Pose calculation method and device for six-degree-of-freedom parallel mechanism in underground scene

Abstract

The invention discloses a pose calculation method and device for a six-degree-of-freedom parallel mechanism in an underground scene. The method comprises the following steps: acquiring a training sample set; dividing each training sample into a corresponding geometric region based on training joint variables in the training sample set; each geometric region is trained, dictionary atoms are updated, meanwhile, the region weight of each geometric region is determined, and the dictionary atoms are specifically a column of elements of a dictionary and comprise basic features of a Jacobian matrix; determining a target area in each geometric area based on the area weight, searching in the target area to obtain a closest training joint variable closest to the current joint variable, and taking a training pose corresponding to the closest training joint variable as a coarse sample; determining a rough sample approximate Jacobian matrix corresponding to the rough sample based on the dictionary atoms; and carrying out iterative optimization on the rough sample approximate Jacobian matrix to obtain a current pose. The pose of the parallel mechanism can be accurately determined in real time in an underground scene.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of parallel mechanism, and particularly relates to a pose calculation method and device of a six-degree-of-freedom parallel mechanism in an underground scene. BACKGROUND

[0002] Parallel mechanism, also known as parallel robot (Parallel Mechanism), is a closed-loop mechanism in which a moving platform and a fixed platform are connected through at least two independent kinematic chains, the mechanism has two or more degrees of freedom, and is driven in a parallel manner.

[0003] In the prior art, the methods for solving the pose of the parallel mechanism include a Jacobian matrix solving method based on kinematic analysis, a numerical differential method, a solving method based on swarm intelligence algorithm, and a Jacobian matrix prediction model based on neural network. However, the above methods perform well in conventional scenes, but in underground scenes such as in-situ measurement of ground stress, precise monitoring of in-situ induced seismic signals by underground fault activity, and the like, the terrain is complex, and due to the existence of the non-homogeneous fault fracture zone and the vibration noise around the fault in the underground scene, the accuracy of the existing scheme is reduced. For example, in the Jacobian matrix solving method based on kinematic analysis, due to the non-homogeneous characteristics of the fault fracture zone, there is a micro-slip effect on the contact surface between the mechanism branch and the surrounding rock which is difficult to model. When the fault dip angle exceeds a certain angle, the displacement calculation error of this method increases to more than twice the theoretical value due to the failure to consider the rock-mechanism coupling dynamics, which seriously affects the inversion accuracy of the fault displacement. Moreover, in the fault creep stage, the mechanism is easy to enter a singular pose, and at this time, the Jacobian matrix solving method based on kinematic analysis directly fails.

[0004] Therefore, how to determine the pose of the parallel mechanism in real time and accurately in the underground scene is a technical problem to be solved by those skilled in the art. SUMMARY

[0005] The purpose of the present application is to solve the technical problem that the prior art cannot quickly and accurately determine the pose of the parallel mechanism in the underground application scene.

[0006] To achieve the above technical purpose, in one aspect, the present application provides a pose calculation method of a six-degree-of-freedom parallel mechanism in an underground scene, the method comprising: obtaining a training sample set based on a kinematic model of the six-degree-of-freedom parallel mechanism, wherein each training sample in the training sample set is specifically a corresponding training joint variable and a training pose; dividing each training sample into a corresponding geometric region based on the training joint variable in the training sample set, wherein the geometric region is specifically a plurality of working regions divided according to the activity range of the center point of the moving platform and the maximum angle that the moving platform can reach in space; Each of the geometric regions is trained and the dictionary atoms are updated. At the same time, the region weights of each of the geometric regions are determined. The dictionary atoms are specifically a column of elements in a dictionary, including the basic features of the Jacobian matrix. Based on the region weights, target regions are determined in each geometric region, and the closest training joint variable that is most similar to the current joint variable is obtained by searching in the target region. The training pose corresponding to the closest training joint variable is used as a coarse sample. Based on the dictionary atoms, the coarse sample approximate Jacobian matrix corresponding to the coarse sample is determined; The current pose is obtained by iteratively optimizing the coarse sample approximation Jacobian matrix.

[0007] Furthermore, the kinematic model Specifically represented as ,in, Let the position of the center point of the moving platform in the stationary coordinate system be given. The rotation angle of the moving platform relative to the stationary coordinate system is given. The kinematic model based on the six-degree-of-freedom parallel mechanism obtains training samples, specifically by generating training samples through inverse kinematic model and uniform grid sampling.

[0008] Furthermore, the division of the geometric region is specifically implemented through the following formula: , In the formula, For geometric regions, For the k-th geometric region, For the first i The center of each region To train joint variables.

[0009] Furthermore, the updating of dictionary atoms specifically includes: Randomly select a finite number of training samples in the geometric region to obtain multiple initial dictionary atoms; The final dictionary atoms are obtained by updating multiple initial dictionary atoms.

[0010] Furthermore, the initial dictionary atoms are updated using the following formula: ; In the formula, This is the recombined form of the Jacobian matrix corresponding to the training samples. The dictionary within the working region where the training samples are located. This is the sparse coding representation corresponding to the training samples. It is the Frobenius norm. For the first ia sparse representation vector of the signal, a maximum number of non-zero elements allowed in each sparse coding vector.

[0011] Further, the region weight is determined by the following formula: ; In the formula, is the region weight of the jth geometric region, i is the region weight of the jth geometric region, is the number of individuals in the jth region, i is the number of individuals in the jth region, is the number of individuals in the jth region, j is the number of individuals in the jth region, is the total number of divided regions.

[0012] Further, the iterative formula is specifically as follows: ; In the formula, is the data update direction and value, is the coarse sample approximate Jacobian matrix, is the residual between the rod length of the coarse sample and the target rod length, i.e., the rod length corresponding to the current joint variable, is the updated pose, is the damping coefficient, is the pose before updating.

[0013] On the other hand, the present application also provides a pose calculation device of a six-degree-of-freedom parallel mechanism in an underground scene, the device comprising: an acquisition module configured to acquire a training sample set based on a kinematics model of the six-degree-of-freedom parallel mechanism, wherein each training sample in the training sample set is specifically corresponding training joint variables and a training pose; a division module configured to divide each training sample into a corresponding geometric region based on the training joint variables in the training sample set, wherein the geometric region is specifically a plurality of working regions divided according to the activity range of the center point of the moving platform and the maximum angle that the moving platform can reach in space; a weight module configured to train each geometric region and update dictionary atoms, while determining the region weight of each geometric region, wherein the dictionary atom is specifically a column element of a dictionary, including the basic characteristics of the Jacobian matrix; a search module configured to determine a target region in each geometric region based on the region weight, and search in the target region to obtain the closest training joint variable most similar to the current joint variable, and take the training pose corresponding to the closest training joint variable as a coarse sample; determining a coarse sample approximate Jacobian matrix corresponding to the coarse sample based on the dictionary atom; updating the coarse sample approximate Jacobian matrix to obtain a current pose after iterative optimization.

[0014] The pose calculation method and device of the six-degree-of-freedom parallel mechanism in the underground scene provided by the application can determine the pose of the parallel mechanism in the underground scene in real time and accurately. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments described in the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0016] Figure 1 The flowchart of the pose calculation method of the six-degree-of-freedom parallel mechanism in the underground scene provided by the embodiments of the present application is shown. Figure 2 The structural diagram of the pose calculation device of the six-degree-of-freedom parallel mechanism in the underground scene provided by the embodiments of the present application is shown. DETAILED DESCRIPTION

[0017] In order for those skilled in the art to better understand the technical solutions in the specification, the technical solutions in the embodiments of the present application will be clearly and completely described in the specification below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0018] As Figure 1 shown is a flowchart of a pose calculation method of a six-degree-of-freedom parallel mechanism in an underground scene provided by an embodiment of the specification. Although the specification provides the method operation steps or device structures shown in the following embodiments or drawings, more or part of the operation steps or module units can be included in the method or device based on the conventional or without creative labor, and the execution order of the steps or the module structure of the device is not limited to the execution order or module structure shown in the embodiments or drawings of the specification. The method or module structure can be sequentially executed or executed in parallel (for example, parallel processor or multi-thread processing environment, even including distributed processing, server cluster implementation environment) according to the method or module structure shown in the embodiments or drawings when the method or module structure is applied in actual device, server or terminal product.

[0019] The pose calculation method of the six-degree-of-freedom parallel mechanism in the underground scene provided in the embodiments of the specification can be applied in terminal devices such as clients and servers, as Figure 1 shown, the method specifically includes the following steps: Step S101, obtaining a training sample set based on a kinematics model of the six-degree-of-freedom parallel mechanism, and the training sample in the training sample set is specifically a corresponding training joint variable and a training pose.

[0020] Specifically, the kinematics model specifically includes a position of a center point of a moving platform in a stationary coordinate system and a rotation angle of the moving platform relative to the stationary coordinate system, and the rotation angle of the moving platform relative to the stationary coordinate system is represented by Euler angles, and the kinematics model is specifically represented as , wherein, is the position of the center point of the moving platform in the stationary coordinate system, is the rotation angle of the moving platform relative to the stationary coordinate system.

[0021] In order to construct the above kinematics model, a stationary coordinate system needs to be established on a fixed base, i.e., a static platform , a moving coordinate system is established on a moving platform ; then six hinge points are determined on the static platform Determine the coordinates in the stationary coordinate system and identify six hinge points on the moving platform. The coordinates in the moving coordinate system; then a coordinate transformation is performed, moving coordinate system Relative to the stationary coordinate system The change can be achieved using a rotation matrix R and a position vector. The rotation matrix R is represented by Euler angles. Decision, for example, Then calculate the vector loop. For each connecting rod i, its vector in the static coordinate system can form a closed loop: ,in, The vector representing the i-th rod (direction from) point to ), This represents the position vector of the center of the moving platform. Indicates the coordinates of the hinge point on the moving platform. Transform to static coordinate system middle, Represent the coordinates of the hinge point on the static platform; finally, solve for the length of the rod, which is also the length of the connecting rod and the driving rod, and the length of the i-th rod. Vector Modulus length: The above is the core formula for inverse kinematics in constructing a kinematic model. For each given position X, the lengths of the six rods can be directly calculated using this formula.

[0022] The kinematic model based on the six-degree-of-freedom parallel mechanism obtains training samples specifically through inverse kinematics modeling and uniform grid sampling. However, it is necessary to determine the working range. Based on the machine's physical constraints (minimum / maximum link length, hinge angle limits, avoidance of inter-link collisions, etc.), the maximum range of motion that the machine's motion platform can achieve is calculated. Here, we use the limitation of the motion platform's center coordinate movement range and the range of three angle parameters. Sampling is performed within the working space, ensuring that each sampling point is within the limited constraints and physically accessible range. The inverse kinematics is used to calculate whether the corresponding link length is within the limits; the corresponding joint variables are calculated, and the lengths of the six driving links are calculated using the aforementioned inverse kinematics formula. Training samples and validation data are generated according to the corresponding positions. The training samples are specifically the corresponding training joint variables and training poses. Finally, the dataset is constructed.

[0023] Based on the above, the coordinates of three points (B1, B2, B3) on the moving platform and the coordinates of three points (b1, b2, b3) on the fixed platform are defined in the six-degree-of-freedom parallel mechanism. The initial value of the joint variable is obtained by subtracting the coordinates of the lower platform from the coordinates of the upper platform at the initial position. Using rotation matrix The pose rotation angle is calculated, where: , Let α, β, and γ represent rotations about the corresponding axes, and α, β, and γ represent the angles of rotation about the corresponding axes, respectively. Calculation of branch length (change in rod length): ,in The central position of the moving platform, and These are the coordinates of the hinge points on the upper and lower platforms, respectively; the corresponding Jacobian matrix solution method is as follows: Jacobian matrix row k: ,in , , , The derivative of the rotation matrix is ​​calculated as follows: , , .

[0024] Step S102: Based on the training joint variables in the training sample set, divide each training sample into a corresponding geometric region. Specifically, the geometric region is a number of working areas divided according to the range of motion of the center point of the moving platform and the maximum angle that the moving platform can reach in space.

[0025] Specifically, the geometric working area can be interpreted as the maximum angle and displacement range that the moving platform can achieve relative to the static coordinate system (base) through the extension and retraction of the six rods below, resulting in changes in the platform's roll, yaw, pitch angles, and displacement. After using the obtained samples (i.e., the training samples) and their corresponding joint variables, the samples need to be redistributed according to a fixed number of workspaces (i.e., the number of geometric regions). The center point of each workspace is selected as the origin of that region. Starting from this point, the difference between each sample point and the center point is calculated, and the joint variables closest to this center point are used as sample points for that region.

[0026] In this embodiment of the application, the division of the geometric region is specifically implemented using the following formula: , In the formula, For geometric regions, For the k-th geometric region, For the first i The center of each region To train the joint variables, the obtained geometric region segmentation information is stored, and the region adjacency relationship is calculated.

[0027] Step S103: Train and update the dictionary atoms for each of the geometric regions, and determine the region weights for each of the geometric regions. The dictionary atoms are specifically a column of elements in the dictionary, including the basic features of the Jacobian matrix.

[0028] Specifically, dictionary atoms are the basic elements that make up a dictionary. A dictionary is composed of multiple dictionary atoms, and each dictionary atom is a column of elements in the dictionary. This column of elements contains the basic characteristics of the Jacobian matrix.

[0029] First, the dictionary is divided into segments, which are the geometric regions. , The dictionary is updated in segments. To address abrupt changes at region boundaries, neighboring region data is introduced to initialize atoms. Let be the dictionary for the k-th region, and SVD be the singular value decomposition. for The corresponding reconstructed Jacobian matrix has a dimension of 36*1. Let be the joint variable within the k-th geometric region.

[0030] In this embodiment of the application, updating the dictionary atom specifically includes: Randomly select a finite number of training samples in the geometric region to obtain multiple initial dictionary atoms; The final dictionary atoms are obtained by updating multiple initial dictionary atoms.

[0031] Specifically, within the current region, select the Jacobian matrix corresponding to the joint variables, and within the neighboring regions, select the Jacobian matrix corresponding to the joint variables. From these, randomly select a finite number of elements as initial dictionary atoms. For each region's initial dictionary atoms, use the following formula: renew, This is the recombined form of the Jacobian matrix corresponding to the training samples. Specifically, the recombined form is a 36*1 form that has been vertically concatenated. The dictionary within the working region where the training samples are located. This is the sparse coding representation corresponding to the training samples. It is the Frobenius norm. For the first i A sparse representation vector of a signal. The maximum number of non-zero elements allowed in each sparse coding vector. Here, the dictionary atom is a dictionary atom formed by each of the finite number of columns in D. This is achieved through alternating optimization of the sparse coding. Atomic dictionary update ,in The error matrix is ​​obtained by using... The calculation is performed, i represents the selected non-empty sparse coding representation, k represents the kth dictionary atom; the alternating optimization refers to fixing the dictionary D or the sparse coding representation s, selecting the minimum deviation from J as the latest sparse coding representation or the dictionary D by approximately fitting J (Jacobi matrix), and updating the other sparse coding representation s or the dictionary D after updating one of the dictionary D or the sparse coding representation s, so that the deviation will become smaller and smaller, and the purpose of fitting J is achieved; the error matrix is the residual reconstruction error after removing the contribution of the kth atom in the reconstruction, which represents the error when reconstructing the sample using only other atoms, and the first column of the left singular vector U provides the direction of the updated atom after SVD decomposition of E, and the product of the first column of the right singular vector V and the singular value ∑ provides the updated sparse coefficient, so that the quality of the dictionary can be gradually improved in each iteration, so that the input data can be better represented.

[0032] In step S104, a target region is determined in each geometric region based on the region weight, and a search is performed in the target region to obtain a nearest training joint variable most similar to the current joint variable, and a training pose corresponding to the nearest training joint variable is taken as a coarse sample.

[0033] The nearest sample, i.e., the nearest training joint variable, is obtained by searching a target region according to region weight screening, and then performing error analysis on the current joint variable and each sample in the target region to obtain a sample with the minimum error as the nearest sample. At the same time, to avoid the existence of closedness between region centers, screening of the surrounding adjacent regions is also performed when screening the regions, and the adjacent regions are also searched, realizing the continuity between regions and helping to improve the continuity of the boundaries of each working area.

[0034] The distribution and average fitness of the current population in each region are calculated using the clustering partition information, and then the weight is dynamically allocated.

[0035] The region weight is determined by the following formula: ; In the formula, is the region weight of the mth geometric region, i is the number of individuals in the mth region, is the number of individuals in the mth region, i is the number of individuals in the mth region, is the number of individuals in the mth region, j is the number of individuals in the mth region, and is the total number of divided regions.

[0036] Specifically, when the hybrid search strategy is executed, global search is performed with a certain probability, and region search is performed with a certain probability, wherein is the center of the target region, ; wherein the geometric segmentation is used to train and update the dictionary atoms and sparse representation, and precise training is achieved for each region, especially in the working region with singular configuration points, which is more precise than the traditional analytical method.

[0037] In step S105, a coarse sample approximate Jacobian matrix corresponding to the coarse sample is determined based on the dictionary atoms.

[0038] In the embodiments of the present application, the coarse sample generated by the region guidance is used as fine adjustment data input, the clustering is used to query the region to which the coarse sample belongs, and the dictionary atoms corresponding to the part are obtained, and the sparse representation of the dictionary learning is used to The coarse sample approximate Jacobian matrix corresponding to the coarse sample is quickly obtained.

[0039] In step S106, a current pose is obtained after the coarse sample approximate Jacobian matrix is iteratively optimized.

[0040] Specifically, the fine sample is obtained based on the coarse sample approximate Jacobian matrix, and specifically, the fine sample is obtained by iteratively adjusting the coarse sample based on the target pose and the coarse sample.

[0041] The iteration formula is specifically as follows: ; In the formula, is the data update direction and numerical value, is the coarse sample approximate Jacobian matrix, is the residual error between the rod length of the coarse sample and the target rod length, i.e., the rod length corresponding to the current joint variable, is the updated pose, is the damping coefficient, is the pose before updating.

[0042] Specifically, after the Jacobian matrix corresponding to the coarse sample is obtained, the Newton method is used to directly iterate from the coarse sample position to realize fine adjustment of the coarse sample, reduce the difference from the target pose, and thus obtain the fine sample, i.e., the current position, so as to realize the forward kinematics solution of the parallel mechanism. In this way, it is not necessary to calculate step by step from the initial guess zero position, the convergence time is shortened, and the convergence speed is improved The method provided by the application segments the dictionary according to the geometric posture of the parallel mechanism, divides the working range of the mechanism into multiple classes through clustering, and performs segmented dictionary atom updating, which is different from the conventional global dictionary atom updating method, adopts an improved dictionary updating method, considers the correlation of neighboring areas, and improves the boundary continuity; in terms of intelligent optimization algorithms, the DBO optimization algorithm is improved, the global search range and the local search range are dynamically adjusted according to probabilities and weights, and intelligent optimization initial values are realized; the Jacobian matrix is estimated by using dictionary learning approximation, and fine adjustment is performed in Newton iteration, and finally the change amount of the posture of the mechanism corresponding to the current joint variable is output.

[0043] In traditional six-bar parallel mechanisms, the calculation of the forward position solution usually adopts a single solution method such as an analytical method or a Newton iteration method, consumes a large amount of time for calculation, and needs to derive complex kinematic equations; in recent years, due to the rise of deep learning technology, a large number of models for the forward position solution of parallel mechanisms have emerged, the posture is predicted through the model, so that the posture of the parallel mechanism is quickly solved, but the method is not ideal in terms of accuracy, and the prediction performance at the singular posture point is not good; a pure intelligent optimization algorithm (such as PSO) is sensitive to initial values, has strong global search ability but insufficient local precision, and needs a large number of iterations; the present application aims to use a geometric segmented dictionary learning method to quickly approximate the real sample Jacobian matrix, adopts a hybrid optimization method, and is divided into an offline training stage and an online test stage; at the same time, for the test sample, an improved intelligent optimization algorithm, DBO (DBO, hereinafter referred to as DBO), is used to realize the function of regional guidance, the global and local searches are divided according to probabilities, the global search and the local accurate search are performed, finally the dictionary atom of the corresponding area is queried, the Jacobian matrix is quickly approximated, the numerical differentiation is replaced by the estimated Jacobian matrix, the Newton iteration fine adjustment is performed, and the forward position solution is quickly calculated. The main purpose of the present application is to solve the three problems of poor real-time performance, insufficient stability of singular postures, and low global convergence probability in the solution of the Jacobian matrix of a six-bar parallel mechanism, the geometric segmented dictionary is used to quickly fit the approximate Jacobian matrix of the coarse sample, the segmented dictionary helps to solve the singular postures in different areas, and the addition of the neighboring areas ensures the boundary continuity.

[0044] Based on the above-mentioned pose calculation method of the six-degree-of-freedom parallel mechanism in the underground scene, one or more embodiments of the specification also provide a platform and a terminal for pose calculation of the six-degree-of-freedom parallel mechanism in the underground scene. The platform or terminal can include an apparatus, software, module, plug-in, server, client, etc. using the method described in the embodiments of the specification and combining the necessary implementation hardware. Based on the same innovative concept, the system in one or more embodiments provided by the embodiments of the specification is described in the following embodiments. Since the implementation scheme of the system solves the problem is similar to the method, the implementation of the specific system of the embodiments of the specification can be referred to the implementation of the foregoing method, and the repeated parts will not be described here. The terms "unit" or "module" used below can be a combination of software and / or hardware that implements a predetermined function. Although the system described in the following embodiments is preferably implemented in software, hardware, and a combination of software and hardware are also possible and are conceived.

[0045] Specifically, Figure 2 is a module structure schematic diagram of one embodiment of the pose calculation apparatus of the six-degree-of-freedom parallel mechanism in the underground scene provided by the specification, as Figure 2 shown, the pose calculation apparatus of the six-degree-of-freedom parallel mechanism in the underground scene provided in the specification includes: The acquisition module 201 is configured to acquire a training sample set based on a kinematics model of the six-degree-of-freedom parallel mechanism. The training sample in the training sample set is specifically a corresponding training joint variable and a training pose. The division module 202 is configured to divide each training sample into a corresponding geometric region based on the training joint variable in the training sample set. The geometric region is specifically a plurality of working regions divided according to the activity range of the center point of the moving platform and the maximum angle that the moving platform can reach in space. The weight module 203 is configured to train each geometric region and update a dictionary atom, while determining the region weight of each geometric region. The dictionary atom is specifically a column of elements of a dictionary, including the basic characteristics of the Jacobian matrix. The search module 204 is configured to determine a target region in each geometric region based on the region weight, and search in the target region to obtain the closest training joint variable closest to the current joint variable, and take the training pose corresponding to the closest training joint variable as a coarse sample. The determination module 205 is configured to determine a coarse sample approximate Jacobian matrix corresponding to the coarse sample based on the dictionary atom. The update module 206 is configured to obtain the current pose after iterative optimization of the coarse sample approximate Jacobian matrix.

[0046] It should be noted that the system described above can also include other embodiments according to the description of the corresponding method embodiments, and the specific implementation can refer to the description of the corresponding method embodiments described above, which will not be described here.

[0047] The embodiment of the present application also provides an electronic device, comprising: a processor; a memory for storing executable instructions of the processor; The processor is configured to execute the method provided by the above embodiment.

[0048] The electronic device provided by the embodiment of the present application stores the executable instructions of the processor through the memory, and when the processor executes the executable instructions, the following can be achieved: obtaining a training sample set based on a kinematics model of a six-degree-of-freedom parallel mechanism, wherein the training sample in the training sample set is specifically a corresponding training joint variable and a training pose; dividing each training sample into a corresponding geometric region based on the training joint variable in the training sample set, wherein the geometric region is specifically a plurality of working regions divided according to the activity range of the center point of the moving platform and the maximum angle that the moving platform can reach in space; training each geometric region and updating a dictionary atom, wherein the dictionary atom is specifically a column of elements of a dictionary, including the basic characteristics of the Jacobian matrix; determining a target region in each geometric region based on the region weight, and searching in the target region to obtain the closest training joint variable closest to the current joint variable, and taking the training pose corresponding to the closest training joint variable as a coarse sample; determining a coarse sample approximate Jacobian matrix corresponding to the coarse sample based on the dictionary atom; and obtaining the current pose after iterative optimization of the coarse sample approximate Jacobian matrix. The pose of the parallel mechanism can be determined in real time and accurately in an underground scene.

[0049] The above describes specific embodiments of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different than the order in the embodiments and still achieve the desired result. In addition, the processes depicted in the figures do not necessarily require the particular order shown or sequential order to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous or possible.

[0050] The method or device described in the above embodiments provided by the present specification can implement business logic through a computer program and be recorded on a storage medium, which can be read and executed by a computer to achieve the effects of the schemes described in the embodiments of the present specification, such as: Obtaining a training sample set based on a kinematics model of the six-degree-of-freedom parallel mechanism, and each training sample in the training sample set is specific to a corresponding training joint variable and a training pose; Dividing each training sample into a corresponding geometric region based on the training joint variable in the training sample set, and the geometric region is specific to a plurality of working regions divided according to the movement range of the center point of the moving platform and the maximum angle that the moving platform can reach in space; Training each geometric region and updating a dictionary atom, while determining a region weight of each geometric region, the dictionary atom is specific to a column of elements of a dictionary, including basic characteristics of a Jacobian matrix; Determining a target region in each geometric region based on the region weight, and searching in the target region to obtain a closest training joint variable closest to the current joint variable, and taking the training pose corresponding to the closest training joint variable as a coarse sample; Determining a coarse sample approximate Jacobian matrix corresponding to the coarse sample based on the dictionary atom; Obtaining the current pose after iterative optimization of the coarse sample approximate Jacobian matrix.

[0051] The storage medium can include a physical device for storing information, usually after digitizing the information and storing it in a medium using electrical, magnetic, or optical methods. The storage medium can include devices that store information using electrical energy, such as various types of memory, such as RAM, ROM, etc.; devices that store information using magnetic energy, such as hard disks, floppy disks, magnetic tapes, magnetic core memories, bubble memories, U disks; devices that store information using optical methods, such as CDs or DVDs. Of course, there are other ways of readable storage medium, such as quantum memory, graphene memory, etc.

[0052] The embodiments of the present specification are not limited to the cases that must comply with the industry communication standards, the standard computer resource data updating and data storage rules, or the cases described in one or more embodiments of the present specification. Certain industry standards or implementation schemes slightly modified based on the implementation described in the embodiments can also achieve the same, equivalent or similar, or modified implementation effects as described in the above embodiments. The embodiments obtained by applying these modifications or modifications to the data acquisition, storage, judgment, processing methods, etc. still belong to the scope of optional implementation schemes of the embodiments of the present specification.

[0053] The controller can be implemented in any suitable manner, for example, the controller can take the form of, for example, a microprocessor or processor and a computer readable medium storing computer readable program code, such as software or firmware, executable by the (micro)processor, logic gates, switches, application specific integrated circuits (asic), programmable logic controllers and embedded microcontrollers, examples of controllers include but are not limited to the following microcontrollers: arc 625d, atmel at91sam, microchip pic18f26k20 and silicone labs c8051f320, the memory controller can also be implemented as part of the control logic of the memory. It is also known to the person skilled in the art that, in addition to implementing the controller purely in the form of computer readable program code, it is also possible to implement the controller in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers and embedded microcontrollers, etc. to perform the same functions by means of logical programming of the method steps. Such a controller can therefore be regarded as a hardware component, and the means comprised within it for implementing the various functions can also be regarded as structures within the hardware component. Alternatively, the means for implementing the various functions can even be regarded as both a software module implementing the method and a structure within a hardware component.

[0054] The apparatus embodiments described above are only illustrative, for example, the division of the units is only a logical functional division, and in actual implementation, there can be another division manner, for example, a plurality of units or plug-ins can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection between the units through some interfaces, and can be electrical, mechanical or other forms.

[0055] These computer program instructions can also be loaded onto a computer or other programmable data updating device to cause a series of operational steps to be performed on the computer or other programmable device to generate a computer implemented process, so that the instructions executed on the computer or other programmable device provide a process for implementing the functions specified in the flow Figure 1 One flow or multiple flows and / or blocks Figure 1 One block or multiple blocks.

[0056] Each of the embodiments described in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the difference from other embodiments. In particular, for the system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the description of the method embodiments. In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the specification. In the specification, the illustrative description of the above terms is not necessarily the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples without contradiction.

[0057] Those skilled in the art will appreciate that the embodiments described herein are presented for the purpose of helping the reader to understand the principles of the present application, and should be understood as the protection scope of the present application is not limited to such specific descriptions and embodiments. Those skilled in the art can make various other specific modifications and combinations according to the technical inspiration disclosed in the present application without departing from the essence of the present application, and these modifications and combinations are still within the protection scope of the present application.

Claims

1. A pose calculation method of a six-degree-of-freedom parallel mechanism in an underground scene, characterized by, The method comprises: obtaining a training sample set based on a kinematics model of a six-degree-of-freedom parallel mechanism, wherein each training sample in the training sample set is specific to a corresponding training joint variable and a training pose; dividing each training sample into a corresponding geometric region based on the training joint variable in the training sample set, wherein the geometric region is specific to a plurality of working regions divided according to the range of movement of the center point of the moving platform and the maximum angle that the moving platform can reach in space; training each geometric region and updating dictionary atoms, wherein the dictionary atom is specific to a column of elements of a dictionary and includes the basic characteristics of a Jacobian matrix; determining a target region in each geometric region based on the region weight, searching for the closest training joint variable that is closest to the current joint variable in the target region, and taking the training pose corresponding to the closest training joint variable as a coarse sample; determining a coarse sample approximate Jacobian matrix corresponding to the coarse sample based on the dictionary atom; iteratively optimizing the coarse sample approximate Jacobian matrix to obtain the current pose.

2. The method of claim 1, wherein, The kinematic model Specifically represented as Wherein, is the position of the center point of the moving platform in the static coordinate system, is the rotation angle of the moving platform relative to the static coordinate system, and the kinematic model based on the six-degree-of-freedom parallel mechanism obtains a training sample, specifically by generating the training sample through an inverse kinematic model and uniform grid sampling.

3. The method of claim 1, wherein, The division of the geometric region is implemented by the following formula: ， wherein is a geometric region, is the kth geometric region, is the center of the i th region, is a training joint variable.

4. The method of claim 1, wherein, The updating of the dictionary atom specifically includes: randomly selecting the Jacobian matrices corresponding to a limited number of training samples in the geometric region to obtain a plurality of initial dictionary atoms; updating the plurality of initial dictionary atoms to obtain the final dictionary atom.

5. The method of claim 4, wherein, The initial dictionary atom is updated by the following formula: ； wherein, is a reorganized form of the Jacobian matrix corresponding to the training samples, is a dictionary within the working region where the training samples are located, is a sparse coding representation corresponding to the training samples, is the Frobenius norm, is the sparse representation vector of the i th signal, is the maximum number of non-zero elements allowed in each sparse coding vector.

6. The method of claim 1, wherein, The region weight is determined by the following formula: ； wherein is the region weight of the jth geometric region, i is the total number of regions, is the individual number in the jth region, i is the individual number in the jth region, is the individual number in the jth region, j is the individual number in the jth region, is the total number of regions.

7. The method of claim 1, wherein, The coarse sample approximate Jacobian matrix is iteratively optimized by the following formula: ； In the formula, is the data update direction and value, is the coarse sample approximate Jacobian matrix, is the residual error between the coarse sample length and the target length, i.e., the length corresponding to the current joint variable, is the updated pose, is the damping coefficient, is the pose before updating.

8. A device for calculating a pose of a six-degree-of-freedom parallel mechanism in an underground scene, characterized by The device comprises: an obtaining module configured to obtain a training sample set based on a kinematics model of a six-degree-of-freedom parallel mechanism, wherein each training sample in the training sample set is specific to a corresponding training joint variable and a training pose; a dividing module configured to divide each training sample into a corresponding geometric region based on the training joint variable in the training sample set, wherein the geometric region is specific to a plurality of working regions divided according to the range of movement of the center point of the moving platform and the maximum angle that the moving platform can reach in space; a weight module configured to train each geometric region and update dictionary atoms, wherein the dictionary atom is specific to a column of elements of a dictionary and includes the basic characteristics of a Jacobian matrix; a searching module configured to determine a target region in each geometric region based on the region weight, search for the closest training joint variable that is closest to the current joint variable in the target region, and take the training pose corresponding to the closest training joint variable as a coarse sample; a determining module configured to determine a coarse sample approximate Jacobian matrix corresponding to the coarse sample based on the dictionary atom; an updating module configured to iteratively optimize the coarse sample approximate Jacobian matrix to obtain the current pose.

Images