Riemannian Manifold Excavator Behavior Cloning Method and Device Based on Improved Dual Quaternion

Abstract

The present application relates to a Riemannian manifold excavator behavior cloning method and device based on improved biquaternions. The method includes: constructing a forward dynamics model of the excavator by the DH parameter method, obtaining the transfer matrix in the Cartesian coordinate system of the excavator end, the translation variables and rotation variables in the Cartesian coordinate system of the excavator according to the forward dynamics model, obtaining the quaternion representation of the rotation part of the excavator end, obtaining the biquaternion representation of the position and attitude of the entire excavator according to the quaternion representation, representing the position and attitude of the excavator end in the Cartesian coordinate system as the Riemannian manifold biquaternion representation in the tangent space of the Riemannian manifold through Riemannian manifold mapping, constructing a non-parametric Bayesian Gaussian process model, calculating the mean and variance of the output behavior data components given the query timestamp, and finally obtaining the behavior cloning result. Using this method can improve the expression ability of the excavator operation behavior.

Description

Technical Field

[0001] This application relates to the technical field of excavator behavior cloning, and particularly to a method and device for excavator behavior cloning based on the Riemannian manifold of improved dual quaternions. Background Art

[0002] In the development process of the intelligentization of construction machinery, although intelligent unmanned excavators have broad prospects, they currently face many problems that need to be solved urgently. Their autonomous excavation efficiency is extremely low. During actual operations, it is difficult to complete excavation tasks quickly and accurately. Compared with manual excavation, the efficiency gap is obvious, and they cannot truly replace operators for efficient excavation, resulting in a significant increase in construction costs. At the same time, the embodied intelligent technology for autonomous excavation is not yet mature, especially in imitating and cloning the excavation behavior of operators, lacking effective methods. This makes intelligent unmanned excavators unable to accurately learn and reproduce human operation skills and difficult to adapt to complex and changeable excavation conditions, severely restricting the development and practical application of this technology. Therefore, seeking an effective method to improve the imitation and learning ability of intelligent unmanned excavators has become the key to promoting the progress of embodied intelligent technology for autonomous excavation. Summary of the Invention

[0003] Based on this, in view of the above technical problems, it is necessary to provide a method and device for excavator behavior cloning based on the Riemannian manifold of improved dual quaternions.

[0004] A method for excavator behavior cloning based on the Riemannian manifold of improved dual quaternions, the method comprising:

[0005] Construct a forward dynamics model of the excavator by the DH parameter method, and obtain a transfer matrix in the Cartesian coordinate system of the excavator end, a translation variable and a rotation variable in the Cartesian coordinate system of the excavator end according to the forward dynamics model;

[0006] Obtain a quaternion representation of the rotating part at the end of the excavator according to the transfer matrix and the rotation variable, and obtain a dual quaternion representation of the position and attitude of the entire excavator according to the translation variable and the quaternion representation;

[0007] According to the dual quaternion representation, represent the position and attitude of the end of the excavator in the Cartesian coordinate system as a dual quaternion representation in the tangent space of the Riemannian manifold through Riemannian manifold mapping;

[0008] Project the behavior data of the excavator operator onto the tangent space of the Riemannian manifold, construct a non-parametric Bayesian Gaussian process model, and given a query timestamp according to the non-parametric Bayesian Gaussian process model, calculate and output the mean and variance of the behavior data component, and obtain the final behavior cloning result through the inverse mapping of the Riemannian manifold mapping.

[0009] In one of the embodiments, it further includes: electronically transforming the excavator; the transformation includes: arranging IMU sensors at the boom, arm, bucket, and vehicle slewing mechanism parts of the excavator, and installing proportional valves on the manual mechanical drive of the excavator and connecting a data acquisition card.

[0010] In one of the embodiments, it further includes: constructing a forward kinematic model of the excavator by the DH parameter method; the DH parameters include: the distance along the x-axis between two adjacent axes , the angle of rotation of two adjacent axes along the x-axis , the distance of translation of two adjacent axes along the z-axis , the angle of rotation of two adjacent axes along the z-axis ;

[0011] Obtain the transfer matrix in the Cartesian coordinate system at the end of the excavator according to the forward kinematic model as:

[0012] ;

[0013] wherein, , , , . , , , , , , , are respectively the attitude data collected by the IMU sensors placed at the boom, arm, bucket, and vehicle slewing mechanism of the excavator;

[0014] The translation variables in the Cartesian coordinate system of the excavator are:

[0015] .

[0016] The rotation variables are expressed in Euler angles as:

[0017] .

[0018] In one of the embodiments, it further includes: obtaining the quaternion representation of the rotating part at the end of the excavator according to the transfer matrix, translation variables, and rotation variables as:

[0019] .

[0020] wherein, , , and are in Cartesian coordinate representation.

[0021] In one embodiment, it further includes: obtaining a biquaternion representation of the position and attitude of the entire excavator according to the quaternion representation, including:

[0022] .

[0023] In one embodiment, it further includes: obtaining the representation of the end position and attitude of the excavator as:

[0024] .

[0025] Wherein, is the position of the end of the excavator in the Cartesian coordinate system, is the subscript of the position and attitude of the excavator at a certain moment.

[0026] In one embodiment, it further includes: representing the position and attitude of the end of the excavator in the Cartesian coordinate system as a Riemannian manifold biquaternion in the tangent space of the Riemannian manifold through Riemannian manifold mapping as:

[0027] .

[0028] In one embodiment, it further includes: projecting the behavior data of the excavator onto the tangent space of the Riemannian manifold to construct a non-parametric Bayesian Gaussian process model; the squared exponential covariance function of the non-parametric Bayesian Gaussian process model is:

[0029] .

[0030] Wherein, , and are the hyperparameters of the Gaussian process model, and are the timestamps corresponding to the components of the excavator behavior data respectively, is the Kronecker symbol;

[0031] Given the timestamp of the query, the mean and variance of the output excavator behavior data component are respectively:

[0032] ;

[0033] ;

[0034] Wherein, and are the covariance matrices of the timestamps corresponding to the excavator behavior data and the query data timestamp respectively, is the corresponding cross-covariance matrix. is the set of operator's behavior cloning data vectors;

[0035] The final behavior cloning result is obtained through the inverse mapping of the Riemann manifold mapping as:

[0036] ;

[0037] where represents the behavior cloning result.

[0038] A Riemann manifold excavator behavior cloning device based on improved dual quaternions, the device includes:

[0039] A forward dynamics model construction module, configured to construct a forward dynamics model of the excavator by the DH parameter method, and obtain a transfer matrix in the Cartesian coordinate system of the excavator end, a translation variable and a rotation variable in the Cartesian coordinate system of the excavator;

[0040] A quaternion representation module, configured to obtain a quaternion representation of the rotating part of the excavator end according to the transfer matrix and the rotation variable, and obtain a dual quaternion representation of the position and attitude of the entire excavator according to the translation variable and the quaternion representation;

[0041] A Riemann mapping module, configured to represent the position and attitude of the excavator end in the Cartesian coordinate system as a Riemann manifold dual quaternion representation in the tangent space of the Riemann manifold through Riemann manifold mapping according to the dual quaternion representation;

[0042] A behavior cloning module, configured to project the behavior data of the excavator onto the tangent space of the Riemann manifold, construct a non-parametric Bayesian Gaussian process model, calculate the mean and variance of the output behavior data components given a query timestamp according to the non-parametric Bayesian Gaussian process model, and obtain the final behavior cloning result through the inverse mapping of the Riemann manifold mapping.

[0043] A computer device, including a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0044] Construct a forward dynamics model of the excavator by the DH parameter method, and obtain a transfer matrix in the Cartesian coordinate system of the excavator end, a translation variable and a rotation variable in the Cartesian coordinate system of the excavator;

[0045] Obtain a quaternion representation of the rotating part of the excavator end according to the transfer matrix and the rotation variable, and obtain a dual quaternion representation of the position and attitude of the entire excavator according to the translation variable and the quaternion representation;

[0046] According to the dual quaternion representation, the end position and attitude representation of the excavator are obtained. Through the Riemannian manifold mapping, the position and attitude of the end of the excavator in the Cartesian coordinate system are represented as the dual quaternion representation in the tangent space of the Riemannian manifold.

[0047] Project the behavior data of the excavator onto the tangent space of the Riemannian manifold, construct a non-parametric Bayesian Gaussian process model. According to the non-parametric Bayesian Gaussian process model, given the query timestamp, calculate the mean and variance of the output behavior data components, and obtain the final behavior cloning result through the inverse mapping of the Riemannian manifold mapping.

[0048] In the above Riemannian manifold excavator behavior cloning method and device based on the improved dual quaternion, in the case where the traditional Euler angle method is easily affected by the gimbal lock problem and a single quaternion is difficult to comprehensively represent complex postures, this application can more accurately and comprehensively describe the posture state of the excavator through the improved dual quaternion method, avoiding possible mathematical singularity problems during the operation. The method based on the Riemannian manifold can utilize the geometric structure of the manifold to accurately model in the non-linear space and capture complex behaviors. In addition, the introduction of the improved dual quaternion can effectively describe multi-degree-of-freedom motion, while the Riemannian manifold theory can better depict the connection of the behavior trajectory in time and space, thereby enhancing the expression ability of the model for the excavator operation behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic flow chart of a Riemannian manifold excavator behavior cloning method based on an improved dual quaternion in an embodiment;

[0050] Figure 2 It is a structural block diagram of a Riemannian manifold excavator behavior cloning device based on an improved dual quaternion in an embodiment;

[0051] Figure 3 It is an internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0053] In one embodiment, as Figure 1 shown, a Riemannian manifold excavator behavior cloning method based on an improved dual quaternion is provided, including the following steps:

[0054] Step 102: Construct the forward dynamics model of the excavator using the DH parameter method. Obtain the transfer matrix in the Cartesian coordinate system of the excavator's end, as well as the translational variables and rotational variables in the Cartesian coordinate system of the excavator.

[0055] The DH (Denavit - Hartenberg) parameter method is a standard method widely used in robot kinematic modeling. By defining a series of link and joint parameters, it mathematically describes the relative positions and orientations of the various joints and links of the robot (here, the excavator). For a mechanical system like an excavator composed of multiple links (boom, arm, bucket, etc.) and joints, the DH parameter method can systematically establish its kinematic model.

[0056] Through the above steps, the transfer matrix in the Cartesian coordinate system of the excavator's end, the translational variables, and the rotational variables can be obtained based on the forward dynamics model constructed using the DH parameter method, thereby accurately describing the position and orientation of the excavator's end effector in the Cartesian space.

[0057] Step 104: Obtain the quaternion representation of the rotating part of the excavator's end based on the transfer matrix and rotational variables. Based on the translational variables and the quaternion representation, obtain the bi - quaternion representation of the position and orientation of the entire excavator.

[0058] A bi - quaternion is an extension of a quaternion. It can represent both rotation and translation simultaneously and is suitable for describing the complete pose of a rigid body in three - dimensional space.

[0059] Step 106: Obtain the representation of the excavator's end position and orientation based on the bi - quaternion representation. Through Riemannian manifold mapping, represent the position and orientation of the excavator's end in the Cartesian coordinate system as a bi - quaternion representation in the tangent space of the Riemannian manifold.

[0060] A Riemannian manifold is a smooth manifold with a Riemannian metric, which provides a mathematical framework for studying curved spaces. In robot kinematics, the space formed by the position and orientation of the end of an excavator is usually non-linear, and Riemannian manifolds can well model such non-linear spaces. The tangent space is a linear approximation space at a certain point on the manifold, which facilitates linear operations and calculations on the non-linear manifold. Through the logarithmic map, it can be mapped from the biquaternion space (a specific form belonging to Riemannian manifolds) to the tangent space. The logarithmic map converts the non-linear biquaternion representation into a linear tangent space representation, facilitating subsequent mathematical analysis and processing. After the logarithmic map, the position and orientation of the end of the excavator originally represented by biquaternions in the Cartesian coordinate system are converted into a vector in the tangent space of the Riemannian manifold, and this vector is the biquaternion representation of the Riemannian manifold. In the tangent space, operations such as interpolation and differentiation can be carried out more conveniently, and it also helps to analyze and process the motion data of the excavator by using methods such as linear algebra and machine learning.

[0061] Step 108: Project the behavior data of the excavator onto the tangent space of the Riemannian manifold, construct a non-parametric Bayesian Gaussian process model, and according to the non-parametric Bayesian Gaussian process model, given the query timestamp, calculate the mean and variance of the output behavior data components, and obtain the final behavior cloning result through the inverse mapping of the Riemannian manifold mapping.

[0062] In the above Riemannian manifold excavator behavior cloning method based on the improved biquaternion, when the traditional Euler angle method is easily affected by the gimbal lock problem and a single quaternion is difficult to comprehensively represent complex postures, this application can more accurately and comprehensively describe the posture state of the excavator through the improved biquaternion method, avoiding possible mathematical singularity problems during the operation process. The method based on the Riemannian manifold can utilize the geometric structure of the manifold to accurately model in the non-linear space and capture complex behaviors. In addition, the introduction of the improved biquaternion can effectively describe multi-degree-of-freedom motions, while the Riemannian manifold theory can better characterize the connection of the behavior trajectory in time and space, thereby enhancing the model's expression ability for the operation behavior of the excavator.

[0063] In one embodiment, the excavator is electrically controlled and transformed. The transformation includes: setting IMU sensors at the boom, arm, bucket, and vehicle body slewing mechanism parts of the excavator, and installing proportional valves on the manual mechanical drive of the excavator and connecting a data acquisition card.

[0064] Specifically, attitude acquisition sensors such as IMUs are placed on the boom (arm), stick (forearm), bucket, and vehicle slewing mechanism of the excavator, and the data is transmitted back to the Jetson Orin Nano industrial control computer via the CAN bus to achieve the function of collecting behavior cloning data and the attitude data of the end-effector Cartesian space coordinate points of the excavator. Additionally, for the drive system of the excavator, a proportional valve needs to be installed on the traditional manual mechanical drive and then connected to a data acquisition card to achieve the mutual conversion from analog to digital quantities, and further, to achieve the electric control of the excavator.

[0065] In one embodiment, the forward dynamics model of the excavator is constructed by the DH parameter method; the DH parameters include: the distance along the x-axis between two adjacent axes , the angle of rotation of two adjacent axes along the x-axis , the distance of translation of two adjacent axes along the z-axis , and the angle of rotation of two adjacent axes along the z-axis ;

[0066] The transfer matrix in the end-effector Cartesian coordinate system of the excavator is obtained according to the forward dynamics model as:

[0067] ;

[0068] where , , , . , , , , , , , is the attitude data collected by the IMU sensor;

[0069] The translation variables in the excavator Cartesian coordinate system are:

[0070] ;

[0071] The rotation variables are expressed in Euler angles as:

[0072] .

[0073] The forward dynamics model of the excavator is the kinematic model of a spatial four-bar linkage, and the method of DH ( ) parameters is shown in the following table:

[0074]

[0075] The position and attitude of the excavator in the Cartesian coordinate system can be fully obtained through the IMU data placed on each body of the excavator.

[0076] In one embodiment, according to the transfer matrix, translation variables, and rotation variables, the quaternion representation of the rotating part at the end of the excavator is obtained as:

[0077] ;

[0078] Among them, , , and are represented in the Cartesian coordinate system.

[0079] Furthermore, based on the quaternion representation and adding the translation part, the biquaternion representation of the position and attitude of the entire excavator is obtained, including:

[0080] .

[0081] In addition, based on the biquaternion representation, the position and attitude of the end of the excavator are represented as:

[0082] ;

[0083] Among them, is the position of the end of the excavator in the Cartesian coordinate system, is the subscript of the position and attitude of the excavator at a certain moment.

[0084] Through Riemannian manifold mapping, the position and attitude of the end of the excavator in the Cartesian coordinate system are represented as a biquaternion in the tangent space of the Riemannian manifold as:

[0085] .

[0086] In one embodiment, the behavior data of the excavator is projected onto the tangent space of the Riemannian manifold to construct a nonparametric Bayesian Gaussian process model; the squared exponential covariance function of the nonparametric Bayesian Gaussian process model is:

[0087] ;

[0088] Among them, , and are the hyperparameters of the Gaussian process model, and are the timestamps corresponding to the components of the excavator behavior data respectively, is the Kronecker symbol;

[0089] Given the timestamp , obtain the mean value of the output mining behavior data component and variance respectively as:

[0090] ;

[0091] ;

[0092] Among them, and are respectively the covariance matrices of the time stamps corresponding to the excavator behavior data and the query data time stamps, is the corresponding cross-covariance matrix. is the set of operator's behavior cloning data vectors;

[0093] The final behavior cloning result obtained through the inverse mapping of the Riemannian manifold mapping is:

[0094] ;

[0095] Among them, represents the behavior cloning result.

[0096] It should be understood that although Figure 1 each step in the flowchart of Figure 1 is shown in sequence according to the indication of the arrow, these steps are not necessarily executed in sequence according to the indication of the arrow. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover,

[0097] In one embodiment, as Figure 2 shown, a Riemannian manifold excavator behavior cloning device based on an improved biquaternion is provided, including: a forward dynamics model construction module 202, a quaternion representation module 204, a Riemannian mapping module 206, and a cloning module 208, where:

[0098] The forward dynamics model construction module 202 is used to construct the forward dynamics model of the excavator by the DH parameter method, and obtain the transfer matrix in the end Cartesian coordinate system of the excavator, the translation variable and the rotation variable in the excavator Cartesian coordinate system according to the forward dynamics model.

[0099] The quaternion representation module 204 is configured to obtain the quaternion representation of the rotating part at the end of the excavator according to the transfer matrix and the rotation variable, and obtain the biquaternion representation of the position and attitude of the entire excavator according to the translation matrix and the quaternion representation.

[0100] The Riemann mapping module 206 is configured to obtain the position and attitude representation of the end of the excavator according to the biquaternion representation, and represent the position and attitude of the end of the excavator in the Cartesian coordinate system as the Riemann manifold biquaternion representation in the tangent space of the Riemann manifold through Riemann manifold mapping.

[0101] The behavior cloning module 208 is configured to project the behavior data of the excavator onto the tangent space of the Riemann manifold, construct a non-parametric Bayesian Gaussian process model, calculate the mean and variance of the output behavior data component given the query timestamp according to the non-parametric Bayesian Gaussian process model, and obtain the final behavior cloning result through the inverse mapping of the Riemann manifold mapping.

[0102] In one embodiment, a transformation module is further included, which is configured to perform an electric control transformation on the excavator; the transformation includes: setting IMU sensors at the boom, arm, bucket, and vehicle slewing mechanism parts of the excavator, and installing proportional valves on the manual mechanical drive of the excavator and connecting a data acquisition card.

[0103] In one embodiment, the forward dynamics model construction module 202 is further configured to construct the forward dynamics model of the excavator by the DH parameter method; the DH parameters include: the distance along the x-axis between two adjacent axes , the angle of rotation along the x-axis between two adjacent axes , the distance of translation along the z-axis between two adjacent axes , and the angle of rotation along the z-axis between two adjacent axes ;

[0104] The transfer matrix of the end of the excavator in the Cartesian coordinate system is obtained according to the forward dynamics model as:

[0105] ;

[0106] where , , , . , , , , , , , are the attitude data collected by the IMU sensor;

[0107] The translation variables of the excavator in the Cartesian coordinate system are:

[0108] ;

[0109] The rotation variables are represented by Euler angles as:

[0110] .

[0111] In one embodiment, the quaternion representation module 204 is further configured to obtain the quaternion representation of the rotating part at the end of the excavator according to the transfer matrix, translation variables, and rotation variables as:

[0112] ;

[0113] Wherein, , , and are represented in the Cartesian coordinate system.

[0114] In one embodiment, the quaternion representation module 204 is further configured to obtain the biquaternion representation of the position and attitude of the entire excavator according to the quaternion representation as:

[0115] .

[0116] In one embodiment, the quaternion representation module 204 is further configured to obtain the position and attitude representation of the end of the excavator according to the biquaternion representation as:

[0117] ;

[0118] Wherein, is the position of the end of the excavator in the Cartesian coordinate system, is the subscript of the position and attitude of the excavator at a certain moment.

[0119] In one embodiment, the Riemann mapping module 206 is further configured to represent the position and attitude of the end of the excavator in the Cartesian coordinate system as a Riemann manifold biquaternion representation in the tangent space of the Riemann manifold through Riemann manifold mapping as:

[0120] .

[0121] In one embodiment, the behavior cloning module 208 is further configured to project the behavior data of the excavator onto the tangent space of the Riemann manifold to construct a non-parametric Bayesian Gaussian process model; the squared exponential covariance function of the non-parametric Bayesian Gaussian process model is:

[0122] ;

[0123] Among them, , and are hyperparameters of the Gaussian process model, and are the timestamps corresponding to the excavator behavior data components respectively, is the Kronecker symbol;

[0124] Given the timestamp of the query , the mean value and variance of the output excavating behavior data component are respectively:

[0125] ;

[0126] ;

[0127] Among them, and are the covariance matrices of the timestamps corresponding to the excavator behavior data and the query data timestamp respectively, is the corresponding cross-covariance matrix. is the set of operator's behavior cloning data vectors;

[0128] The final behavior cloning result obtained through the inverse mapping of the Riemann manifold mapping is:

[0129] ;

[0130] Among them, represents the behavior cloning result.

[0131] For the specific limitations of the Riemann manifold excavator behavior cloning device based on the improved biquaternion, reference can be made to the limitations of the Riemann manifold excavator behavior cloning method based on the improved biquaternion in the above text, which will not be elaborated here. Each module in the above Riemann manifold excavator behavior cloning device based on the improved biquaternion can be implemented in whole or in part through software, hardware and their combination. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0132] In one embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as shown in Figure 3As shown. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, it implements a Riemann manifold excavator behavior cloning method based on improved dual quaternions. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0133] Those skilled in the art can understand that Figure 3 the structure shown in [the figure] is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0134] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it implements the steps of the method in the above embodiment.

[0135] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0136] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0137] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A Riemannian manifold excavator behavior cloning method based on improved dual quaternions, characterized in that The method includes: Constructing a forward dynamics model of an excavator by the DH parameter method, and obtaining a transfer matrix in the Cartesian coordinate system of the excavator end, a translational variable and a rotational variable in the Cartesian coordinate system of the excavator end according to the forward dynamics model; Obtaining a quaternion representation of the rotating part of the excavator end according to the transfer matrix and the rotational variable, and obtaining a biquaternion representation of the position and attitude of the entire excavator according to the translational variable and the quaternion representation; According to the biquaternion representation, representing the position and attitude of the excavator end in the Cartesian coordinate system as a biquaternion representation of the Riemannian manifold tangent space through Riemannian manifold mapping; Projecting the behavior data of the excavator operator onto the Riemannian manifold tangent space, constructing a non-parametric Bayesian Gaussian process model, and given a query timestamp according to the non-parametric Bayesian Gaussian process model, calculating the mean and variance of the output behavior data components, and obtaining the final behavior cloning result through the inverse mapping of the Riemannian manifold mapping.

2. The method according to claim 1, wherein The method further includes: Performing an electric control transformation on the excavator; the transformation includes: setting IMU sensors at the boom, arm, bucket and vehicle body slewing mechanism parts of the excavator, and installing proportional valves on the manual mechanical drive of the excavator and connecting a data acquisition card.

3. The method according to claim 2, characterized in that Constructing a forward dynamics model of an excavator by the DH parameter method, and obtaining a transfer matrix in the Cartesian coordinate system of the excavator end, a translational variable and a rotational variable in the Cartesian coordinate system of the excavator, including: Construct the forward dynamic model of the excavator by the DH parameter method; the DH parameters include: the distance along the x-axis between two adjacent axes , the angle of rotation along the x-axis of two adjacent axes , the distance of translation along the z-axis of two adjacent axes , the angle of rotation along the z-axis of two adjacent axes ; The transfer matrix in the Cartesian coordinate system of the end of the excavator is obtained according to the forward dynamics model as follows: Among them, , , , , , , , , , , , are respectively the attitude data collected by the IMU sensors placed on the boom, arm, bucket and slewing mechanism of the excavator. The translational variable in the Cartesian coordinate system of the excavator is: The rotational variable is represented by Euler angles as: 。 4. The method according to claim 3, characterized in that, Obtaining a quaternion representation of the rotating part of the excavator end according to the transfer matrix and the rotational variable, including: Obtaining a quaternion representation of the rotating part of the excavator end according to the transfer matrix, translational variable and rotational variable as: Among them, , , and are represented in a Cartesian coordinate system.

5. The method according to claim 4, wherein Obtaining a biquaternion representation of the position and attitude of the entire excavator according to the translational variable and the quaternion representation as: 。 6. The method according to claim 5, characterized in that, Obtaining the position and attitude representation of the excavator end according to the biquaternion representation, including: Obtaining the position and attitude representation of the excavator end according to the biquaternion representation as: Among them, is the position of the end of the excavator in the Cartesian coordinate system, is the subscript of the position and attitude of the excavator at a certain moment.

7. The method according to claim 6, wherein Representing the position and attitude of the excavator end in the Cartesian coordinate system as a biquaternion representation in the Riemannian manifold tangent space through Riemannian manifold mapping, including: Representing the position and attitude of the excavator end in the Cartesian coordinate system as a biquaternion representation in the Riemannian manifold tangent space through Riemannian manifold mapping as: 。 8. The method according to claim 7, wherein Projecting the behavior data of the excavator onto the Riemannian manifold tangent space, constructing a non-parametric Bayesian Gaussian process model, and given a query timestamp according to the non-parametric Bayesian Gaussian process model, calculating the mean and variance of the output behavior data components, and obtaining the final behavior cloning result through the inverse mapping of the Riemannian manifold mapping, including: Projecting the behavior data of the excavator onto the Riemannian manifold tangent space, constructing a non-parametric Bayesian Gaussian process model; the squared exponential covariance function of the non-parametric Bayesian Gaussian process model is: Among them, , and are the hyperparameters of the Gaussian process model, and are the timestamps corresponding to the excavator behavior data components respectively, is the Kronecker symbol; Timestamp of the given query , to obtain the mean of the mined behavior data component of the output and variance respectively as follows: Among them, and are the covariance matrices corresponding to the timestamp of the excavator behavior data and the query data timestamp respectively, is the corresponding cross-covariance matrix, is the set of operator's behavior cloning data vectors; Obtaining the final behavior cloning result through the inverse mapping of the Riemannian manifold mapping as: Among them, represents the result of behavior cloning.

9. A Riemannian manifold excavator behavior cloning device based on improved dual quaternions, characterized in that, The device includes: The forward dynamics model construction module is used to construct the forward dynamics model of the excavator by the DH parameter method, and obtain the transfer matrix in the Cartesian coordinate system of the end of the excavator, the translation variables and rotation variables in the Cartesian coordinate system of the excavator; The quaternion representation module is used to obtain the quaternion representation of the rotating part at the end of the excavator according to the transfer matrix and rotation variables, and obtain the biquaternion representation of the position and attitude of the whole excavator according to the translation variables and the quaternion representation; The Riemann mapping module is used to obtain the representation of the end position and attitude of the excavator according to the biquaternion representation, and represent the position and attitude of the end of the excavator in the Cartesian coordinate system as the Riemann manifold biquaternion representation in the tangent space of the Riemann manifold through Riemann manifold mapping; The behavior cloning module is used to project the behavior data of the excavator onto the tangent space of the Riemann manifold, construct a nonparametric Bayesian Gaussian process model, and calculate the mean and variance of the output behavior data components given the query timestamp according to the nonparametric Bayesian Gaussian process model, and obtain the final behavior cloning result through the inverse mapping of the Riemann manifold mapping.

Images