accurate matching of stroke width; through model predictive control, it makes forward prediction of ink diffusion trend and actively adjusts pressure in advance, intervening before ink bleeding occurs, effectively avoiding the lag problem of traditional feedback control, and significantly improving the adaptability and artistic expression of calligraphy robot writing.

[0020] Other features and beneficial effects of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other beneficial effects of the invention can be realized and obtained by means of the structures particularly pointed out in the description, claims and drawings. Attached Figure Description

[0021] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Unless otherwise specified, the positional relationships shown in the drawings in the following description are based on the direction in which the components are drawn in the figure.

[0022] Figure 1 A flowchart illustrating the steps of a paper-adaptive calligraphy robot writing control method according to an embodiment of the present invention;

[0023] Figure 2 A flowchart of the writing control steps for a calligraphy robot;

[0024] Figure 3 A comparison diagram of the traditional calligraphy robot writing method and the one-stroke writing method of this invention;

[0025] Figure 4 A comparative curve of the relationship between ink spread width and pressure on different papers under different ink absorption coefficients;

[0026] Figure 5 Illustration of ink absorption effect on raw Xuan paper and cardstock;

[0027] Figure 6 This is a structural block diagram of a paper-adaptive calligraphy robot writing control system provided in an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all embodiments. The technical features designed in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0029] In the description of this invention, it should be noted that all terms used in this invention (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains, and should not be construed as limiting the invention; it should be further understood that the terms used in this invention should be understood to have the same meaning as those in the context of this specification and in the relevant field, and should not be understood in an idealized or overly formal sense, except as expressly defined in this invention.

[0030] Existing calligraphy robots mostly rely on fixed pressure parameter libraries that are manually calibrated. After changing paper, these parameters need to be manually recalibrated, which is cumbersome and has extremely low adaptation efficiency. Furthermore, existing control logic generally suffers from lag and fragmentation. Most systems rely solely on preset force control curves or simple post-visual feedback, resulting in delayed control response when ink is about to bleed, making it difficult to avoid bloated and distorted strokes. There is a lack of effective coordination between visual detection and force control; pressure adjustment cannot be dynamically adjusted according to real-time ink diffusion, and there is no ability to predict ink diffusion trends. The writing effect is mechanical and stiff, far removed from the artistic effect of human writing. To address the problems of the existing technology, this invention provides a paper-adaptive calligraphy robot writing control method and system. It automatically identifies the ink absorption characteristics of paper through a paper ink absorption model construction step, establishes a dual-closed-loop collaborative control through force-based pressure stabilization and visual calibration, and combines model predictive control to achieve proactive intervention in ink diffusion, thereby significantly improving the calligraphy robot's adaptability to different types of paper and its artistic expression.

[0031] The technical solution of the present invention will be described and explained in detail below with reference to specific embodiments and accompanying drawings through various specific implementation methods.

[0032] Example 1

[0033] Please see Figure 1This embodiment provides a paper-adaptive calligraphy robot writing control method, which includes at least a robotic arm, a force sensing module, a vision sensing module, and a main control unit. The robotic arm drives the brush to perform writing actions; the force sensing module collects the pressure value of the brush tip contacting the paper in real time; the vision sensing module collects images of ink stains on the paper in real time; and the main control unit is connected to the robotic arm, the force sensing module, and the vision sensing module to execute the writing control method. Preferably, the force sensing module can be a six-dimensional force / torque sensor, installed between the end of the robotic arm and the brush holding device to collect pressure values ​​in real time; the vision sensing module can be an industrial camera, mounted above the writing area to collect images in real time.

[0034] In this embodiment, the control method includes a paper ink absorption model construction step and a writing control step.

[0035] The paper ink absorption model construction step is used to control the robotic arm to drive the brush to perform stepped pressure pressing on the blank area of ​​the paper, collect and record the pressure value, pressing duration and final stable ink width data of each pressing; based on the collected multiple sets of pressure values, pressing duration and ink width data, the ink absorption coefficient reflecting the ink absorption characteristics of the paper is fitted to construct a paper ink absorption model characterizing the relationship between pressure, time and ink width.

[0036] In practice, the first step is to construct a paper ink absorption model. The main control unit controls a robotic arm to drive a brush to perform stepped pressure presses on the blank area (non-writing area) of the paper. The pressure gradient can be set from 0.5N to 5N. At each pressure level, the brush is held for a preset time (e.g., 1 to 2 seconds) before being lifted to ensure that the ink spread reaches a stable state. During this process, the force sensing module records the actual pressure value of each press in real time, the visual sensing module simultaneously collects the entire ink spread process and records the final stable ink width (diameter for circular ink marks, average width for stripe ink marks), and the main control unit records the duration of each press. Based on the collected data of multiple sets of pressure values, press durations, and ink widths, the main control unit uses a fitting algorithm to obtain the ink absorption coefficient reflecting the paper's ink absorption characteristics, thereby constructing a paper ink absorption model that characterizes the relationship between pressure, time, and ink width. The fitting algorithm can employ, but is not limited to, regression analysis methods such as linear regression, weighted least squares, ridge regression, and principal component regression, or machine learning methods such as support vector regression and shallow neural networks. This embodiment uses the linear regression algorithm as an example for illustration, but it does not constitute a limitation on the type of fitting algorithm.

[0037] Through this step, the system can automatically identify the ink absorption characteristics of different papers such as raw Xuan paper, cardstock, and rough-edged paper before writing, and generate a unique ink absorption model without manual intervention, providing accurate parameter basis for subsequent writing control.

[0038] The writing control step is used to perform writing control based on the paper ink absorption model, the writing control including:

[0039] Force-sensory pressure stabilization control dynamically adjusts the brush tip pressure based on a preset target pressure and the actual pressure fed back in real time by the force-sensory perception module to maintain stable pressure. Visual calibration control extracts the actual stroke width from the stroke image collected in real time by the visual perception module, compares it with the preset target width, dynamically corrects the preset target pressure based on the comparison result, and uses the corrected preset target pressure as the input for force-sensory pressure stabilization control. Model prediction control, based on the paper ink absorption model, predicts the ink spread width within a preset time period based on the current actual pressure. When the predicted ink spread width exceeds the preset width boundary, the brush tip pressure is proactively adjusted in advance to intervene before ink smudging occurs.

[0040] In specific implementation, such as Figure 2 As shown, after constructing the paper ink absorption model, the writing control step begins. This step comprises three collaborative stages: force-based pressure stabilization control, visual calibration control, and model predictive control. Force-based pressure stabilization control uses a preset target pressure as a benchmark. The force sensing module provides real-time feedback on the actual contact pressure between the pen tip and the paper. The main control unit uses a PID algorithm to dynamically adjust the Z-axis displacement of the robotic arm. When the actual pressure deviates from the preset target pressure, the system quickly outputs an adjustment amount through the collaborative calculation of proportional, integral, and derivative terms, precisely stabilizing the pen tip pressure near the preset target pressure, effectively counteracting external disturbances such as uneven paper and brush deformation. Visual calibration control involves the visual sensing module acquiring stroke images in real-time at a frame rate of at least 30fps during the writing process. An image processing algorithm extracts the actual width of the current stroke and compares it with the preset target width. If the actual width is too thin, the preset target pressure is dynamically increased; if the actual width is too thick, the preset target pressure is decreased or a pen-lifting operation is performed. The corrected preset target pressure is used as the input for force-based pressure stabilization control, forming a closed-loop collaborative control of "visual guidance and force execution." Model predictive control is based on an established paper ink absorption model. Within each control cycle, it predicts the ink spread width within a preset time period (e.g., 0.05 to 0.1 seconds) based on the current actual pressure. When the prediction indicates that maintaining the current pressure will cause the ink spread width to exceed the preset width boundary, the system does not wait for visual feedback but immediately and proactively adjusts the pen tip pressure in advance. This avoids the problem of ink bleeding beyond the stroke boundary from the source, achieving proactive control that "prevents problems before they occur."

[0041] Please see Figure 3 Traditional calligraphy robots require outlining the strokes and filling in the blanks before writing. However, this embodiment achieves continuous stroke writing with a brush by using the synergistic effect of force-sensing voltage regulation control, visual calibration control, and model prediction control. This eliminates the need for complex outline and filling processes, resulting in higher writing efficiency and more natural and smooth strokes.

[0042] Optionally, in the paper ink absorption model construction step, the pressure gradient of the stepped pressure pressing is 0.5N to 5N, and the pen is lifted after each press for a preset time; the pressure value, the pressing duration and the ink width data are fitted using a linear regression algorithm to obtain the ink absorption coefficient.

[0043] In practice, the pressure gradient can start from 0.5N and gradually increase to 5N, i.e., the pressure gradient levels are set to 0.5N, 1N, 1.5N, 2N, 2.5N, 3N, 3.5N, 4N, 4.5N, and 5N, for a total of 10 pressure levels. At each pressure level, the brush is held in a pressed state for a preset time (e.g., 1.5 seconds) before being lifted to ensure that the ink has fully diffused and reached a stable state. Of course, the specific pressure gradient levels can be reasonably set according to actual needs, and this embodiment does not limit them. The linear regression algorithm can effectively establish a linear relationship between pressure value, pressing duration, and ink width. By minimizing the sum of squared errors to solve for the optimal fitting parameters, the ink absorption coefficient K value, which reflects the ink absorption characteristics of the paper, is obtained. Experiments have verified that the K value range for raw Xuan paper is 0.6~0.9, and the K value range for cardboard paper is 0.05~0.2. This range can be used as a reference for determining the paper type.

[0044] Optionally, the linear regression algorithm employs the least squares method, and the expression for calculating the ink absorption coefficient is as follows:

[0045]

[0046] In the formula, For the first The pressure value of each press. For the first The duration of each press. For the first The stable ink width corresponding to each press. for The average value, for The average value; that is, , .

[0047] In practice, the least squares method calculates the ink absorption coefficient K by determining the ratio of covariance to variance, ensuring the optimality of the fitting result. That is, the covariance term... This reflects the degree of linear correlation between the pressure-time product and the ink width, with the variance term... This reflects the dispersion of the pressure-time product, and the ratio of the two is the optimal fitting coefficient in the least squares sense. Mathematically, this method guarantees the minimization of the sum of squared fitting errors, thus obtaining the coefficient that best reflects the ink absorption characteristics of the paper. value.

[0048] Taking raw Xuan paper as an example, during the stepped pressing process, the ink diffusion is not obvious under low pressure (e.g., 0.5N), while under high pressure (e.g., 5N), the ink diffuses rapidly to form a larger ink blot. A strong linear relationship exists between the pressure-time product and the ink width. After collecting 10 sets of pressing data (pressure gradient range set to 0.5N~5N, pressing duration fixed at 1.5 seconds), the value was calculated using the above formula to obtain K≈0.75. This value is within the preset K value range (0.6~0.9) for raw Xuan paper, indicating that raw Xuan fibers are loose, absorb ink quickly, and diffuse ink significantly, verifying its strong ink absorption characteristics. Taking cardboard as an example, under the same pressure, the ink diffusion range is much smaller than that of raw Xuan paper. The linear relationship between the pressure-time product and the ink width is also significant, but the slope is significantly reduced. After collecting 10 sets of data using the same procedure and substituting them into the formula, K≈0.12 was obtained, which is within the K value range (0.05~0.2) for cardboard. This indicates that the cardboard fibers are dense, the ink absorption speed is slow, and the ink diffusion range is small, verifying its weak ink absorption characteristics. Please refer to [link / reference]. Figure 4 Comparison curves of ink spread width versus pressure for different papers with varying ink absorption coefficients reveal a distinct phased characteristic in ink spread width changes with pen tip pressure: when pressure is in the low-pressure suppression zone, ink spread is inhibited, and width increases slowly; when pressure reaches the critical breakthrough point, ink begins to effectively penetrate the paper fibers; after entering the rapid spread zone, ink spread width increases approximately linearly with increasing pressure, this region corresponds to the main working range of the paper ink absorption model described in this embodiment; when pressure is too high and enters the saturation-ultra-slow zone, ink spread tends to saturate, and width increases extremely slowly. Therefore, this application limits the pressure range of stepped pressure application and subsequent writing control to the rapid spread zone of 0.5N to 5N to ensure the linear fitting accuracy of the ink absorption model and the effectiveness of writing control. Simultaneously, the comparison curves show that the larger the ink absorption coefficient, the more significant the ink spread, verifying the rationality of quantifying paper characteristics through the ink absorption coefficient in this embodiment. Using the above formula, only a limited number of pressing data sets are needed to complete the quantitative characterization of paper characteristics within seconds. Compared with the traditional method that relies on repeated adjustments based on human experience, this design significantly improves the adaptation efficiency and intelligence level between different types of paper.

[0049] Optionally, in the visual calibration control, extracting the actual stroke width includes: sequentially performing grayscale conversion, binarization, contour detection, and minimum circumcircle fitting processing on the acquired stroke image, and obtaining the actual stroke width through the diameter of the minimum circumcircle.

[0050] In specific implementation, the process for extracting the actual stroke width is as follows: First, the original stroke image acquired by the visual perception module is converted to grayscale. Grayscale conversion converts the color image into a single-channel grayscale image, preserving the brightness difference information between the stroke and the background while compressing the data volume to one-third of the original, effectively reducing the subsequent computational complexity and improving processing speed. After grayscale conversion, a binarization algorithm is used to convert the grayscale image into a black and white binary image, achieving separation of the stroke and the background. This embodiment uses the OTSU (Maximum Inter-Class Variance) algorithm to adaptively determine the binarization threshold. This algorithm searches for the threshold that maximizes the variance between the foreground and background pixels by traversing the grayscale levels as the optimal segmentation point. The OTSU algorithm does not require manual threshold setting and can adapt to image changes under different lighting conditions and different paper colors, exhibiting good adaptability and robustness. After binarization, the stroke region appears as a connected set of foreground pixels. The contour detection algorithm traverses the binary image and extracts the closed contour of each stroke region by tracing the boundaries of the foreground pixels. For strokes that may have internal voids (such as tiny blanks caused by uneven ink distribution), the contour detection algorithm can simultaneously extract both the outer and inner contours, ensuring a complete representation of the stroke's shape. After contour extraction, a minimum circumcircle fitting algorithm is used to calculate the minimum circumcircle that completely contains the contour. This algorithm iteratively optimizes the center position and radius to minimize the radius and ensure that all points on the contour are either inside or on the circle. Since brushstrokes typically present an approximately circular ink mark shape during writing, and the stroke width is defined as the span of the ink mark perpendicular to the stroke's direction, the diameter of the minimum circumcircle objectively reflects this span, thus serving as the output width value of the current stroke. This image processing workflow can quickly and accurately extract stroke widths, providing reliable feedback data for visual calibration control. In practical applications, the accuracy of stroke width extraction can reach the pixel level, accurately capturing subtle changes in stroke width and meeting the precision requirements for width control in calligraphy writing.

[0051] It should be noted that the grayscale conversion, OTSU binarization, contour detection, and minimum circumcircle fitting processes used in this embodiment are merely examples of preferred image processing procedures. Those skilled in the art can reasonably adjust or replace the above processing steps in practical applications based on specific conditions such as hardware performance, lighting environment, paper color, and ink density. For example, in addition to OTSU, the binarization algorithm can also employ adaptive threshold segmentation, globally fixed threshold segmentation, or semantic segmentation based on deep learning; contour detection can use the Canny operator based on edge detection followed by contour tracking; and the stroke width can be represented not only by the minimum circumcircle diameter but also by the minimum circumcircle rectangle width or the statistical value of the stroke skeleton normal distance. All techniques that use a visual perception module to acquire stroke images and extract stroke widths fall within the protection scope of this invention.

[0052] Optionally, in the model predictive control, a predictive model is established based on the paper ink absorption model, the optimal pressure adjustment amount that satisfies the ink diffusion constraint is solved by rolling optimization, and the paper ink absorption model is corrected by the real-time feedback of the force perception module and the visual perception module.

[0053] In practice, the predictive model is built upon an established paper ink absorption model, using the current actual pressure as input to predict the ink spread width at future moments. Rolling optimization solves a constrained optimization problem within each control cycle, aiming to minimize the deviation between the predicted ink width and the preset target width, while simultaneously controlling the pressure adjustment to maintain writing stability. Feedback correction is a crucial part of the model predictive control. The system utilizes the actual pressure value fed back by the force perception module and the actual stroke width fed back by the visual perception module to correct the ink absorption model parameters online, continuously improving the model's prediction accuracy over time and enhancing the system's adaptability to changes in paper characteristics and ink concentration.

[0054] Optionally, in the model predictive control, the proactive adjustment of the pen tip pressure includes: predicting the ink spread width within a preset time domain based on the paper ink absorption model in each control cycle. A cost function is constructed that includes a weighted sum of tracking error and control energy. The solution is then used to minimize the cost function while satisfying the ink spread width requirement. The optimal pressure adjustment sequence, which does not exceed the preset width boundary constraint, is executed only in the first step of the sequence as the output of the current cycle. At the beginning of the next control cycle, the paper ink absorption model is corrected using force and visual feedback information, and the optimal pressure adjustment sequence is solved again with the corrected model.

[0055] In practical implementation, the core of model predictive control lies in the optimization strategy of the "rolling time domain". Taking a sampling time of 0.05 seconds as an example, the prediction time domain is set to 5 steps (i.e., predicting the ink spread trend within the next 0.25 seconds), and the control time domain is set to 2 steps (i.e., solving for the optimal control quantity for the next 2 steps). The cost function can be designed as follows:

[0056]

[0057] In the formula, This is the cost function, used to evaluate the effectiveness of control; the smaller the value, the better the control effect. For the future prediction based on the paper ink absorption model Step ink spread width; The target stroke width (i.e., the desired stroke width, which can be preset according to the actual writing font and artistic effect); For the future The pressure adjustment step is indicated by a positive value, which represents an increase in pressure, and a negative value, which represents a decrease in pressure. In the time domain, it represents the number of steps for forward prediction; This is the step index, with values ​​ranging from 0 to... ,in, Corresponding to the current moment; and These are the weighting coefficients for tracking error and control energy, respectively. "and" "represents the weighted coefficients respectively" and The norm square operation. The cost function contains two terms, where... To track the error term and reflect the deviation between the predicted ink width and the target width, the controller is prompted to make the ink width as close as possible to the target value. To control the energy term and reflect the intensity of pressure regulation, the controller is instructed to smoothly adjust the pressure and avoid drastic fluctuations. This is achieved by adjusting the weighting coefficients. and The ratio allows for a flexible balance between stroke width accuracy and control stability.

[0058] When solving the above cost function minimization problem, it is also necessary to meet the preset constraints, including but not limited to pressure safety range constraints (to prevent the strokes from being discontinuous due to insufficient pressure or the paper from being damaged by excessive pressure), adjustment rate constraints (to prevent stroke jitter caused by sudden pressure changes), and ink diffusion constraints (to ensure that the predicted ink width does not exceed the preset width boundary and avoid smudging), etc. The specific constraints should be selected reasonably according to the actual needs.

[0059] After solving the constrained optimization problem and obtaining the optimal pressure adjustment sequence, the system executes only the first step of the sequence (i.e., the pressure adjustment amount for the current cycle) as the output for the current cycle. At the beginning of the next cycle, the model is corrected using actual data from force and visual feedback, and then the optimization problem is solved again. This "execute one step, roll one step" strategy ensures both optimal control and timely response to changes in actual operating conditions, achieving precise and robust pressure control.

[0060] Optionally, the force-sensing pressure control employs a PID algorithm, and the target pressure of the PID algorithm is dynamically corrected by the visual calibration control based on the deviation between the actual stroke width and the preset target width.

[0061] In practical implementation, the PID algorithm achieves rapid and accurate tracking of the target pressure through the coordinated action of proportional, integral, and derivative components, calculating the pressure deviation in real time and outputting the adjustment amount. Specifically, the force-sensing pressure regulation control uses an incremental PID algorithm to dynamically adjust the pen tip pressure, and its control formula is as follows:

[0062]

[0063]

[0064] In the formula, For the current number The increment of the control quantity at any given time is the adjustment amount of the Z-axis displacement of the robotic arm; For the current number Pressure error at any time, , To preset the target pressure, The actual pressure collected by the force sensing module; For the first Pressure error at any given moment; For the first Pressure error at any given moment; This is a proportionality coefficient used for rapid response to pressure deviations; It is the integral coefficient, used to eliminate steady-state errors and counteract disturbances such as uneven paper and brush deformation; The differential coefficient is used to suppress overshoot and oscillation during pressure regulation. By dynamically controlling the minute displacement of the robotic arm's Z-axis through this incremental PID algorithm, the pen tip pressure can be quickly and accurately stabilized near the preset target pressure, providing a stable force feedback basis for writing control.

[0065] Among them, the proportionality coefficient The proportional gain determines the speed of pressure response; the larger the proportional gain, the faster the system responds to pressure deviations. However, an excessively large proportional gain may lead to overshoot and oscillation. In this embodiment, a proportional gain of [missing value] is preferred. The integral coefficient is 0.8~1.2. To eliminate steady-state errors, compensation is provided for the accumulation of pressure deviations to ensure that the actual pressure converges to the target value without steady-state error. This embodiment preferably uses an integral coefficient. The value is 0.1 to 0.3; the differential coefficient. To suppress overshoot and oscillation, predictive compensation is performed on the rate of change of pressure deviation, and reverse regulation is applied in advance to improve the dynamic stability of the system. In this embodiment, the differential coefficient is preferred. The value is 0.05~0.15. In this embodiment, the target pressure of the PID algorithm is not a fixed preset value, but is dynamically corrected by the visual calibration control based on the deviation between the actual stroke width and the preset target width. In practical applications, this parameter can be fine-tuned according to the dynamic response characteristics of the robotic arm and the flatness of the paper surface.

[0066] Optionally, in the visual calibration control, the step of dynamically correcting the preset target pressure based on the comparison result includes: if the actual stroke width is less than the preset target width, then increasing the preset target pressure; if the actual stroke width is greater than the preset target width, then decreasing the preset target pressure or performing a pen lifting operation.

[0067] In practice, visual calibration control extracts the actual stroke width from the real-time acquired stroke images. and compare it with the preset target width A comparison is made. The specific correction strategy for its visual calibration control is as follows: When < If the pressure is insufficient or the ink spread is inadequate, the system will gradually increase the preset target pressure in increments of 0.1N to 0.2N until the stroke width is close to the target pressure. ;when > If the pressure is too high or smudging is imminent, the system will reduce the preset target pressure in increments of 0.2N to 0.3N, or immediately perform a small pen lift (e.g., 0.1mm) followed by a brief pause (e.g., 0.05 seconds) before resuming the pen stroke to effectively suppress further ink spread. The corrected preset target pressure serves as the new input for force-sensitive pressure regulation control. The PID algorithm then tracks the new target pressure and adjusts the pressure accordingly, changing the pen tip's downward pressure depth by adjusting the Z-axis displacement of the robotic arm, thus rapidly bringing the actual pressure close to the corrected target value. This correction strategy can quickly respond to dynamic changes in stroke width, ensuring that the stroke width remains within the target range throughout the writing process.

[0068] Through the above setup, a cascaded control structure of "visual calibration and force-sensory execution" is formed. Specifically, visual calibration control dynamically corrects the control target (preset target pressure) based on the writing result (stroke width), while force-sensory pressure regulation control is responsible for accurately tracking the corrected target pressure. Together, they achieve a closed-loop control architecture of outer-loop calibration and inner-loop execution. This design effectively solves the problems of single force-sensory control's inability to perceive ink diffusion effects and single visual control's response lag. It ensures both the speed and stability of pressure adjustment and the precise matching of stroke width with target requirements. Experiments show that after adopting this collaborative control strategy, the control error of stroke width can be controlled within ±0.1mm, significantly improving the calligraphy robot's adaptability and writing accuracy to different types of paper and stroke shapes.

[0069] Optionally, the formula for the paper ink absorption model is: In the formula, The width of the ink mark. The ink absorption coefficient is... For pen tip pressure, The duration of pressing or writing.

[0070] In practical implementation, this model simplifies the complex physical processes between pressure, time, and ink width through a linear relationship. The model is based on the following physical mechanism: when the brush tip contacts the paper, ink penetrates the paper fibers under pressure; the ink diffusion area is positively correlated with the pressure and the duration of contact; and the ink absorption coefficient... This comprehensively reflects the influence of material properties such as paper fiber density, surface roughness, and ink flow on ink diffusion.

[0071] The model was experimentally verified through a stepped pressure test, in which multiple sets of pressure values ​​were collected within a pressure range of 0.5N to 5N. Duration of press and the corresponding stable ink width The ink absorption coefficient obtained by least squares fitting After being substituted into the model, the fitting error between the model's predicted value and the measured data is less than 5%, indicating that the linear model has high fitting accuracy and engineering applicability within this pressure range.

[0072] It should be noted that when the pressure exceeds this range (e.g., ink diffusion is unstable when below 0.5N, and may exceed the paper's ink absorption saturation zone when above 5N), the ink width and pressure may exhibit a non-linear relationship. However, the pressure range of the stepped pressure pressing and subsequent writing control described in this invention is limited to the linear range of 0.5N to 5N. Therefore, this linear model is sufficient to meet the accuracy requirements.

[0073] Taking raw Xuan paper as an example, the ink absorption coefficient was obtained through pre-detection and fitting.K ≈0.75, at the nib pressure F =2.0 N and the pressing duration T =0.2 s, the model predicts the ink width D =0.75×2.0×0.2 = 0.3 mm. The actual measured value is about 0.29 mm to 0.31 mm. The prediction error is within a reasonable range, verifying the characteristics of strong ink absorption and significant ink diffusion of raw rice paper. Taking cardboard as an example, the ink absorption coefficient is obtained by fitting K ≈0.12. Under the same pressure and time conditions, the predicted ink width D =0.12×2.0×0.2 = 0.048 mm. The actual measured value is about 0.045 mm to 0.05 mm, verifying the characteristics of weak ink absorption and small ink diffusion range of cardboard. That is, this model can be used for paper property identification in the pre-detection stage and can also be used as the core prediction model for model predictive control.

[0074] To verify the effectiveness of the calligraphy robot writing control method based on paper adaptability provided by the present invention, in this embodiment, raw rice paper (with strong ink absorption) and cardboard paper (with weak ink absorption) are used as objects, and the writing experiment of the character "Hua" is carried out respectively. As Figure 5 shown, the control process of two steps and the application details of each parameter are fully presented. The experimental conditions are as follows: The robotic arm uses the DOBOT CR5 type, the force perception module uses the SRI six-axis force / torque sensor, and the visual perception module uses the JHSM300f type industrial camera.

[0075] In the step of constructing the paper ink absorption model, first, a stepped pressure pressing operation is performed on the blank area of the raw rice paper. The pressure gradient is set to 0.5 N, 1 N, 1.5 N, 2 N, 2.5 N, 3 N, 3.5 N, 4 N, 4.5 N, 5 N. The pressing duration of each group is fixed at 0.2 s. After the pressing is completed, the pen is lifted. During this process, the force perception module records the pressure value of each pressing in real time, the visual perception module synchronously captures the whole process of ink diffusion and the final stable ink width, and the main control unit synchronously records the pressing duration. A total of 10 groups of effective data are collected. Based on the collected multi-group data, the least squares method is used for fitting, and the ink absorption coefficient is solved according to the aforementioned calculation expression of the ink absorption coefficient. After calculation, the ink absorption coefficient K =0.733. This value is within the preset range of 0.6 to 0.9 of the ink absorption coefficient of raw rice paper, meeting the expected requirements. Finally, the ink absorption model of raw rice paper D= 0.733·F·T is established. The same stepped pressure pressing and data collection process is used to conduct experiments on cardboard. After fitting and solving by the least squares method, the ink absorption coefficient K =0.12, which is within the range of 0.05 to 0.2 of the ink absorption coefficient of cardboard, and the corresponding ink absorption model of cardboard is established D=0.12· F · T 。

[0076] After entering the writing control step, for writing the character "Hua" on the raw rice paper, the target stroke width is set to 3 mm, and the ink diffusion boundary threshold (i.e., the preset width boundary) is set to 3.2 mm (if it exceeds this threshold, it is determined as ink bleeding). The core control parameters are set as follows: the proportional coefficient of the PID algorithm is 1.0, the integral coefficient is 0.2, and the differential coefficient is 0.1; the prediction horizon of the model predictive control is set to 5 steps, the control horizon is 2 steps, and the sampling time is 0.05 s. The force-sensing pressure stabilization control is based on the preset target pressure of 2.0 N (calculated by inverse calculation based on the raw rice paper ink absorption model D =0.733· F · T , combined with the target stroke width of 3 mm). Relying on the real-time feedback signal of the force-sensing perception module, the Z-axis displacement of the robotic arm is dynamically adjusted by means of the PID algorithm. When it is detected that the actual pressure deviates from the target pressure, the adjustment amount is output through the collaborative calculation of the proportional term, integral term, and differential term, and the nib pressure is quickly corrected to the target value. The visual calibration control is started synchronously. The industrial camera collects the stroke images during the writing process in real time, and sequentially performs grayscale conversion, binarization (OTSU algorithm), contour detection, and minimum bounding circle fitting processing. The actual stroke width is obtained through the diameter of the minimum bounding circle and compared with the preset target width for comparison. When it is detected that the actual stroke width is 2.8 mm (less than ), the nib pressure is finely adjusted and increased to 2.1 N through the PID algorithm; when the actual stroke width is 3.1 mm (close to ), the pressure is finely adjusted and decreased to 1.9 N; when the actual stroke width is 3.3 mm (exceeding ), the pen is immediately lifted by 0.1 mm and paused for 0.05 s and then lowered to correct. The model predictive control is based on the established raw rice paper ink absorption model, and predicts the ink diffusion width in the next 0.05 to 0.1 s under the current pressure. When the real-time pressure is 2.0 N and the writing time is 0.2 s at a certain moment, it is predicted that the ink diffusion width in the next 0.05 s (i.e., the writing time of 0.25 s) is 3.665 mm, far exceeding the 3.2 mm threshold. The model predictive control algorithm immediately outputs a control command, and the pressure is reduced to 1.8 N through the PID algorithm, so that the predicted ink width is adjusted to 3.2985 mm, close to but not exceeding, avoiding the ink bleeding problem from the source.

[0077] For writing on cardboard paper, the core control logic is the same as that of raw rice paper, and only the relevant parameters are adjusted according to the cardboard ink absorption model and the ink absorption coefficient characteristics. Based on the cardboard ink absorption model D =0.12·F · T For example, when the target stroke width is 3 mm, the preset target pressure is calculated to be 12.5 N. The parameters of the PID algorithm remain unchanged, and the parameters of the model predictive control also remain unchanged. The ink diffusion boundary threshold is set to 3.1 mm (since the paper absorbs ink weakly, the ink diffusion range is smaller). During the writing process, the force-sense voltage stabilization control, visual calibration control, and model predictive control work together. Due to the small ink absorption coefficient of the paper, the ink diffusion speed is slow and the range is small. The adjustment frequency of visual calibration is lower than that on raw rice paper, and the fluctuation of the ink diffusion width predicted by the model predictive control is smaller. Finally, the stroke width of the character "华" on the paper surface is stabilized at 2.9 to 3.1 mm, without problems such as ink bleeding and stroke breakage.

[0078] To further verify the adaptability and robustness of the present invention under different control parameters, another set of experimental parameters is used for verification in this embodiment. The experimental conditions are the same as above, and the pressing duration for each pressure level in the stepped pressure pressing is set to 1.5 seconds. After collecting 10 groups of data and fitting them using the least squares method, the ink absorption coefficient of the raw rice paper K = 0.75, and the ink absorption model is D = 0.75· F ·1.5 = 1.125 F ; the ink absorption coefficient of the paper K = 0.12, and the ink absorption model is D = 0.12· F ·1.5 = 0.18 F . To verify the accuracy of the model, a pressure of 2 N is applied on the raw rice paper. According to the model, the predicted ink width is 2.25 mm, and the actually measured width is 2.3 mm, with an error less than 3%; a pressure of 2 N is applied on the paper. The predicted ink width is 0.36 mm, and the actually measured width is 0.35 mm, with an error less than 3%, indicating that the model has a high prediction accuracy.

[0079] In the writing control step, it is set to write the regular script character "华". The preset target width for raw rice paper is 2 mm and for paper is 0.8 mm. For writing on raw rice paper, the initial target pressure of the force-sense voltage stabilization control is set to 1.8 N, and the PID algorithm maintains the pressure stability; the visual calibration control detects the stroke width in real time. When the actually detected stroke width is 1.8 mm (too thin), the pressure is adjusted to 1.9 N, and the actually measured stroke width is stabilized at 2.0 mm; when the model predictive control reaches the stroke turning point, it predicts that after maintaining a pressure of 1.9 N for 0.1 second, the ink width will reach 2.3 mm (greater than ), reduce the pressure to 1.7N in advance to effectively avoid bleeding. For writing on cardboard, the initial target pressure of the force-sensing stable pressure control is set to 4.4N; when the visual calibration control detects that the actual stroke width is 0.7mm (too thin), adjust the pressure to 4.5N, and the actual stroke width stabilizes at 0.8mm; the model prediction control predicts that after maintaining the pressure of 4.5N for 0.1 second, its predicted ink width will reach 0.9mm (greater than ), reduce the pressure to 4.3N in advance to ensure that the stroke width complies with the regulations. The writing effect shows that the stroke widths of the character "Hua" on both types of paper accurately match the target requirements, without bleeding, and a natural ink diffusion effect is shown on the raw rice paper, and the stroke edges on the cardboard are clear, all with good calligraphy artistic expressiveness.

[0080] The experimental results show that whether it is raw rice paper or cardboard, through the two-step collaborative control of the control method provided by the present invention, combined with the least squares method to fit the ink absorption coefficient, PID pressure regulation, OTSU algorithm image processing and model prediction control, the adaptive writing of papers with different ink absorption characteristics can be achieved, ensuring uniform stroke width and no ink bleeding, verifying the feasibility and universality of the control method.

[0081] Embodiment 2

[0082] Please refer to Figure 6 , the embodiment of the present invention also provides a calligraphy robot writing control system based on paper adaptability, at least including:

[0083] A calligraphy robot writing control system for driving a writing brush to perform writing actions;

[0084] A force-sensing module for real-time collecting the pressure value when the pen tip touches the paper;

[0085] A visual sensing module for real-time collecting images of the ink on the paper;

[0086] A main control unit is respectively connected to the robotic arm, the force-sensing module and the visual sensing module, and the main control unit includes:

[0087] A paper ink absorption model construction module for controlling the robotic arm to drive the writing brush to perform stepped pressure pressing in the blank area of the paper, collecting and recording the pressure value, the pressing duration and the final stable ink width data for each pressing; based on the collected multiple groups of the pressure values, the pressing duration and the ink width data, fitting to obtain an ink absorption coefficient reflecting the paper ink absorption characteristics, and constructing a paper ink absorption model representing the relationship between pressure, time and ink width;

[0088] The force-sensing pressure control module is used to dynamically adjust the robotic arm to stabilize the pen tip pressure based on the preset target pressure and the actual pressure fed back by the force-sensing module.

[0089] The visual calibration control module is used to extract the actual stroke width from the stroke image acquired in real time by the visual perception module, compare it with the preset target width, dynamically correct the preset target pressure according to the comparison result, and use the corrected preset target pressure as the input of the force-sensing pressure stabilization control.

[0090] The model prediction control module is used to predict the ink spread width within a preset time period based on the paper ink absorption model and the current actual pressure. When the predicted ink spread width exceeds the preset width boundary, the pen tip pressure is proactively adjusted in advance to intervene before the ink smudges occur.

[0091] The paper-adaptive calligraphy robot writing control system provided in this embodiment, through the coordinated operation of the robotic arm, force perception module, vision perception module, and main control unit, can automatically execute paper ink absorption model construction, force-sensor voltage stabilization control, vision calibration control, and model predictive control. This system can automatically adapt to different types of paper with varying ink absorption characteristics, such as raw Xuan paper and cardstock, without human intervention. During the writing process, it achieves precise closed-loop control of stroke width through deep collaboration between vision and force perception, and utilizes model predictive control to proactively predict and intervene in ink diffusion trends, effectively avoiding ink smudging and significantly improving the adaptability and artistic expression of the calligraphy robot.

[0092] The specific methods for executing the operations of each module in Embodiment 2 have been described in detail in Embodiment 1, and will not be elaborated upon here. Those skilled in the art should understand that the paper ink absorption model construction module, force-sensory voltage regulation control module, visual calibration control module, and model prediction control module respectively execute the paper ink absorption model construction steps, force-sensory voltage regulation control, visual calibration control, and model prediction control in Embodiment 1. Their specific implementation methods, parameter configurations, and collaborative control logic are consistent with those described in Embodiment 1, and will not be repeated here.

[0093] In summary, the paper-adaptive calligraphy robot writing control method and system provided by this invention, compared with existing technologies, automatically identifies the ink absorption characteristics of paper and establishes an ink absorption model through the paper ink absorption model construction step, realizing adaptive adaptation to different papers and completely solving the industry pain point of "re-adjusting after changing paper"; through the force-sensory pressure stabilization control and visual calibration control forming a dual closed-loop structure, it realizes the coordinated correction of pressure stability and visual feedback, ensuring that the stroke width accurately matches the target requirements; through model predictive control, it makes forward-looking predictions of the ink diffusion trend and actively adjusts the pressure in advance, intervening before ink bleeding occurs, perfectly replicating the ink control effect of "precise strokes, without overflow or smudging" in calligraphy creation, and significantly improving the adaptability and artistic expression of calligraphy robot writing.

[0094] Furthermore, those skilled in the art should understand that although many problems exist in the prior art, each embodiment or technical solution of the present invention can be improved in only one or a few aspects, without necessarily solving all the technical problems listed in the prior art or background art simultaneously. Those skilled in the art should understand that any content not mentioned in a claim should not be construed as a limitation on that claim.

[0095] Although this paper uses terms such as robotic arm, force sensing module, vision sensing module, main control unit, paper ink absorption model, force-sensing voltage regulation control, vision calibration control, and model predictive control frequently, the possibility of using other terms is not excluded. These terms are used merely for the convenience of describing and explaining the essence of this invention; interpreting them as any additional limitation would contradict the spirit of this invention.

[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A paper-adaptive calligraphy robot writing control method, characterized in that, Includes the following steps: The paper ink absorption model construction steps involve controlling a robotic arm to drive a brush to perform stepped pressure pressing on a blank area of ​​the paper, collecting and recording the pressure value, pressing duration, and final stable ink width data for each press. Based on the collected data of multiple sets of pressure values, pressing duration and ink width, an ink absorption coefficient reflecting the ink absorption characteristics of paper is fitted, and a paper ink absorption model characterizing the relationship between pressure, time and ink width is constructed. The writing control step involves performing writing control based on the paper ink absorption model, and the writing control includes: Force-sensing pressure control dynamically adjusts the pressure at the tip of the brush based on the preset target pressure and the actual pressure fed back in real time by the force sensing module, in order to maintain stable pressure. Visual calibration control extracts the actual stroke width from the stroke image acquired in real time by the visual perception module, compares it with the preset target width, dynamically corrects the preset target pressure based on the comparison result, and uses the corrected preset target pressure as the input of the force-sensing pressure stabilization control. The model predictive control, based on the paper ink absorption model, predicts the ink spread width within a preset time period according to the current actual pressure. When the predicted ink spread width exceeds the preset width boundary, the pen tip pressure is proactively adjusted in advance to intervene before ink smudging occurs.

2. The paper-adaptive calligraphy robot writing control method according to claim 1, characterized in that, In the paper ink absorption model construction step, the pressure gradient of the stepped pressure pressing is 0.5N to 5N, and the pen is lifted after each press for a preset time. The pressure value, the duration of the pressure, and the width of the ink mark are fitted using a linear regression algorithm to obtain the ink absorption coefficient.

3. The paper-adaptive calligraphy robot writing control method according to claim 2, characterized in that: The linear regression algorithm uses the least squares method, and the expression for calculating the ink absorption coefficient is as follows: In the formula, For the first The pressure value of each press. For the first The duration of each press. For the first The stable ink width corresponding to each press. for The average value, for The average value.

4. The paper-adaptive calligraphy robot writing control method according to claim 1, characterized in that, In the visual calibration control, extracting the actual stroke width includes: sequentially performing grayscale conversion, binarization, contour detection, and minimum circumcircle fitting on the acquired stroke image, and obtaining the actual stroke width through the diameter of the minimum circumcircle.

5. The paper-adaptive calligraphy robot writing control method according to claim 1, characterized in that: In the model predictive control, a predictive model is established based on the paper ink absorption model. The optimal pressure adjustment amount that satisfies the ink diffusion constraint is solved by rolling optimization. The paper ink absorption model is corrected by the real-time feedback of the force perception module and the visual perception module.

6. The paper-adaptive calligraphy robot writing control method according to claim 5, characterized in that, In the model predictive control, the proactive adjustment of the pen tip pressure includes: within each control cycle, predicting the ink spread width in a preset time domain based on the paper ink absorption model, constructing a cost function that includes a weighted sum of tracking error and control energy, solving for the optimal pressure adjustment sequence that minimizes the cost function and satisfies the constraint that the ink spread width does not exceed a preset width boundary, executing only the first step of the pressure adjustment in the sequence as the output of the current cycle, and at the beginning of the next control cycle, using force and visual feedback information to correct the paper ink absorption model, and resolving the optimal pressure adjustment sequence with the corrected model.

7. The paper-adaptive calligraphy robot writing control method according to claim 1, characterized in that, The force-sensing pressure control employs a PID algorithm, and the target pressure of the PID algorithm is dynamically corrected by the visual calibration control based on the deviation between the actual stroke width and the preset target width.

8. The paper-adaptive calligraphy robot writing control method according to claim 1, characterized in that, In the visual calibration control, the step of dynamically correcting the preset target pressure based on the comparison result includes: If the actual stroke width is less than the preset target width, then increase the preset target pressure; If the actual stroke width is greater than the preset target width, then reduce the preset target pressure or perform a pen lifting operation.

9. The paper-adaptive calligraphy robot writing control method according to claim 1, characterized in that, The formula for the paper ink absorption model is as follows: In the formula, The width of the ink mark. The ink absorption coefficient is... For pen tip pressure, The duration of pressing or writing.

10. A paper-adaptive calligraphy robot writing control system, characterized in that, include: A robotic arm used to drive a calligraphy brush to perform writing actions; The force sensing module is used to collect the pressure value between the pen tip and the paper in real time; The visual perception module is used to acquire images of ink stains on paper in real time; The main control unit is connected to the robotic arm, the force sensing module, and the vision sensing module, respectively. The main control unit includes: The paper ink absorption model construction module is used to control the robotic arm to drive the brush to perform stepped pressure pressing on the blank area of ​​the paper, collect and record the pressure value, pressing duration and final stable ink width data of each pressing; based on the collected multiple sets of pressure values, pressing duration and ink width data, the ink absorption coefficient reflecting the ink absorption characteristics of the paper is fitted to construct a paper ink absorption model characterizing the relationship between pressure, time and ink width. The force-sensing pressure control module is used to dynamically adjust the robotic arm to stabilize the pen tip pressure based on the preset target pressure and the actual pressure fed back by the force-sensing module. The visual calibration control module is used to extract the actual stroke width from the stroke image acquired in real time by the visual perception module, compare it with the preset target width, dynamically correct the preset target pressure according to the comparison result, and use the corrected preset target pressure as the input of the force-sensing pressure stabilization control. The model prediction control module is used to predict the ink spread width within a preset time period based on the paper ink absorption model and the current actual pressure. When the predicted ink spread width exceeds the preset width boundary, the pen tip pressure is proactively adjusted in advance to intervene before the ink smudges occur.

Images