CN112428263B - Manipulator control method, device and clustering model training method - Google Patents

Abstract

The application discloses a mechanical arm control method and device and a clustering model training method. The mechanical arm control method comprises the steps of obtaining a force group of a mechanical arm; performing parameter regression on the force group to obtain a parameter regression value of the force group; determining KL divergence between the force group and a clustering model according to the parameter regression value and the clustering model, wherein the clustering model is a Gaussian mixture model trained through the force group of the mechanical arm; and automatically switching the virtual constraint according to the KL divergence. The method and the device solve the problems of poor flexibility and low efficiency of the statically generated virtual constraints.

Description

Translated fromChinese

机械臂控制方法、装置及聚类模型训练方法Manipulator control method, device and clustering model training method

技术领域technical field

本申请涉及机械臂领域，具体而言，涉及一种机械臂控制方法、装置及聚类模型训练方法。The present application relates to the field of robotic arms, and in particular, to a robotic arm control method, device, and clustering model training method.

背景技术Background technique

骨科手术控制中出现的钻、铣等任务控制是基于机械臂的直线约束来完成的，因此机械臂的直线约束具有较强的实际意义。在人手拖动机械臂作业过程中，由于人手存在定位误差，很难使机械臂沿直线移动进行工作。The task control of drilling and milling in orthopedic surgery control is based on the linear constraints of the manipulator, so the linear constraints of the manipulator have strong practical significance. In the process of manipulating the manipulator, it is difficult to make the manipulator move in a straight line due to the positioning error of the manipulator.

机械臂的直线约束解决方案是引入虚拟约束对机械臂运动进行约束控制。常用的方式是引入静态的虚拟约束，但是静态生成的虚拟约束缺乏灵活性与人交互性较差，操作者需要通过上位机配置才能对虚拟约束进行切换，效率低。The linear constraint solution of the manipulator is to introduce virtual constraints to control the motion of the manipulator. The commonly used method is to introduce static virtual constraints, but the statically generated virtual constraints lack flexibility and have poor human interaction. The operator needs to configure the upper computer to switch the virtual constraints, which is inefficient.

针对相关技术中静态生成的虚拟约束灵活性差、效率低的问题，目前尚未提出有效的解决方案。Aiming at the problems of poor flexibility and low efficiency of statically generated virtual constraints in the related art, no effective solution has been proposed yet.

发明内容SUMMARY OF THE INVENTION

本申请的主要目的在于提供一种机械臂控制方法，以解决静态生成的虚拟约束灵活性差、效率低的问题。The main purpose of this application is to provide a robotic arm control method to solve the problems of poor flexibility and low efficiency of statically generated virtual constraints.

为了实现上述目的，本申请提供了一种机械臂控制方法。In order to achieve the above purpose, the present application provides a method for controlling a robotic arm.

第一方面，本申请提供了一种机械臂控制方法In a first aspect, the present application provides a method for controlling a robotic arm

根据本申请的机械臂控制方法包括：The robotic arm control method according to the present application includes:

获取机械臂的力组；Get the force group of the robotic arm;

将所述力组进行参数回归，得到所述力组的参数回归值；Perform parameter regression on the force group to obtain the parameter regression value of the force group;

根据所述参数回归值和聚类模型，确定所述力组和所述聚类模型之间的KL散度，其中，所述聚类模型为通过机械臂的力组训练的高斯混合模型；Determine the KL divergence between the force group and the clustering model according to the parameter regression value and the clustering model, wherein the clustering model is a Gaussian mixture model trained by the force group of the robotic arm;

根据所述KL散度对虚拟约束进行自动切换。Virtual constraints are automatically switched according to the KL divergence.

进一步的，所述根据所述KL散度对虚拟约束进行自动切换包括：Further, the automatic switching of virtual constraints according to the KL divergence includes:

判断所述KL散度是否大于KL散度阈值；Determine whether the KL divergence is greater than the KL divergence threshold;

若所述KL散度大于KL散度阈值，则对机械臂产生各向异性虚拟约束；If the KL divergence is greater than the KL divergence threshold, an anisotropic virtual constraint is generated on the robotic arm;

若所述KL散度不大于KL散度阈值，则不对机械臂进行虚拟约束。If the KL divergence is not greater than the KL divergence threshold, no virtual constraint is performed on the robotic arm.

进一步的，在所述判断所述KL散度是否大于KL散度阈值之后，所述方法还包括：Further, after judging whether the KL divergence is greater than the KL divergence threshold, the method further includes:

若所述KL散度不大于KL散度阈值，则对机械臂产生各向同性虚拟约束。If the KL divergence is not greater than the KL divergence threshold, an isotropic virtual constraint is generated for the robotic arm.

进一步的，所述将所述力组进行参数回归，得到所述力组的参数回归值包括：Further, performing parameter regression on the force group to obtain the parameter regression value of the force group includes:

将所述力组在球坐标系下分解，得到所述力组的角度坐标；Decompose the force group in a spherical coordinate system to obtain the angular coordinates of the force group;

利用最大似然估计，对所述力组的角度坐标进行参数回归，得到力组的参数回归值。Using maximum likelihood estimation, parameter regression is performed on the angular coordinates of the force group to obtain the parameter regression value of the force group.

进一步的，所述若所述KL散度大于KL散度阈值，则对机械臂产生各向异性虚拟约束之后，所述方法还包括：Further, if the KL divergence is greater than the KL divergence threshold, after generating anisotropic virtual constraints on the robotic arm, the method further includes:

通过所述聚类模型对所述机械臂产生虚拟约束；generating virtual constraints on the robotic arm through the clustering model;

根据所述虚拟约束，确定机械臂的关节速度，以便以关节速度控制的方式控制机械臂沿直线运动。According to the virtual constraint, the joint speed of the manipulator is determined, so as to control the manipulator to move in a straight line in a joint speed control manner.

第二方面，本申请提供了一种聚类模型训练方法，用于得到第一方面中的机械臂控制方法中的聚类模型。In a second aspect, the present application provides a clustering model training method for obtaining the clustering model in the robotic arm control method in the first aspect.

根据本申请的聚类模型训练方法包括：The clustering model training method according to the present application includes:

建立高斯混合模型，其中高斯混合模型具有第一预设数量的分模型；establishing a Gaussian mixture model, wherein the Gaussian mixture model has a first preset number of sub-models;

获取不少于两组的机械臂在不同位姿下的训练力组；Obtain no less than two groups of training force groups of the robotic arms in different poses;

将所述训练力组作为所述高斯混合模型的参数并进行聚类，得到聚类模型。The training force group is used as a parameter of the Gaussian mixture model and clustering is performed to obtain a clustering model.

进一步的，所述将所述训练力组作为所述高斯混合模型的参数并进行聚类，得到聚类模型包括：Further, using the training force group as a parameter of the Gaussian mixture model and performing clustering to obtain a clustering model includes:

将所述训练力组转换为球坐标系下的角度坐标，得到训练力组角度坐标；Converting the training force group into angular coordinates under the spherical coordinate system to obtain the training force group angular coordinates;

将所述训练力组角度坐标作为高斯混合模型的回归变量进行回归计算，得到第一预设数量的分模型的聚类结果；Using the training force group angle coordinates as a regression variable of the Gaussian mixture model to perform regression calculation to obtain the clustering results of the first preset number of sub-models;

将所述聚类结果进行参数集合，得到聚类模型。The clustering results are parameterized to obtain a clustering model.

第三方面，本申请提供了一种机械臂控制装置。In a third aspect, the present application provides a robotic arm control device.

根据本申请的机械臂控制装置包括：The robotic arm control device according to the present application includes:

获取模块，用于获取机械臂的力组；Get the module, which is used to get the force group of the robotic arm;

参数回归模块，用于将所述力组进行参数回归，得到所述力组的参数回归值；A parameter regression module, for performing parameter regression on the force group to obtain a parameter regression value of the force group;

确定模块，用于根据所述参数回归值和聚类模型，确定所述力组和所述聚类模型之间的KL散度，其中，所述聚类模型为通过机械臂的力组训练的高斯混合模型；A determination module for determining the KL divergence between the force group and the clustering model according to the parameter regression value and the clustering model, wherein the clustering model is trained by the force group of the robotic arm Gaussian mixture model;

切换模块，用于根据所述KL散度对虚拟约束进行自动切换。A switching module, configured to automatically switch the virtual constraints according to the KL divergence.

进一步的，所述参数回归模块包括：Further, the parameter regression module includes:

分解单元，用于将所述力组在球坐标系下分解，得到所述力组的角度坐标；a decomposition unit for decomposing the force group in a spherical coordinate system to obtain the angular coordinates of the force group;

参数回归单元，用于利用最大似然估计，对所述力组的角度坐标进行参数回归，得到所述力组的参数回归值。The parameter regression unit is used for performing parameter regression on the angle coordinates of the force group by using maximum likelihood estimation to obtain the parameter regression value of the force group.

进一步的，切换模块包括：Further, the switching module includes:

判断单元，用于判断所述KL散度是否大于KL散度阈值；a judgment unit for judging whether the KL divergence is greater than the KL divergence threshold;

各向异性约束单元，用于若所述KL散度大于KL散度阈值，则对机械臂产生各向异性虚拟约束；an anisotropic constraint unit, used for generating anisotropic virtual constraints on the robotic arm if the KL divergence is greater than the KL divergence threshold;

各向同性约束单元，用于若所述KL散度不大于KL散度阈值，则对机械臂产生各向同性虚拟约束。The isotropic constraint unit is configured to generate an isotropic virtual constraint on the robotic arm if the KL divergence is not greater than the KL divergence threshold.

进一步的，切换模块还包括：Further, the switching module also includes:

控制单元，用于根据所述各向异性虚拟约束，确定机械臂的关节速度，以便以关节速度控制的方式控制机械臂运动。The control unit is configured to determine the joint speed of the robot arm according to the anisotropic virtual constraint, so as to control the movement of the robot arm in a joint speed control manner.

第四方面，本申请提供一种非暂态计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时实现第一方面提供的机械臂控制方法和/或第二方面提供的聚类模型训练方法的步骤。In a fourth aspect, the present application provides a non-transitory computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the robotic arm control method provided by the first aspect and/or provided by the second aspect The steps of the clustering model training method.

第五方面，本申请提供一种机器人，包括机械臂、传感器、存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述程序时实现第一方面提供的机械臂控制方法和/或第二方面提供的聚类模型训练方法的步骤。In a fifth aspect, the present application provides a robot, comprising a robotic arm, a sensor, a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor implements the first aspect when executing the program The steps of the provided manipulator control method and/or the cluster model training method provided by the second aspect.

在本申请实施例中，采用获取力组的方式，通过对力组进行参数回归，并确定参数回归值和聚类模型之间的KL散度，达到了通过KL散度对虚拟约束进行自动切换的目的，从而实现了对虚拟约束进行自动切换的技术效果，进而解决了静态生成的虚拟约束灵活性差、效率低的问题。In the embodiment of the present application, the method of acquiring the force group is adopted, and by performing parameter regression on the force group, and determining the KL divergence between the parameter regression value and the clustering model, the automatic switching of virtual constraints through KL divergence is achieved. Therefore, the technical effect of automatic switching of virtual constraints is realized, and the problems of poor flexibility and low efficiency of statically generated virtual constraints are solved.

附图说明Description of drawings

构成本申请的一部分的附图用来提供对本申请的进一步理解，使得本申请的其它特征、目的和优点变得更明显。本申请的示意性实施例附图及其说明用于解释本申请，并不构成对本申请的不当限定。在附图中：The accompanying drawings, which constitute a part of this application, are used to provide a further understanding of the application and make other features, objects and advantages of the application more apparent. The accompanying drawings and descriptions of the exemplary embodiments of the present application are used to explain the present application, and do not constitute an improper limitation of the present application. In the attached image:

图1是根据本申请实施例的机械臂控制方法的流程示意图；1 is a schematic flowchart of a method for controlling a robotic arm according to an embodiment of the present application;

图2是根据本申请实施例的聚类模型训练方法的流程示意图；2 is a schematic flowchart of a clustering model training method according to an embodiment of the present application;

图3是根据本申请实施例的机械臂控制装置的结构框图；3 is a structural block diagram of a robotic arm control device according to an embodiment of the present application;

图4是根据本申请另一实施例的机械臂控制方法的流程示意图；4 is a schematic flowchart of a method for controlling a robotic arm according to another embodiment of the present application;

图5是根据本申请实施例的机械臂控制方法中机械臂的虚拟约束切换的示意图。FIG. 5 is a schematic diagram of virtual constraint switching of a manipulator in a manipulator control method according to an embodiment of the present application.

具体实施方式Detailed ways

为了使本技术领域的人员更好地理解本申请方案，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分的实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都应当属于本申请保护的范围。In order to make those skilled in the art better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only The embodiments are part of the present application, but not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the scope of protection of the present application.

需要说明的是，本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本申请的实施例。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。It should be noted that the terms "first", "second", etc. in the description and claims of the present application and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific sequence or sequence. It is to be understood that the data so used are interchangeable under appropriate circumstances for the embodiments of the application described herein. Furthermore, the terms "comprising" and "having" and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those expressly listed Rather, those steps or units may include other steps or units not expressly listed or inherent to these processes, methods, products or devices.

需要说明的是，在不冲突的情况下，本申请中的实施例及实施例中的特征可以相互组合。下面将参考附图并结合实施例来详细说明本申请。It should be noted that the embodiments in the present application and the features of the embodiments may be combined with each other in the case of no conflict. The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

根据本申请实施例，提供了一种机械臂控制方法，如图1所示，该方法包括如下的步骤S11至步骤S14：According to an embodiment of the present application, a method for controlling a robotic arm is provided. As shown in FIG. 1 , the method includes the following steps S11 to S14:

S11：获取机械臂的力组。S11: Obtain the force group of the robotic arm.

力组可以通过安装在机械臂上的力传感器直接获取。具体的，力传感器为多维力传感器。在该实施例中，示例的，力传感器通过三维力传感器或六维力传感器获取。通过力传感器获取的力组包括X、Y、Z轴对应的三个力分量。The force group can be directly acquired by the force sensor mounted on the robotic arm. Specifically, the force sensor is a multi-dimensional force sensor. In this embodiment, for example, the force sensor is acquired by a three-dimensional force sensor or a six-dimensional force sensor. The force group obtained by the force sensor includes three force components corresponding to the X, Y, and Z axes.

具体的，采集机械臂在X、Y、Z轴其中一个或多个方向拖动的不同姿态的第二预设数量的力组，力传感器获取的力组如式(1)所示，可以表达为：Specifically, a second preset number of force groups with different postures dragged by the robotic arm in one or more directions of the X, Y, and Z axes are collected. The force group obtained by the force sensor is shown in formula (1), which can be expressed as for:

其中，f为力传感器获取的坐标系的X、Y、Z轴对应的三个力分量集合，f_x为力传感器获取的坐标系的X轴对应的力，f_y为力传感器获取的坐标系的Y轴对应的力，f_z为力传感器获取的坐标系的Z轴对应的力。Among them, f is the set of three force components corresponding to the X, Y, and Z axes of the coordinate system obtained by the force sensor, f_x is the force corresponding to the X axis of the coordinate system obtained by the force sensor, and f_y is the coordinate system obtained by the force sensor. The force corresponding to the Y axis of , f_z is the force corresponding to the Z axis of the coordinate system obtained by the force sensor.

具体的，第二预设数量可以为3-10。示例的，第二预设数量为5，即获取一个或多个方向拖动的不同姿态的5组力组。Specifically, the second preset number may be 3-10. Exemplarily, the second preset number is 5, that is, 5 groups of force groups of different gestures of dragging in one or more directions are obtained.

S12：将力组进行参数回归，得到力组的参数回归值。S12: Perform parameter regression on the force group to obtain the parameter regression value of the force group.

“将力组进行参数回归，得到力组的参数回归值”具体为：将力组在球坐标系下分解，得到力组的角度坐标；利用最大似然估计，对力组的角度坐标进行参数回归，得到力组的参数回归值。"Regressing the parameters of the force group to obtain the parameter regression value of the force group" is specifically: decompose the force group in the spherical coordinate system to obtain the angle coordinates of the force group; use the maximum likelihood estimation to parameterize the angle coordinates of the force group Regression to get the parameter regression value of the force group.

力组可以通过上述步骤S11中获取，并将获取的包括X、Y、Z轴对应的三个力分量的力组在球坐标系下分解，得到力组的角度坐标α。The force group can be obtained in the above step S11, and the obtained force group including the three force components corresponding to the X, Y, and Z axes is decomposed in the spherical coordinate system to obtain the angle coordinate α of the force group.

如式(2)所示，力组的角度坐标α可以表达为：As shown in formula (2), the angle coordinate α of the force group can be expressed as:

可以得到的多组力组对应的多个角度坐标{α₁,α₂，…，α_k}，其中，k为第二预设数量。对多个角度坐标进行参数回归，具体的，参数回归可以是线性回归，可以是梯度下降法，也可以是最大似然估计。在该实施例中，示例的，参数回归是最大似然估计。利用最大似然估计，对多个角度坐标{α₁,α₂，…，α_k}进行参数回归，得到多个角度坐标{α₁,α₂，…，α_k}对应的最大似然估计α_r，即力组的参数回归值为α_r。Multiple_angular coordinates {α₁ , α₂ , . Parameter regression is performed on multiple angle coordinates. Specifically, the parameter regression can be linear regression, gradient descent method, or maximum likelihood estimation. In this embodiment, parametric regression is, by way of example, maximum likelihood estimation. Using maximum likelihood estimation, perform parameter regression on multiple angular coordinates {α₁ , α₂ ,..., α_k } to obtain the maximum likelihood estimates corresponding to multiple angular coordinates {α₁ , α₂ ,..., α_k } α_r , that is, the parameter regression value of the force group is α_r .

S13：根据参数回归值和聚类模型，确定力组和聚类模型之间的KL散度，其中，聚类模型为通过机械臂的力组训练的高斯混合模型。S13: Determine the KL divergence between the force group and the cluster model according to the parameter regression value and the cluster model, wherein the cluster model is a Gaussian mixture model trained by the force group of the robotic arm.

参数回归值可以通过上述步骤S12中求解得到，聚类模型可以是机械臂的控制系统中预存的高斯混合模型，也可以是对力传感器获取的力组训练得到的高斯混合模型。在该实施例中，示例的，聚类模型为对力传感器获取的力组训练得到的高斯混合模型。其中，高斯混合模型是由不少于两个高斯分模型组成的模型，即，聚类模型具有m个高斯分模型(m≥2)。分别计算m个高斯分模型和参数回归值α_r之间的KL散度，确定m个KL散度。The parameter regression value can be obtained by solving the above step S12, and the clustering model can be a Gaussian mixture model pre-stored in the control system of the manipulator, or a Gaussian mixture model obtained by training the force group obtained by the force sensor. In this embodiment, for example, the clustering model is a Gaussian mixture model obtained by training the force group obtained by the force sensor. The Gaussian mixture model is a model composed of no less than two Gaussian sub-models, that is, the clustering model has m Gaussian sub-models (m≧2). Calculate the KL divergences between the m Gaussian partial models and the parameter regression value α_r respectively, and determine the m KL divergences.

S14：根据KL散度对虚拟约束进行切换。S14: Switch the virtual constraints according to the KL divergence.

“根据KL散度对虚拟约束进行切换”具体为：判断KL散度是否大于KL散度阈值；若KL散度大于KL散度阈值，则对机械臂产生各向异性虚拟约束；若KL散度不大于KL散度阈值，则不对机械臂进行虚拟约束或对机械臂产生各向同性虚拟约束。"Switching virtual constraints according to KL divergence" is specifically: judging whether the KL divergence is greater than the KL divergence threshold; if the KL divergence is greater than the KL divergence threshold, an anisotropic virtual constraint is generated for the robotic arm; If it is not greater than the KL divergence threshold, no virtual constraint is applied to the manipulator or an isotropic virtual constraint is generated for the manipulator.

如图2所示，为人机交互过程中机械臂的虚拟约束切换的示意图，其中，虚线圆圈部分为产生虚拟约束的地方，{C_h1}为各向同性虚拟约束，{C_h2}为各向异性虚拟约束，{S_h}为上述步骤S12中的交互力组参考坐标系。As shown in Figure 2, it is a schematic diagram of the virtual constraint switching of the robotic arm in the process of human-computer interaction, in which the dotted circle part is the place where the virtual constraint is generated, {C_h1 } is the isotropic virtual constraint, and {C_h2 } is the isotropic virtual constraint. The opposite sex virtual constraint, {S_h } is the reference coordinate system of the interaction force group in the above step S12.

KL散度阈值可以是机械臂的控制系统中预存的KL散度阈值，也可以是用户自己设定的KL散度阈值。在该实施例中，示例的，KL散度阈值为用户自己设定的KL散度阈值。The KL divergence threshold may be a pre-stored KL divergence threshold in the control system of the robotic arm, or a KL divergence threshold set by the user. In this embodiment, for example, the KL divergence threshold is a KL divergence threshold set by the user.

具体的，分别判断上述步骤S13中得到的m个KL散度是否大于KL散度阈值，当m个KL散度中存在大于KL散度阈值的KL散度时，则对机械臂产生各向异性虚拟约束(如KL散度阈值为-20，3个KL散度分别为-10、-50、-100，存在-10＞-20，则对机械臂产生各向异性虚拟约束)，其中，机械臂产生的各向异性虚拟约束为基于上述步骤S13中的聚类模型方向(其中，聚类模型方向为聚类模型对应的方向的单位矢量)产生的各向异性的虚拟约束；Specifically, it is determined whether the m KL divergences obtained in the above step S13 are greater than the KL divergence threshold, and when there is a KL divergence greater than the KL divergence threshold among the m KL divergences, anisotropy is generated for the robotic arm. Virtual constraints (for example, the KL divergence threshold is -20, the three KL divergences are -10, -50, and -100 respectively, and if -10>-20 exists, anisotropic virtual constraints are generated for the robotic arm), where the mechanical The anisotropic virtual constraint generated by the arm is an anisotropic virtual constraint generated based on the direction of the clustering model in the above step S13 (wherein the direction of the clustering model is the unit vector of the direction corresponding to the clustering model);

当m个KL散度中不存在大于KL散度阈值的KL散度时，则对机械臂产生各向同性虚拟约束或不对机械臂进行虚拟约束(如KL散度阈值为-20，3个KL散度分别为-30、-50、-100，不存在大于KL散度阈值的KL散度，则对机械臂产生各向同性虚拟约束或不对机械臂进行虚拟约束)；其中，机械臂产生的各向同性虚拟约束，该虚拟约束采用默认值，由用户自行设定。When there is no KL divergence greater than the KL divergence threshold in the m KL divergences, an isotropic virtual constraint is generated for the manipulator or no virtual constraint is performed on the manipulator (for example, the KL divergence threshold is -20, 3 KL divergences The divergences are -30, -50, -100, respectively, and there is no KL divergence greater than the KL divergence threshold, then an isotropic virtual constraint is generated for the robotic arm or no virtual constraint is performed on the robotic arm); Isotropic virtual constraint, the virtual constraint adopts the default value and is set by the user.

进一步的，在若KL散度大于KL散度阈值，则对机械臂产生各向异性虚拟约束之后，该方法还包括：Further, if the KL divergence is greater than the KL divergence threshold, after generating anisotropic virtual constraints on the manipulator, the method further includes:

根据各向异性虚拟约束，确定机械臂的关节速度，以便以关节速度控制的方式控制机械臂运动。According to the anisotropic virtual constraint, the joint speed of the manipulator is determined so as to control the motion of the manipulator in a joint speed control manner.

通过上述步骤S11中获取的力和获取的力对应的力矩，根据获取到的力和力矩得到虚拟约束力方向和虚拟约束力矩方向，如式(3)所示，可以表示为：Through the force obtained in the above step S11 and the torque corresponding to the obtained force, the virtual restraint force direction and the virtual restraint moment direction are obtained according to the obtained force and moment, as shown in formula (3), which can be expressed as:

其中，a_i为到虚拟约束力方向，f_i为通过机械臂上的力传感器获取的力，b_i为虚拟约束力矩方向，n_i为通过机械臂上的力传感器获取的力矩。Among them, a_i is the direction to the virtual restraint force, f_i is the force obtained by the force sensor on the manipulator, b_i is the direction of the virtual restraint moment, and_ni is the torque obtained by the force sensor on the manipulator arm.

可以通过机械臂上的力传感器获取力和力矩，得到虚拟力子空间C_f＝[f₁ f₂ …f_i]和虚拟力矩子空间C_n＝[n₁ n₂ … n_i]。则生成的由力和力矩合成的力旋量投影空间如式(4)所示，可以表示为：The force and moment can be acquired by the force sensor on the manipulator, and the virtual force subspace C_f =[f₁ f₂ ··· f_i ] and the virtual moment subspace C_n =[n₁ n₂ ··· n_i ] can be obtained. Then the generated force screw projection space synthesized by force and moment is shown in formula (4), which can be expressed as:

其中，P_w为力旋量投影空间对应的无量纲对角矩阵，C_f为虚拟力子空间，C_n为虚拟力矩子空间。在虚拟力子空间C_f和虚拟力矩子空间C_n已知的情况下，可以求得力旋量投影空间对应的无量纲对角矩阵P_w。Among them, P_w is the dimensionless diagonal matrix corresponding to the force screw projection space, C_f is the virtual force subspace, and C_n is the virtual moment subspace. When the virtual force subspace C_f and the virtual moment subspace C_n are known, the dimensionless diagonal matrix P_w corresponding to the force screw projection space can be obtained.

那么，对应的速度旋量投影空间为力旋量投影空间的补空间如式(5)所示，可以表示为：Then, the corresponding velocity screw projection space is the complementary space of the force screw projection space, as shown in formula (5), which can be expressed as:

其中，P_t为速度旋量投影空间对应的无量纲对角矩阵，I为单位矩阵。在虚拟力子空间C_f、虚拟力矩子空间C_n、单位矩阵I C_n已知的情况下，可以求得速度旋量投影空间对应的无量纲对角矩阵P_t。Among them, P_t is the dimensionless diagonal matrix corresponding to the velocity screw projection space, and I is the unit matrix. When the virtual force subspace C_f , the virtual moment subspace C_n , and the identity matrix IC_n are known, the dimensionless diagonal matrix P_t corresponding to the velocity screw projection space can be obtained.

那么，机械臂的关节速度的求解方式如式(6)所示，可以表示为：Then, the solution method of the joint velocity of the manipulator is shown in formula (6), which can be expressed as:

其中，

为机械臂的关节速度；J^*为雅可比矩阵，属于机器人学共识内容，可以通过码盘测量从而进行计算；P_w为力旋量投影空间对应的无量纲对角矩阵，可以通过式(4)得到；K_cart为导纳增益矩阵，可以由调试得到，例如对于优傲公司的UR5机器人往往采用0.4乘六行六列的单位矩阵；f_i通过机械臂上的力传感器获取的人机交互作用力。在雅可比矩阵J^*、力旋量投影空间对应的无量纲对角矩阵P_w、导纳增益矩阵K_cart、人机交互作用力f_i已知的情况下，可以求得机械臂的关节速度

in,

is the joint speed of the manipulator; J^* is the Jacobian matrix, which belongs to the consensus content of robotics and can be calculated by the code disc measurement;_Pw is the dimensionless diagonal matrix corresponding to the force screw projection space, which can be calculated by formula (4 ) is obtained; K_cart is the admittance gain matrix, which can be obtained by debugging. For example, the UR5 robot of Universal Robots often uses a unit matrix of 0.4 times six rows and six columns; f_i is obtained by the force sensor on the manipulator arm. force. When the Jacobian matrix J^* , the dimensionless diagonal matrix P_w corresponding to the force screw projection space, the admittance gain matrix K_cart , and the human-machine interaction force f_i are known, the joint velocity of the manipulator can be obtained

对机械臂的关节速度

进行控制，该控制属于高刚度控制，采用现有的机器人控制器或机械臂内置功能，从而可以对速度跟踪，达到控制机械臂的效果。Joint Velocities for Robotic Arms

Control, the control belongs to high stiffness control, using the existing robot controller or the built-in function of the robot arm, so that the speed can be tracked to achieve the effect of controlling the robot arm.

对虚拟约束切换的方法已由上述步骤S14给出，通过上述步骤S12中的力组的参数回归值α_r生成单位向量，并通过式(4)产生虚拟约束，得到虚拟约束矩阵。例如，通过参数回归值α_r(50°，50°)生成单位向量，并将其单位向量通过式(4)产生虚拟约束，得到虚拟约束矩阵

The method for switching the virtual constraints is given in the above step S14. The unit vector is generated by the parameter regression value α_r of the force group in the above step S12, and the virtual constraint is generated by formula (4) to obtain a virtual constraint matrix. For example, the unit vector is generated by the parameter regression value α_r (50°, 50°), and the unit vector is used to generate virtual constraints through formula (4), and the virtual constraint matrix is obtained

本申请的另一实施例的机械臂控制方法的流程如图3所示，包括：通过力传感器获取输入信息，作为训练数据进行聚类分析，其中聚类分析包括：利用EM算法(即，最大似然估计)对输入的训练数据进行参数估计，并生成GMM人类意图模型，即高斯混合模型人类意图模型；A flow chart of a method for controlling a robotic arm according to another embodiment of the present application is shown in FIG. 3 , including: acquiring input information through a force sensor, and performing cluster analysis as training data, wherein the cluster analysis includes: using an EM algorithm (that is, a maximum Likelihood estimation) to perform parameter estimation on the input training data, and generate a GMM human intent model, that is, a Gaussian mixture model human intent model;

通过力传感器获取力的输入信息，作为实时数据，并将输入的力在球坐标系下分解，获取交互力冲量，并将交互力冲量进行最大似然估计，基于聚类分析中的GMM人类意图模型对KL散度阈值进行判断，若大于KL散度阈值，则进行各向异性约束；若小于KL散度阈值，则进行各向同性约束。The force input information is obtained through the force sensor as real-time data, and the input force is decomposed in the spherical coordinate system to obtain the interaction force impulse, and the maximum likelihood estimation of the interaction force impulse is performed based on the GMM human intention in the cluster analysis. The model judges the KL divergence threshold. If it is greater than the KL divergence threshold, anisotropic constraints are implemented; if it is less than the KL divergence threshold, isotropic constraints are implemented.

从以上的描述中，可以看出，本申请实现了如下技术效果：From the above description, it can be seen that the application has achieved the following technical effects:

本申请的实施例通过获取多组力组，利用最大似然估计对力组的角度坐标进行参数回归，并确定力组的角度坐标和聚类模型中的各个高斯分模型之间的KL散度，并将各个KL散度与预设的KL散度进行判断，进而可以根据判断结果对虚拟约束进行切换，从而可以达到自动切换虚拟约束的效果。The embodiment of the present application obtains multiple sets of force groups, uses maximum likelihood estimation to perform parameter regression on the angle coordinates of the force groups, and determines the KL divergence between the angle coordinates of the force groups and each Gaussian model in the clustering model , and judge each KL divergence with the preset KL divergence, and then switch the virtual constraint according to the judgment result, so as to achieve the effect of automatically switching the virtual constraint.

根据本申请实施例，还提供了一种用于得到上述机械臂控制方法中的聚类模型的方法，如图4所示，该聚类模型训练方法包括如下的步骤S21、步骤S22：According to an embodiment of the present application, a method for obtaining a clustering model in the above-mentioned robotic arm control method is also provided. As shown in FIG. 4 , the clustering model training method includes the following steps S21 and S22:

S21：获取不少于两组的机械臂在不同位姿下的训练力组。S21: Obtain the training force groups of not less than two groups of robotic arms in different poses.

训练力组可以通过安装在机械臂上的力传感器直接获取。具体的，力传感器为多维力传感器。在该实施例中，示例的，力传感器通过三维力传感器或六维力传感器获取。通过力传感器获取的训练力组包括X、Y、Z轴对应的三个训练力分量。The training force group can be obtained directly through the force sensor installed on the robotic arm. Specifically, the force sensor is a multi-dimensional force sensor. In this embodiment, for example, the force sensor is acquired by a three-dimensional force sensor or a six-dimensional force sensor. The training force group obtained by the force sensor includes three training force components corresponding to the X, Y, and Z axes.

具体的，采集机械臂在X、Y、Z轴其中一个或多个方向拖动的不同姿态的第三预设数量的训练力组，力传感器获取的训练力组同理上述步骤S11中式(1)所示，在此不再赘述。Specifically, a third preset number of training force groups with different postures dragged by the robotic arm in one or more directions of the X, Y, and Z axes are collected, and the training force group obtained by the force sensor is the same as the formula in step S11 above (1 ), which will not be repeated here.

第三预设数量的训练力组可以和上述步骤S11中的第二预设数量的力组是通过力传感器获取的相同的力组，也可以是通过力传感器获取的不同的力组。在该实施例中，示例的，第三预设数量的训练力组和第二预设数量的力组是通过力传感器获取的不同的力组。示例的，第三预设数量的范围可以是2-10。实例的，第三预设数量是5。The third preset number of training force groups may be the same force group acquired by the force sensor as the second preset number of force groups in the above step S11, or may be different force groups acquired by the force sensor. In this embodiment, for example, the third preset number of training force groups and the second preset number of force groups are different force groups acquired by the force sensor. Exemplarily, the range of the third preset number may be 2-10. By way of example, the third preset number is five.

具体的，力传感器获取的训练力组如式(1)所示，可以表达为：Specifically, the training force group obtained by the force sensor is shown in formula (1), which can be expressed as:

S22：将多组训练力组对高斯混合模型进行训练并聚类，得到聚类模型。S22: Train and cluster the Gaussian mixture model with multiple training force groups to obtain a clustering model.

“将多组训练力组对高斯混合模型进行训练并聚类，得到聚类模型。”具体为：将多组训练力组转换为球坐标系下的角度坐标，得到多组训练力组角度坐标；将多组训练力组角度坐标作为回归变量，对进行高斯混合模型进行回归计算，得到第一预设数量的分模型的聚类结果；将聚类结果进行参数集合，得到聚类模型。"Train and cluster the Gaussian mixture model with multiple training force groups to obtain a clustering model." Specifically: convert the multiple training force groups into angular coordinates in the spherical coordinate system, and obtain the angle coordinates of the multiple training force groups Using the angle coordinates of multiple groups of training force groups as regression variables, performing regression calculation on the Gaussian mixture model to obtain the clustering results of the first preset number of sub-models; the clustering results are parameterized to obtain a clustering model.

具体的，建立高斯混合模型，其中高斯混合模型具有第一预设数量的分模型。高斯混合模型为具有不少于两个高斯分模型的模型，即具有第一预设数量不小于2。示例的，第一预设数量为2-6。实例的，第一预设数量为3，即高斯混合模型具有3个高斯分模型。Specifically, a Gaussian mixture model is established, wherein the Gaussian mixture model has a first preset number of sub-models. The Gaussian mixture model is a model with not less than two Gaussian sub-models, that is, the first preset number is not less than 2. For example, the first preset number is 2-6. For example, the first preset number is 3, that is, the Gaussian mixture model has 3 Gaussian sub-models.

具体的，训练力组可以通过上述步骤S21得到，将训练力组转换为球坐标系下的角度坐标的方法同理上述步骤S12中式(2)的求解方法，在此不再赘述。可以得到多组训练力组对应的多个角度坐标{α₁,α₂，…，α_e}，其中，e为第三预设数量，将多组训练力组对应的多个角度坐标{α₁,α₂，…，α_e}作为高斯混合模型的回归变量进行回归计算，可以得到一个分模型聚类结果，经过多次计算最终可以得到第一预设数量的分模型聚类结果，并将结果以参数结合的形式进行展示(如，存在2个分模型的高斯混合模型，可以是(0.2，(30°，7°)，(50°，5°))和(0.8，(53°，2°)，(10°，0°))的训练结果。)Specifically, the training force group can be obtained through the above step S21, and the method of converting the training force group into the angular coordinates in the spherical coordinate system is the same as the solution method of formula (2) in the above step S12, and will not be repeated here. Multiple_angular coordinates {α₁ , α₂ , .₁ , α₂ , ..., α_e } are used as the regression variables of the Gaussian mixture model for regression calculation, and a sub-model clustering result can be obtained. After several calculations, the first preset number of sub-model clustering results can be obtained, and The results are displayed in the form of a combination of parameters (for example, there are 2 sub-models of the Gaussian mixture model, which can be (0.2, (30°, 7°), (50°, 5°)) and (0.8, (53°) , 2°), (10°, 0°)) training results.)

需要说明的是，在附图的流程图示出的步骤可以在诸如一组计算机可执行指令的计算机系统中执行，并且，虽然在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤。It should be noted that the steps shown in the flowcharts of the accompanying drawings may be executed in a computer system, such as a set of computer-executable instructions, and, although a logical sequence is shown in the flowcharts, in some cases, Steps shown or described may be performed in an order different from that herein.

根据本申请实施例，还提供了一种用于实施上述机械臂控制方法的装置10，如图5所示，该机械臂控制装置10包括：According to an embodiment of the present application, adevice 10 for implementing the above-mentioned method for controlling a robot arm is also provided. As shown in FIG. 5 , thedevice 10 for controlling a robot arm includes:

获取模块11，用于获取机械臂的力组；anacquisition module 11 for acquiring the force group of the robotic arm;

参数回归模块12，用于将力组进行参数回归，得到力组的参数回归值；Theparameter regression module 12 is used to perform parameter regression on the force group to obtain the parameter regression value of the force group;

确定模块13，用于根据参数回归值和聚类模型，确定力组和聚类模型之间的KL散度，其中，聚类模型为通过机械臂的力组训练的高斯混合模型；The determiningmodule 13 is used to determine the KL divergence between the force group and the clustering model according to the parameter regression value and the clustering model, wherein the clustering model is a Gaussian mixture model trained by the force group of the mechanical arm;

切换模块14，用于根据KL散度对虚拟约束进行切换。Theswitching module 14 is used for switching the virtual constraints according to the KL divergence.

进一步的，参数回归模块12包括：Further, theparameter regression module 12 includes:

分解单元，用于将力组在球坐标系下分解，得到力组的角度坐标；The decomposition unit is used to decompose the force group in the spherical coordinate system to obtain the angular coordinates of the force group;

参数回归单元，用于利用最大似然估计，对力组的角度坐标进行参数回归，得到力组的参数回归值。The parameter regression unit is used to perform parameter regression on the angular coordinates of the force group by using maximum likelihood estimation to obtain the parameter regression value of the force group.

进一步的，切换模块14包括：Further, the switchingmodule 14 includes:

判断单元，用于判断KL散度是否大于KL散度阈值；a judgment unit for judging whether the KL divergence is greater than the KL divergence threshold;

各向异性约束单元，用于若KL散度大于KL散度阈值，则对机械臂产生各向异性虚拟约束；The anisotropic constraint unit is used to generate anisotropic virtual constraints on the robotic arm if the KL divergence is greater than the KL divergence threshold;

各向同性约束单元，用于若KL散度不大于第一KL散度阈值，则对机械臂产生各向同性虚拟约束。The isotropic constraint unit is used to generate an isotropic virtual constraint on the robotic arm if the KL divergence is not greater than the first KL divergence threshold.

进一步的，切换模块14还包括：Further, the switchingmodule 14 also includes:

控制单元，用于根据各向异性虚拟约束，确定机械臂的关节速度，以便以关节速度控制的方式控制机械臂运动。The control unit is used for determining the joint speed of the robot arm according to the anisotropic virtual constraint, so as to control the movement of the robot arm in a joint speed control manner.

具体的，本实施例中个模块的实现可以参考方法实施例中的相关实现，不再赘述。Specifically, for the implementation of each module in this embodiment, reference may be made to the relevant implementation in the method embodiment, and details are not described again.

从以上的描述中，可以看出，本申请实现了如下技术效果：From the above description, it can be seen that the application has achieved the following technical effects:

本申请的实施例通过获取多组力组，利用最大似然估计对力组的角度坐标进行参数回归，并确定力组的角度坐标和聚类模型中的各个高斯分模型之间的KL散度，并将各个KL散度与预设的KL散度进行判断，进而可以根据判断结果对虚拟约束进行切换，从而可以达到自动切换虚拟约束的效果。The embodiment of the present application obtains multiple sets of force groups, uses maximum likelihood estimation to perform parameter regression on the angle coordinates of the force groups, and determines the KL divergence between the angle coordinates of the force groups and each Gaussian model in the clustering model , and judge each KL divergence with the preset KL divergence, and then switch the virtual constraint according to the judgment result, so as to achieve the effect of automatically switching the virtual constraint.

显然，本领域的技术人员应该明白，上述的本申请的各模块或各步骤可以用通用的计算装置来实现，它们可以集中在单个的计算装置上，或者分布在多个计算装置所组成的网络上，可选地，它们可以用计算装置可执行的程序代码来实现，从而，可以将它们存储在存储装置中由计算装置来执行，或者将它们分别制作成各个集成电路模块，或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。这样，本申请不限制于任何特定的硬件和软件结合。Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present application can be implemented by a general-purpose computing device, and they can be centralized on a single computing device, or distributed in a network composed of multiple computing devices Alternatively, they can be implemented with program codes executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or they can be integrated into The multiple modules or steps are fabricated into a single integrated circuit module. As such, the present application is not limited to any particular combination of hardware and software.

以上所述仅为本申请的优选实施例而已，并不用于限制本申请，对于本领域的技术人员来说，本申请可以有各种更改和变化。凡在本申请的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本申请的保护范围之内。The above descriptions are only preferred embodiments of the present application, and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims (9)

Translated fromChinese

1.一种机械臂控制方法，其特征在于，包括：1. a mechanical arm control method, is characterized in that, comprises:获取机械臂的力组；Get the force group of the robotic arm;采集机械臂在X、Y、Z轴其中一个或多个方向拖动的不同姿态的第二预设数量的力组以及力组的表达式，力组的表达式为：Collect the second preset number of force groups and the expressions of the force groups in different poses dragged by the robotic arm in one or more directions of the X, Y, and Z axes, and the expressions of the force groups are:

f为力传感器获取的坐标系的X、Y、Z轴对应的三个力分量集合，f_x为力传感器获取的坐标系的X轴对应的力，f_y为力传感器获取的坐标系的Y轴对应的力，f_z为力传感器获取的坐标系的Z轴对应的力；f is the set of three force components corresponding to the X, Y and Z axes of the coordinate system obtained by the force sensor, f_x is the force corresponding to the X axis of the coordinate system obtained by the force sensor, and f_y is the Y of the coordinate system obtained by the force sensor The force corresponding to the axis, f_z is the force corresponding to the Z axis of the coordinate system obtained by the force sensor;将所述力组进行参数回归，得到所述力组的参数回归值；将所述力组在球坐标系下分解，得到所述力组的角度坐标，力组的角度坐标表达式为：Perform parameter regression on the force group to obtain the parameter regression value of the force group; decompose the force group in the spherical coordinate system to obtain the angular coordinates of the force group, and the angular coordinate expression of the force group is:

利用最大似然估计，对所述力组的角度坐标进行参数回归，得到力组的参数回归值；Using maximum likelihood estimation, parameter regression is performed on the angular coordinates of the force group to obtain the parameter regression value of the force group;根据所述参数回归值和聚类模型，确定所述力组和聚类模型之间的KL散度，其中，所述聚类模型为通过机械臂的力组训练的高斯混合模型；According to the parameter regression value and the clustering model, determine the KL divergence between the force group and the clustering model, wherein the clustering model is a Gaussian mixture model trained by the force group of the robotic arm;根据所述KL散度对虚拟约束进行自动切换。Virtual constraints are automatically switched according to the KL divergence.2.根据权利要求1所述的机械臂控制方法，其特征在于，所述根据所述KL散度对虚拟约束进行自动切换包括：2. The robotic arm control method according to claim 1, wherein the automatic switching of virtual constraints according to the KL divergence comprises:判断所述KL散度是否大于KL散度阈值；Determine whether the KL divergence is greater than the KL divergence threshold;若所述KL散度大于KL散度阈值，则对机械臂产生各向异性虚拟约束；If the KL divergence is greater than the KL divergence threshold, an anisotropic virtual constraint is generated on the robotic arm;若所述KL散度不大于KL散度阈值，则不对机械臂进行虚拟约束。If the KL divergence is not greater than the KL divergence threshold, no virtual constraint is performed on the robotic arm.3.根据权利要求2所述的机械臂控制方法，其特征在于，在所述判断所述KL散度是否大于KL散度阈值之后，所述方法还包括：3. The method for controlling a robotic arm according to claim 2, wherein after judging whether the KL divergence is greater than a KL divergence threshold, the method further comprises:若所述KL散度不大于KL散度阈值，则对机械臂产生各向同性虚拟约束。If the KL divergence is not greater than the KL divergence threshold, an isotropic virtual constraint is generated for the robotic arm.4.根据权利要求2所述的机械臂控制方法，其特征在于，在所述若所述KL散度大于KL散度阈值，则对机械臂产生各向异性虚拟约束之后，所述方法还包括：4 . The method for controlling a manipulator according to claim 2 , wherein, after the anisotropic virtual constraint is generated on the manipulator if the KL divergence is greater than the KL divergence threshold, the method further comprises: 5 . :根据所述各向异性虚拟约束，确定机械臂的关节速度，以便以关节速度控制的方式控制机械臂运动。According to the anisotropic virtual constraint, the joint speed of the manipulator is determined, so as to control the motion of the manipulator in a joint speed control manner.5.一种聚类模型训练方法，其特征在于，所述方法用于得到权利要求1-4中任一项所述的机械臂控制方法中的聚类模型，所述聚类模型训练方法包括：5. A clustering model training method, wherein the method is used to obtain the clustering model in the robotic arm control method according to any one of claims 1-4, and the clustering model training method comprises: :获取不少于两组的机械臂在不同位姿下的训练力组；Obtain no less than two groups of training force groups of the robotic arms in different poses;采集机械臂在X、Y、Z轴其中一个或多个方向拖动的不同姿态的第二预设数量的力组以及力组的表达式，力组的表达式为：Collect the second preset number of force groups and the expressions of the force groups in different poses dragged by the robotic arm in one or more directions of the X, Y, and Z axes. The expression of the force groups is:

其中，f为力传感器获取的坐标系的X、Y、Z轴对应的三个力分量集合，f_x为力传感器获取的坐标系的X轴对应的力，f_y为力传感器获取的坐标系的Y轴对应的力，f_z为力传感器获取的坐标系的Z轴对应的力；Among them, f is the set of three force components corresponding to the X, Y, and Z axes of the coordinate system obtained by the force sensor, f_x is the force corresponding to the X axis of the coordinate system obtained by the force sensor, and f_y is the coordinate system obtained by the force sensor. The force corresponding to the Y axis of , f_z is the force corresponding to the Z axis of the coordinate system obtained by the force sensor;将多组训练力组对高斯混合模型进行训练并聚类，得到聚类模型。The Gaussian mixture model is trained and clustered by multiple training force groups to obtain a clustering model.6.根据权利要求5所述的聚类模型训练方法，其特征在于，所述将多组训练力组对高斯混合模型进行训练并聚类，得到聚类模型包括：6. The clustering model training method according to claim 5, wherein the Gaussian mixture model is trained and clustered by multiple groups of training force groups, and the obtained clustering model comprises:将多组训练力组转换为球坐标系下的角度坐标，得到多组训练力组角度坐标，力组角度坐标表达式为：Convert multiple sets of training force groups to angular coordinates in the spherical coordinate system, and obtain the angular coordinates of multiple sets of training force groups. The expression of force group angle coordinates is:

将多组训练力组角度坐标作为回归变量，对进行高斯混合模型进行回归计算，得到第一预设数量的分模型的聚类结果；Using the angle coordinates of the multiple groups of training force groups as regression variables, performing regression calculation on the Gaussian mixture model to obtain the clustering results of the first preset number of sub-models;将所述聚类结果进行参数集合，得到聚类模型。The clustering results are parameterized to obtain a clustering model.7.一种机械臂控制装置，其特征在于，所述装置包括：7. A robotic arm control device, wherein the device comprises:获取模块，用于获取机械臂的力组；采集机械臂在X、Y、Z轴其中一个或多个方向拖动的不同姿态的第二预设数量的力组以及力组的表达式，力组的表达式为：The acquisition module is used to acquire the force group of the manipulator; collect the second preset number of force groups and the expressions of the force groups in different postures of the manipulator dragging in one or more directions of the X, Y, and Z axes, the force The expression for the group is:

其中，f为力传感器获取的坐标系的X、Y、Z轴对应的三个力分量集合，f_x为力传感器获取的坐标系的X轴对应的力，f_y为力传感器获取的坐标系的Y轴对应的力，f_z为力传感器获取的坐标系的Z轴对应的力；Among them, f is the set of three force components corresponding to the X, Y, and Z axes of the coordinate system obtained by the force sensor, f_x is the force corresponding to the X axis of the coordinate system obtained by the force sensor, and f_y is the coordinate system obtained by the force sensor. The force corresponding to the Y axis of , f_z is the force corresponding to the Z axis of the coordinate system obtained by the force sensor;参数回归模块，用于将所述力组进行参数回归，得到所述力组的参数回归值；将力组在球坐标系下分解，得到力组的角度坐标，力组的角度坐标表达式为：The parameter regression module is used to perform parameter regression on the force group to obtain the parameter regression value of the force group; decompose the force group in the spherical coordinate system to obtain the angular coordinates of the force group, and the angular coordinate expression of the force group is: :

利用最大似然估计，对力组的角度坐标进行参数回归，得到多个角度坐标对应的最大似然估计，即力组的参数回归值；Using the maximum likelihood estimation, perform parameter regression on the angle coordinates of the force group, and obtain the maximum likelihood estimation corresponding to multiple angle coordinates, that is, the parameter regression value of the force group;确定模块，用于根据所述参数回归值和聚类模型，确定所述力组和所述聚类模型之间的KL散度，其中，所述聚类模型为通过机械臂的力组训练的高斯混合模型；A determination module for determining the KL divergence between the force group and the clustering model according to the parameter regression value and the clustering model, wherein the clustering model is trained by the force group of the robotic arm Gaussian mixture model;切换模块，用于根据所述KL散度对虚拟约束进行自动切换。A switching module, configured to automatically switch the virtual constraints according to the KL divergence.8.计算机可读存储介质，其特征在于，所述计算机可读存储介质存储有计算机指令，所述计算机指令用于使所述计算机执行权利要求1-4任意一项所述的机械臂控制方法和/或权利要求5-6任意一项所述的聚类模型训练方法。8. A computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to make the computer execute the robotic arm control method according to any one of claims 1-4 and/or the clustering model training method described in any one of claims 5-6.9.一种机器人，包括：机械臂、传感器、至少一个处理器；以及与所述至少一个处理器通信连接的存储器；其中，所述存储器存储有可被所述至少一个处理器执行的计算机程序，所述计算机程序被所述至少一个处理器执行，以使所述至少一个处理器执行权利要求1-4任意一项所述的机械臂控制方法和/或权利要求5-6任意一项所述的聚类模型训练方法。9. A robot comprising: a robotic arm, a sensor, at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor , the computer program is executed by the at least one processor, so that the at least one processor executes the robotic arm control method described in any one of claims 1-4 and/or any one of claims 5-6. The clustering model training method described above.

Images