A method and apparatus for calibrating the load on a multi-axis flexible optical waveguide sensor.

Abstract

This application relates to a calibration method and apparatus for the load on a multi-axis flexible optical waveguide sensor. The method includes: acquiring the response signal of the flexible optical waveguide sensor; and determining the load on the flexible optical waveguide sensor based on the response signal and a load prediction model. This application obtains the load of the flexible optical waveguide sensor based on a load prediction model established using pre-acquired dense calibration data, which significantly improves the accuracy of the model output results compared to a load prediction model established based on sparse data.

Description

Technical Field

[0001] This application relates to the fields of flexible tactile sensor calibration and soft robot technology, and in particular to a calibration method and apparatus for the load on a multi-axis flexible optical waveguide sensor. Background Technology

[0002] Tactile sensors are commonly used to collect tactile information such as force, shape, and pressure, and have wide applications in robotics, wearable devices, virtual reality, smart prosthetics, and human-computer interaction. Flexible tactile sensors, made from soft materials with low Young's modulus, have unique advantages in tactile perception, naturally exhibiting potential in terms of compliance and shock absorption. Their large deformation significantly enhances the sensor's sensitivity, but also indirectly introduces nonlinear response issues, leading to inaccuracies when decoupling loads. Summary of the Invention

[0003] Therefore, it is necessary to provide a calibration method and apparatus for the load on a multi-axis flexible optical waveguide sensor that can improve the decoupling accuracy, in order to address the above-mentioned technical problems.

[0004] In a first aspect, this application provides a method for calibrating the load on a multi-axis flexible optical waveguide sensor, the method comprising:

[0005] Acquire the response signal of the flexible optical waveguide sensor;

[0006] The load of the optical waveguide flexible sensor is determined based on the response signal and the load prediction model. The load prediction model is a model with the response signal as input and the load as output, established based on the dense calibration data obtained in advance.

[0007] In one embodiment, the method further includes:

[0008] Acquire triaxial load data; the triaxial load data is manually generated dense data.

[0009] The three-dimensional response under triaxial load data is predicted based on the pre-established response prediction model, resulting in a dense dataset; the dense dataset includes the triaxial load data and the three-dimensional response data output by the response prediction model.

[0010] A load prediction model is obtained by training the model based on a dense dataset.

[0011] In one embodiment, the process of establishing the above-mentioned response prediction model includes:

[0012] Acquire calibration experimental data and divide the data into training and testing datasets; the calibration experimental data includes three-dimensional response and corresponding triaxial loads;

[0013] The training is conducted using a training dataset and different initial models, with triaxial load as input and three-dimensional response as output, to obtain a response prediction model; among which, different initial models include different model types and different hyperparameter models of the same type.

[0014] In one embodiment, the above-mentioned training based on a training dataset and different initial models, taking triaxial load as input and three-dimensional response as output, yields a response prediction model including:

[0015] Each initial model is trained separately based on the training dataset to obtain multiple candidate response prediction models;

[0016] Based on the test dataset, the performance of multiple candidate response prediction models is evaluated, and the response prediction model is determined from the multiple candidate response prediction models according to the evaluation results.

[0017] In one embodiment, the above-mentioned evaluation of the model performance of multiple candidate models based on a test dataset, and the determination of the response prediction model from multiple candidate response prediction models based on the evaluation results, includes:

[0018] The test dataset is input into the candidate response prediction model to obtain multiple test results output by the candidate response prediction model;

[0019] Determine the root mean square error of each test result, and select the candidate response prediction model with the smallest root mean square error of each test result as the response prediction model.

[0020] In one embodiment, the above-mentioned model training based on a dense dataset to obtain a load prediction model includes:

[0021] The candidate load prediction model is obtained by training the model using a dense dataset; the candidate load prediction model includes different model types and different hyperparameter models of the same type.

[0022] Input the test dataset from the calibration experiment into the candidate load prediction model to obtain the root mean square error of the load prediction results;

[0023] If the root mean square error of the load prediction result does not meet the preset accuracy conditions, the adjustable parameters of the candidate load prediction model and the response prediction model are adjusted, and the process returns to the step of predicting the three-dimensional response under the triaxial load data based on the response prediction model until the load prediction result output by the candidate load prediction model meets the preset accuracy conditions, thus obtaining the load prediction model.

[0024] Secondly, this application also provides a calibration device for the load on a multi-axis flexible optical waveguide sensor. The device includes:

[0025] The signal acquisition module is used to acquire the response signal of the optical waveguide flexible sensor;

[0026] The load determination module is used to determine the load of the optical waveguide flexible sensor based on the response signal and the load prediction model. The load prediction model is a model with the response signal as input and the load as output, established based on the dense calibration data obtained in advance.

[0027] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0028] Acquire the response signal of the flexible optical waveguide sensor;

[0029] The load of the optical waveguide flexible sensor is determined based on the response signal and the load prediction model. The load prediction model is a model with the response signal as input and the load as output, established based on the dense calibration data obtained in advance.

[0030] Fourthly, this application also provides a computer-readable storage medium. This computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0031] Acquire the response signal of the flexible optical waveguide sensor;

[0032] The load of the optical waveguide flexible sensor is determined based on the response signal and the load prediction model. The load prediction model is a model with the response signal as input and the load as output, established based on the dense calibration data obtained in advance.

[0033] Fifthly, this application also provides a computer program product. This computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0034] Acquire the response signal of the flexible optical waveguide sensor;

[0035] The load of the optical waveguide flexible sensor is determined based on the response signal and the load prediction model. The load prediction model is a model with the response signal as input and the load as output, established based on the dense calibration data obtained in advance.

[0036] The aforementioned calibration method and apparatus for the load on a multi-axis flexible optical waveguide sensor first acquires the response signal of the flexible optical waveguide sensor, and then determines the load on the flexible optical waveguide sensor based on the response signal and a load prediction model. In this embodiment, the load on the flexible optical waveguide sensor is obtained based on a load prediction model established using pre-acquired dense calibration data, which significantly improves the accuracy of the model output compared to a load prediction model established based on sparse data. Attached Figure Description

[0037] Figure 1 This is an application environment diagram of a calibration method for the load on a multi-axis flexible optical waveguide sensor in one embodiment;

[0038] Figure 2 This is a flowchart illustrating a method for calibrating the load on a multi-axis flexible optical waveguide sensor in one embodiment.

[0039] Figure 3 This is a flowchart illustrating the calibration method for the load on a multi-axis flexible optical waveguide sensor in another embodiment;

[0040] Figure 4 This is a flowchart illustrating the process of establishing a response prediction model in one embodiment;

[0041] Figure 5 This is a flowchart illustrating the process of obtaining a response prediction model in one embodiment;

[0042] Figure 6 This is a flowchart illustrating the process of determining a candidate response prediction model in one embodiment;

[0043] Figure 7 This is a flowchart illustrating the process of obtaining the load prediction model in one embodiment;

[0044] Figure 8 This is a flowchart illustrating the calibration method for the load on a multi-axis flexible optical waveguide sensor in another embodiment;

[0045] Figure 9 This is a structural block diagram of a calibration device for the load on a multi-axis flexible optical waveguide sensor in one embodiment;

[0046] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] First, before introducing the technical solutions of the embodiments of this application in detail, the technical background on which the embodiments of this application are based will be introduced.

[0049] Tactile perception plays a crucial role in enabling humans to acquire information from their environment and complete interactive tasks. Establishing tactile perception will also significantly improve the performance of robots in interactive tasks in unstructured environments and enhance their safety. Current methods for robots to acquire tactile information generally rely on tactile sensors.

[0050] Flexible sensors made from soft materials with low Young's modulus possess unique advantages in the field of tactile sensing, naturally exhibiting potential in terms of compliance and shock absorption. Their large deformation significantly enhances sensor sensitivity, but also indirectly introduces nonlinear response issues, making load decoupling difficult. The lack of corresponding calibration methods hinders the full potential of high-performance flexible tactile sensors in practical applications.

[0051] In current calibration methods for multi-axis flexible tactile sensors, few consider full-scale calibration across all load directions. Besides the experimental difficulties in applying load conditions, establishing a load prediction model for such sensors is also challenging. Without a high-precision load prediction model, multi-axis flexible tactile sensors cannot be readily applied in real-world scenarios.

[0052] Based on this, this application provides a method and apparatus for calibrating the load on a multi-axis flexible optical waveguide sensor, aiming to solve the above-mentioned technical problems.

[0053] The calibration method for the load on a multi-axis flexible optical waveguide sensor provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or located in the cloud or on other network servers. Terminal 102 acquires the response signal of the optical waveguide flexible sensor and determines the load of the optical waveguide flexible sensor based on the response signal and a load prediction model. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, and tablets. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0054] In one exemplary embodiment, such as Figure 2 As shown, this application provides a method for calibrating the load on a multi-axis flexible optical waveguide sensor, which can be applied to... Figure 1 Taking terminal 102 as an example, the explanation includes the following steps S201 to S202. Wherein:

[0055] S201, acquire the response signal of the optical waveguide flexible sensor.

[0056] The reason why the flexible optical waveguide sensor can respond to multi-axis vector forces is that the cylindrical optical waveguide core buried in the cladding has anisotropic response characteristics.

[0057] In this embodiment, when the optical waveguide flexible sensor is subjected to forces in different directions on the optical waveguide, the optical waveguide will undergo different deformations, establish different light propagation paths, and thus obtain different light intensity responses at the end of the inner core. These light intensity responses are the response signals.

[0058] S202, based on the response signal and load prediction model, determine the load of the optical waveguide flexible sensor.

[0059] Among them, the load prediction model is a model built based on the dense calibration data obtained in advance, with the response signal as input and the load as output.

[0060] In this embodiment, based on the response signal obtained in the above embodiments, the terminal can input the response signal into the load prediction model to determine the load of the flexible optical waveguide sensor. Alternatively, the response signal can be processed and then input into the load prediction model to determine the load of the flexible optical waveguide sensor. This embodiment does not specifically limit the method for determining the load of the flexible optical waveguide sensor.

[0061] The above-mentioned calibration method for the load on the multi-axis flexible optical waveguide sensor first acquires the response signal of the flexible optical waveguide sensor, and then determines the load of the flexible optical waveguide sensor based on the response signal and the load prediction model. In this embodiment, the load of the flexible optical waveguide sensor is obtained based on the load prediction model established by pre-acquired dense calibration data, which greatly improves the accuracy of the model output results compared with the load prediction model established by sparse data.

[0062] In one exemplary embodiment, based on the above embodiments, please refer to... Figure 3 The method in this application embodiment further includes the following S301 to S303. Wherein:

[0063] S301, acquire triaxial load data.

[0064] Among them, the triaxial load data is artificially generated dense data.

[0065] In this embodiment of the application, the terminal can acquire artificially generated dense data. The dense data can be triaxial load data with a spacing of 0.05N. For example, the normal force can be 0N, 0.05N, 0.1N...1N, and the tangential forces in the two directions can be 0N, 0.05N, 0.1N...1N, respectively. The above three components form a dense point set that presents a cube in force space.

[0066] S302, based on a pre-established response prediction model, predicts the three-dimensional response under triaxial load data to obtain a dense dataset.

[0067] The dense dataset includes triaxial load data and three-dimensional response data output by the response prediction model.

[0068] In this embodiment, based on the triaxial load data obtained in the above embodiments, the terminal can input the triaxial load data into a pre-established response prediction model to predict the three-dimensional response under the triaxial load data and obtain a dense dataset. For example, if the triaxial load data has a step size of 0.1N, the dense dataset obtained by predicting the three-dimensional response under the triaxial load data has a load spacing of 0.05N.

[0069] S303, a load prediction model is obtained by training the model based on a dense dataset.

[0070] In this embodiment, based on the dense dataset obtained in the above embodiments, the terminal can use an even denser dataset and a machine learning algorithm (such as the k-nearest neighbor algorithm) to build a sensor load prediction model.

[0071] In this embodiment, directly collecting overly dense datasets through experiments may place an unbearable fatigue burden on the flexible waveguide sensor, thus reducing its lifespan after calibration, and even causing structural damage before the sensor has completed the calibration experiment. However, this problem can be avoided by rationally utilizing the inference capabilities of the response prediction model to obtain denser calibration data based on relatively sparse experimental calibration data.

[0072] In one exemplary embodiment, based on the above embodiments, please refer to... Figure 4 The embodiments of this application relate to the process of establishing a response prediction model, including the following steps S401 to S402. Wherein:

[0073] S401, acquire calibration experimental data and divide the calibration experimental data into training dataset and test dataset.

[0074] The calibration experimental data includes the three-dimensional response and the corresponding triaxial load.

[0075] In this embodiment, a flexible sensor is fixedly installed on a calibration platform, and multiple loads within the range are applied at the designed position. For example, if the designed range of the three-axis components is between 0 and 1N, only a load application scheme with a step size of 0.1N is needed. Within a range slightly larger than the designed range of -0.2 to 1.2N (data slightly larger than the range is collected to improve the prediction accuracy of the load at the edge of the range), the load applied by the sensor and the corresponding response data are recorded.

[0076] S402 is trained based on the training dataset and different initial models, taking triaxial load as input and three-dimensional response as output to obtain a response prediction model.

[0077] Different initial models include different model types and different hyperparameter models of the same type, such as Gaussian process regression equations, neural network models, linear regression equations, regression trees, ensemble trees, and support vector machines.

[0078] In this embodiment of the application, by comparing different types of initial models, the best-performing model can be selected to improve the accuracy of the output results.

[0079] In one exemplary embodiment, based on the above embodiments, please refer to... Figure 5 This application embodiment relates to the process of training a response prediction model based on a training dataset and different initial models, taking a triaxial load as input and a three-dimensional response as output, and including the following S501 to S502. Wherein:

[0080] S501, train each initial model according to the training dataset to obtain multiple candidate response prediction models.

[0081] In this embodiment, the terminal can be trained based on the training dataset and different types of initial models to obtain multiple candidate response prediction models.

[0082] S502 evaluates the performance of multiple candidate models based on the test dataset and determines the response prediction model from among the multiple candidate response prediction models based on the evaluation results.

[0083] In this embodiment, based on the candidate response prediction models obtained in the above embodiments, the terminal can evaluate the model performance of these candidate response prediction models, such as output accuracy, recall, and root mean square error. Then, the initial model corresponding to the optimal evaluation result is selected from these evaluation results as the response prediction model.

[0084] In one exemplary embodiment, based on the above embodiments, please refer to... Figure 6This application embodiment relates to a process of evaluating the performance of multiple candidate models based on a test dataset, and determining the response prediction model from multiple candidate response prediction models based on the evaluation results, including the following steps S601 to S602. Wherein:

[0085] S601, input the test dataset into the candidate response prediction model to obtain multiple test results output by the candidate response prediction model.

[0086] In this embodiment of the application, based on the training dataset obtained in the above embodiments, the terminal can input the training data in the training dataset into the candidate response prediction model to obtain multiple training results output by the candidate response prediction model.

[0087] S602, determine the root mean square error of each test result, and select the candidate response prediction model with the smallest root mean square error of each test result as the response prediction model.

[0088] In this embodiment, based on the multiple training results obtained from the above embodiments, the terminal can determine the root mean square error of each training result and select the candidate response prediction model with the smallest root mean square error of each test result as the response prediction model.

[0089] In one exemplary embodiment, based on the above embodiments, please refer to... Figure 7 This application's embodiments relate to model training based on dense datasets to obtain a load prediction model, including the following steps S701 to S703. Wherein:

[0090] S701 uses a dense dataset for model training to obtain a candidate load prediction model.

[0091] In this embodiment of the application, based on the dense data set after the above thickening process, the terminal can train the initial load prediction model according to the dense data set to obtain the trained candidate load prediction model.

[0092] S702, input the test dataset from the calibration experiment data into the candidate load prediction model to obtain the root mean square error of the load prediction result.

[0093] In this embodiment of the application, the three-dimensional response of the test dataset in the calibration experimental data is input into the candidate load prediction model to obtain the load prediction result, which is compared with the triaxial load of the test dataset to obtain the root mean square error of each candidate load prediction model.

[0094] S703, if the root mean square error of the load prediction result does not meet the preset accuracy condition, adjust the adjustable parameters of the candidate load prediction model and the response prediction model, and return to the step of predicting the three-dimensional response under the triaxial load data according to the response prediction model until the load prediction result output by the candidate load prediction model meets the preset accuracy condition, and obtain the load prediction model.

[0095] In this embodiment of the application, the adjustable parameters of the candidate load prediction model and the response prediction model are adjusted by determining the load prediction results, thereby making the output results of the load prediction model more accurate.

[0096] In one exemplary embodiment, based on the above embodiments, please refer to... Figure 8 The method involved in this application embodiment further includes the following steps S801 to S807. Wherein:

[0097] S801: Obtain calibration experimental data and divide the calibration experimental data into training dataset and test dataset;

[0098] S802: Train each initial model based on the training dataset with triaxial load as input and three-dimensional response as output to obtain multiple candidate response prediction models. Input the triaxial load of the test dataset into the candidate response prediction models to obtain multiple three-dimensional response test results output by the candidate response prediction models. Determine the root mean square error of each test result based on the triaxial response of the test dataset, and select the candidate response prediction model with the smallest root mean square error of each test result as the response prediction model.

[0099] S803: Acquire triaxial load data; the triaxial load data is artificially generated dense data. Based on the pre-established response prediction model, the three-dimensional response under the triaxial load data is predicted to obtain a dense data set.

[0100] S804: Use dense datasets to train the model and obtain the candidate load prediction model;

[0101] S805: Input the three-dimensional response of the test dataset in the calibration experiment data into the candidate load prediction model to obtain the load prediction result, compare it with the triaxial load of the test dataset, and obtain the root mean square error of each candidate load prediction model;

[0102] S806: If the load prediction result does not meet the preset accuracy conditions, adjust the adjustable parameters of the candidate load prediction model and the response prediction model, and return to the step of thickening the initial calibration data according to the pre-established response prediction model until the accuracy of the load result output by the candidate load prediction model meets the preset conditions, and obtain the load prediction model.

[0103] S807: Acquire the response signal of the flexible optical waveguide sensor; determine the load of the flexible optical waveguide sensor based on the response signal and the load prediction model.

[0104] First, for the same sensor, different calibration experiment loading schemes can only be performed serially, and the denser the loading scheme, the longer it takes. However, model training can be performed in parallel, so this method will accelerate sensor calibration.

[0105] Secondly, directly collecting overly dense datasets through experiments can place an unbearable fatigue burden on flexible waveguide sensors, reducing their lifespan after calibration and even causing structural damage before the calibration experiment is completed. This problem can be avoided by utilizing the inference capabilities of response prediction models to obtain denser calibration data based on sparser experimental calibration data.

[0106] Furthermore, the materials used in the fabrication of flexible waveguide sensors, especially the core material polyurethane, possess a certain degree of viscoelasticity. This results in a viscous response and limited response bandwidth, requiring a certain amount of time to stabilize at each load during calibration to obtain a relatively stable response. However, excessively long calibration times can cause the sensor to exhibit different response levels at the beginning and end of the calibration experiment. Therefore, directly using long-term experimental sampling to obtain dense datasets is not conducive to reflecting the sensor's true operating state and may even reduce the accuracy of load prediction.

[0107] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0108] Based on the same inventive concept, this application also provides a calibration device for the load on a multi-axis flexible optical waveguide sensor used to implement the calibration method for the load on the multi-axis flexible optical waveguide sensor described above. The solution provided by this device is similar to the implementation described in the above method. Therefore, the specific limitations in one or more embodiments of the calibration device for the load on a multi-axis flexible optical waveguide sensor provided below can be found in the limitations of the calibration method for the load on a multi-axis flexible optical waveguide sensor described above, and will not be repeated here.

[0109] In one embodiment, such as Figure 9 As shown, a calibration device 900 for the load on a multi-axis flexible optical waveguide sensor is provided, comprising:

[0110] The signal acquisition module 901 is used to acquire the response signal of the optical waveguide flexible sensor;

[0111] The load determination module 902 is used to determine the load of the optical waveguide flexible sensor based on the response signal and the load prediction model; wherein, the load prediction model is a model established based on the dense calibration data obtained in advance, with the response signal as input and the load as output.

[0112] In one embodiment, the above-mentioned apparatus further includes:

[0113] The data acquisition module is used to acquire triaxial load data; the triaxial load data is artificially generated dense data.

[0114] The data calibration module is used to predict the three-dimensional response under triaxial load data based on a pre-established response prediction model, and obtain a dense dataset; the dense dataset includes triaxial load data and three-dimensional response data output by the response prediction model.

[0115] The load prediction model determination module is used to train the model based on a dense dataset to obtain the load prediction model.

[0116] In one embodiment, the above-mentioned apparatus further includes:

[0117] The experimental data acquisition module is used to acquire calibration experimental data and divide the calibration experimental data into training datasets and test datasets; the calibration experimental data includes three-dimensional response and corresponding triaxial loads;

[0118] The response prediction model determination module is used to train a response prediction model based on the training dataset and different initial models, taking triaxial load as input and three-dimensional response as output; among them, different initial models include different model types and different hyperparameter models of the same type.

[0119] In one embodiment, the response prediction model determination module includes:

[0120] The candidate model determination unit is used to train each initial model according to the training dataset to obtain multiple candidate response prediction models;

[0121] The response prediction model determination unit is used to evaluate the model performance of multiple candidate response prediction models based on a test dataset, and determine the response prediction model from the multiple candidate response prediction models based on the evaluation results.

[0122] In one embodiment, the response prediction model determination unit includes:

[0123] The result determination sub-unit is used to input the test dataset into the candidate response prediction model and obtain multiple test results output by the candidate response prediction model;

[0124] The response prediction model determination sub-unit is used to determine the root mean square error of each test result, and the candidate response prediction model with the smallest root mean square error of each test result is selected as the response prediction model.

[0125] In one embodiment, the load prediction model determination module includes:

[0126] The load prediction model determination unit is used to train the model using a dense dataset to obtain candidate load prediction models; the candidate load prediction models include different model types and different hyperparameter models of the same type.

[0127] The error determination unit is used to input the three-dimensional response of the test dataset in the calibration experiment data into the candidate load prediction model to obtain the load prediction result, and compare it with the corresponding triaxial load in the test dataset to obtain the root mean square error.

[0128] The prediction model determination unit is used to adjust the adjustable parameters of the candidate load prediction model and the response prediction model if the root mean square error of the load prediction result does not meet the preset accuracy conditions, and then return to execute the step of predicting the three-dimensional response under the triaxial load data according to the response prediction model until the load prediction result output by the candidate load prediction model meets the preset accuracy conditions, thus obtaining the load prediction model.

[0129] Each module in the calibration device for the load on the aforementioned multi-axis flexible optical waveguide sensor can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0130] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a calibration method for the load on a multi-axis flexible optical waveguide sensor. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0131] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0132] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0133] Acquire the response signal of the flexible optical waveguide sensor;

[0134] The load of the optical waveguide flexible sensor is determined based on the response signal and the load prediction model. The load prediction model is a model with the response signal as input and the load as output, established based on the dense calibration data obtained in advance.

[0135] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0136] Acquire triaxial load data; the triaxial load data is manually generated dense data.

[0137] The three-dimensional response under triaxial load data is predicted based on the pre-established response prediction model, resulting in a dense dataset; the dense dataset includes the triaxial load data and the three-dimensional response data output by the response prediction model.

[0138] A load prediction model is obtained by training the model based on a dense dataset.

[0139] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0140] Acquire calibration experimental data and divide the data into training and testing datasets; the calibration experimental data includes three-dimensional response and corresponding triaxial loads;

[0141] The training is conducted using a training dataset and different initial models, with triaxial load as input and three-dimensional response as output, to obtain a response prediction model; among which, different initial models include different model types and different hyperparameter models of the same type.

[0142] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0143] Each initial model is trained separately based on the training dataset to obtain multiple candidate response prediction models;

[0144] Based on the test dataset, the performance of multiple candidate response prediction models is evaluated, and the response prediction model is determined from the multiple candidate response prediction models according to the evaluation results.

[0145] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0146] The test dataset is input into the candidate response prediction model to obtain multiple test results output by the candidate response prediction model;

[0147] Determine the root mean square error of each test result, and select the candidate response prediction model with the smallest root mean square error of each test result as the response prediction model.

[0148] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0149] The candidate load prediction model is obtained by training the model using a dense dataset; the candidate load prediction model includes different model types and different hyperparameter models of the same type.

[0150] Input the test dataset from the calibration experiment into the candidate load prediction model to obtain the root mean square error of the load prediction results;

[0151] If the root mean square error of the load prediction result does not meet the preset accuracy conditions, the adjustable parameters of the candidate load prediction model and the response prediction model are adjusted, and the process returns to the step of predicting the three-dimensional response under the triaxial load data based on the response prediction model until the load prediction result output by the candidate load prediction model meets the preset accuracy conditions, thus obtaining the load prediction model.

[0152] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0153] Acquire the response signal of the flexible optical waveguide sensor;

[0154] The load of the optical waveguide flexible sensor is determined based on the response signal and the load prediction model. The load prediction model is a model with the response signal as input and the load as output, established based on the dense calibration data obtained in advance.

[0155] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0156] Acquire triaxial load data; the triaxial load data is manually generated dense data.

[0157] The three-dimensional response under triaxial load data is predicted based on the pre-established response prediction model, resulting in a dense dataset; the dense dataset includes the triaxial load data and the three-dimensional response data output by the response prediction model.

[0158] A load prediction model is obtained by training the model based on a dense dataset.

[0159] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0160] Acquire calibration experimental data and divide the data into training and testing datasets; the calibration experimental data includes three-dimensional response and corresponding triaxial loads;

[0161] The training is conducted using a training dataset and different initial models, with triaxial load as input and three-dimensional response as output, to obtain a response prediction model; among which, different initial models include different model types and different hyperparameter models of the same type.

[0162] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0163] Each initial model is trained separately based on the training dataset to obtain multiple candidate response prediction models;

[0164] Based on the test dataset, the performance of multiple candidate response prediction models is evaluated, and the response prediction model is determined from the multiple candidate response prediction models according to the evaluation results.

[0165] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0166] The test dataset is input into the candidate response prediction model to obtain multiple test results output by the candidate response prediction model;

[0167] Determine the root mean square error of each test result, and select the candidate response prediction model with the smallest root mean square error of each test result as the response prediction model.

[0168] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0169] The candidate load prediction model is obtained by training the model using a dense dataset; the candidate load prediction model includes different model types and different hyperparameter models of the same type.

[0170] Input the test dataset from the calibration experiment into the candidate load prediction model to obtain the root mean square error of the load prediction results;

[0171] If the root mean square error of the load prediction result does not meet the preset accuracy conditions, the adjustable parameters of the candidate load prediction model and the response prediction model are adjusted, and the process returns to the step of predicting the three-dimensional response under the triaxial load data based on the response prediction model until the load prediction result output by the candidate load prediction model meets the preset accuracy conditions, thus obtaining the load prediction model.

[0172] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0173] Acquire the response signal of the flexible optical waveguide sensor;

[0174] The load of the optical waveguide flexible sensor is determined based on the response signal and the load prediction model. The load prediction model is a model with the response signal as input and the load as output, established based on the dense calibration data obtained in advance.

[0175] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0176] Acquire triaxial load data; the triaxial load data is manually generated dense data.

[0177] The three-dimensional response under triaxial load data is predicted based on the pre-established response prediction model, resulting in a dense dataset; the dense dataset includes the triaxial load data and the three-dimensional response data output by the response prediction model.

[0178] A load prediction model is obtained by training the model based on a dense dataset.

[0179] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0180] Acquire calibration experimental data and divide the data into training and testing datasets; the calibration experimental data includes three-dimensional response and corresponding triaxial loads;

[0181] The training is conducted using a training dataset and different initial models, with triaxial load as input and three-dimensional response as output, to obtain a response prediction model; among which, different initial models include different model types and different hyperparameter models of the same type.

[0182] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0183] Each initial model is trained separately based on the training dataset to obtain multiple candidate response prediction models;

[0184] Based on the test dataset, the performance of multiple candidate response prediction models is evaluated, and the response prediction model is determined from the multiple candidate response prediction models according to the evaluation results.

[0185] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0186] The test dataset is input into the candidate response prediction model to obtain multiple test results output by the candidate response prediction model;

[0187] Determine the root mean square error of each test result, and select the candidate response prediction model with the smallest root mean square error of each test result as the response prediction model.

[0188] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0189] The candidate load prediction model is obtained by training the model using a dense dataset; the candidate load prediction model includes different model types and different hyperparameter models of the same type.

[0190] Input the test dataset from the calibration experiment into the candidate load prediction model to obtain the root mean square error of the load prediction results;

[0191] If the root mean square error of the load prediction result does not meet the preset accuracy conditions, the adjustable parameters of the candidate load prediction model and the response prediction model are adjusted, and the process returns to the step of predicting the three-dimensional response under the triaxial load data based on the response prediction model until the load prediction result output by the candidate load prediction model meets the preset accuracy conditions, thus obtaining the load prediction model.

[0192] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0193] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0194] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0195] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for calibrating the load on a multi-axis flexible optical waveguide sensor, characterized in that, The method includes: Acquire the response signal of the flexible optical waveguide sensor; The load of the optical waveguide flexible sensor is determined based on the response signal and the load prediction model; wherein the load prediction model is a model established based on pre-acquired dense calibration data, with the response signal as input and the load as output. Acquire triaxial load data; the triaxial load data is artificially generated dense data. The three-dimensional response under the triaxial load data is predicted based on a pre-established response prediction model to obtain a dense dataset; the dense dataset includes the triaxial load data and the three-dimensional response data output by the response prediction model. The load prediction model is obtained by training the model based on the dense dataset.

2. The method according to claim 1, characterized in that, The process of establishing the response prediction model includes: Acquire calibration experimental data and divide the calibration experimental data into training dataset and test dataset; the calibration experimental data includes three-dimensional response and corresponding triaxial load; The response prediction model is obtained by training the model using the training dataset and different initial models, taking the triaxial load as input and the three-dimensional response as output; wherein the different initial models include different model types and different hyperparameter models of the same type.

3. The method according to claim 2, characterized in that, The step of training the model based on the training dataset and different initial models, taking the triaxial load as input and the three-dimensional response as output, to obtain the response prediction model includes: Each of the initial models is trained according to the training dataset to obtain multiple candidate response prediction models; Based on the test dataset, the model performance of multiple candidate response prediction models is evaluated, and the response prediction model is determined from the multiple candidate response prediction models according to the evaluation results.

4. The method according to claim 3, characterized in that, The step of evaluating the performance of multiple candidate response prediction models based on the test dataset, and determining the response prediction model from the multiple candidate response prediction models based on the evaluation results, includes: The test dataset is input into the candidate response prediction model to obtain multiple test results output by the candidate response prediction model; The root mean square error of each test result is determined, and the candidate response prediction model with the smallest root mean square error of each test result is selected as the response prediction model.

5. The method according to claim 2, characterized in that, The process of training the model based on the dense dataset to obtain the load prediction model includes: The dense dataset is used to train the model to obtain a candidate load prediction model; the candidate load prediction model includes different model types and different hyperparameter models of the same type. The test dataset from the calibration experiment data is input into the candidate load prediction model to obtain the root mean square error of the load prediction result. If the root mean square error of the load prediction result does not meet the preset accuracy condition, the adjustable parameters of the candidate load prediction model and the response prediction model are adjusted, and the step of predicting the three-dimensional response under the triaxial load data based on the response prediction model is returned to be executed until the load prediction result output by the candidate load prediction model meets the preset accuracy condition, thus obtaining the load prediction model.

6. A calibration device for the load on a multi-axis flexible optical waveguide sensor, characterized in that, The device includes: The signal acquisition module is used to acquire the response signal of the optical waveguide flexible sensor; The load determination module is used to determine the load of the optical waveguide flexible sensor based on the response signal and the load prediction model; wherein the load prediction model is a model established based on pre-acquired dense calibration data, with the response signal as input and the load as output; The data acquisition module is used to acquire triaxial load data; the triaxial load data is artificially generated dense data. The data calibration module is used to predict the three-dimensional response under triaxial load data based on a pre-established response prediction model, and obtain a dense dataset; the dense dataset includes triaxial load data and three-dimensional response data output by the response prediction model. The load prediction model determination module is used to train the model based on a dense dataset to obtain the load prediction model.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images