CN-122018513-A - Robot obstacle avoidance method and device based on dimension reduction motion parameters and storage medium

CN122018513ACN 122018513 ACN122018513 ACN 122018513ACN-122018513-A

Abstract

The application relates to the technical field of robot obstacle avoidance, and discloses a robot obstacle avoidance method and device based on dimension reduction motion parameters, and a storage medium, wherein the method comprises the steps of recording an initial motion track of an obstacle when the obstacle in a target working area is monitored, and fitting the initial motion track into an initial motion parameter matrix under a robot working coordinate system; the method comprises the steps of obtaining a local base coordinate transformation matrix containing rotation parameter variables and non-orthogonal parameter variables, carrying out multi-round assignment on the rotation parameter variables and the non-orthogonal parameter variables, utilizing the obtained local base coordinate transformation matrix to transform an initial motion parameter matrix into a transformed parameter matrix after each round of assignment, carrying out sparsification processing, calculating a sparse parameter matrix, determining target rotation parameters and target non-orthogonal parameters according to the sparse parameter matrix, taking the target rotation parameters, the target non-orthogonal parameters and the corresponding sparse parameter matrix as dimension reduction parameter representation, predicting the motion trail of an obstacle according to the dimension reduction parameter representation, and generating a robot avoidance path based on the motion trail.

Inventors

ZHANG DINGSEN
YU YIFEI
WANG HONGXIAO
ZHANG ZIYE

Assignees

东北大学

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (9)

1. The robot obstacle avoidance method based on the dimension reduction motion parameters is characterized by comprising the following steps of: the method comprises the steps that a target working area is monitored in real time through a vision sensor carried by a robot, when an obstacle appears in the target working area, an initial motion track of the obstacle in a three-dimensional space is recorded, and the initial motion track is fitted into an initial motion parameter matrix of the robot under a working coordinate system; Obtaining a local base coordinate transformation matrix containing a rotation parameter variable and a non-orthogonal parameter variable, wherein the rotation parameter variable is used for representing a rotation gesture from the working coordinate system to a local base coordinate system, and the non-orthogonal parameter variable is used for representing an included angle between two coordinate axes in the local base coordinate system; Performing multi-round assignment on the rotation parameter variable and the non-orthogonal parameter variable, after each round of assignment, transforming the initial motion parameter matrix into a transformed parameter matrix under the local base coordinate system by using the obtained local base coordinate transformation matrix, performing sparsification processing on the transformed parameter matrix to obtain a sparse parameter matrix, calculating a reconstruction error between a reconstruction motion track corresponding to the sparse parameter matrix and the initial motion track, and taking an assignment corresponding to the minimum reconstruction error in the multi-round assignment as a target rotation parameter and a target non-orthogonal parameter; taking the target rotation parameters, the target non-orthogonal parameters and the corresponding sparse parameter matrix as a dimension reduction parameter representation of the initial motion parameter matrix; and predicting the motion trail of the obstacle according to the dimension reduction parameter representation, generating an avoidance path of the robot based on the predicted motion trail, and controlling the robot to move along the avoidance path.
2. The method of claim 1, wherein the rotation parameter variables include an X-axis rotation angle variable, a Y-axis rotation angle variable, and a Z-axis rotation angle variable, and wherein prior to the obtaining the local basis coordinate transformation matrix comprising rotation parameter variables and non-orthogonal parameter variables, the method further comprises: Constructing respective corresponding basic rotation matrixes based on an X-axis rotation angle variable, a Y-axis rotation angle variable and a Z-axis rotation angle variable respectively, and multiplying the basic rotation matrixes in sequence to obtain an initial rotation matrix for describing posture adjustment of a working coordinate system of the robot; determining two target coordinate axes constrained by the non-orthogonal parameter variables, constructing a non-orthogonal transformation matrix for adapting to the geometric form of the motion track of the obstacle according to the two target coordinate axes and the non-orthogonal parameter variables, and solving an inverse matrix of the non-orthogonal transformation matrix; And multiplying the initial rotation matrix with the inverse matrix to obtain a local base coordinate transformation matrix containing rotation parameter variables and non-orthogonal parameter variables.
3. The method according to claim 1, wherein transforming the initial motion parameter matrix into a transformed parameter matrix in the local base coordinate system by using the obtained local base coordinate transformation matrix, and performing a sparsification process on the transformed parameter matrix to obtain a sparse parameter matrix, includes: Multiplying the obtained local basic coordinate transformation matrix with an initial motion parameter matrix in the working coordinate system, transforming the initial motion parameter matrix into a local basic coordinate system aligned with the main motion direction of the obstacle, and obtaining a transformed parameter matrix; And reserving elements with absolute values exceeding a preset threshold value for each element in the transformed parameter matrix, and setting the rest elements to zero to obtain a sparse parameter matrix, wherein the sparse parameter matrix is used for screening out motion components which contribute significantly to the motion description of the obstacle.
4. The method of claim 1, wherein said fitting the initial motion trajectory to an initial motion parameter matrix in a working coordinate system of the robot comprises: Acquiring track position points corresponding to the initial motion track at a plurality of sampling time points; Calculating basis function value vectors corresponding to each sampling time point based on a predefined basis function set, wherein the basis function set comprises a plurality of basis functions used for representing motion components of the obstacle; Constructing a design matrix based on basis function value vectors corresponding to all sampling time points, and constructing an observation matrix based on track position points corresponding to all sampling time points; substituting the design matrix and the observation matrix into a least square fitting model, and solving to obtain the initial motion parameter matrix.
5. The method of claim 4, wherein the calculating a reconstruction error between the reconstructed motion profile corresponding to the sparse parameter matrix and the initial motion profile comprises: determining a reconstruction motion track corresponding to the initial motion track based on the sparse parameter matrix and the basis function set, and calculating an initial reconstruction error between the reconstruction motion track and the initial motion track, wherein the initial reconstruction error is used for measuring the fitting precision of the obstacle track; Identifying the number of non-zero elements in the sparse parameter matrix; Based on the initial reconstruction errors and the number of non-zero elements, calculating the reconstruction errors between the reconstruction motion trail corresponding to the sparse parameter matrix and the initial motion trail through a comprehensive objective function, wherein the comprehensive objective function is used for balancing the obstacle avoidance precision of the robot and the calculation complexity of a robot control system; the comprehensive objective function is as follows: ; Wherein, the Representing a reconstruction error between the reconstructed motion trail and the initial motion trail corresponding to the sparse parameter matrix, Representing the variation of the rotation parameter in question, Representing the non-orthogonal parameter variables in question, Representing the number of said non-zero elements, Representing the matrix of the sparse parameters, Representing transformed parameter matrices Is a number of L1 norms of (c), Representing the said initial reconstruction error(s), Indicating that a preset error threshold value is to be provided, Representing a preset weight coefficient.
6. The method of claim 1, wherein prior to said multi-round assignment of said rotational parameter variable to said non-orthogonal parameter variable, said method further comprises: constructing a candidate heuristic algorithm set suitable for a robot obstacle avoidance scene, wherein the candidate heuristic algorithm set comprises a plurality of heuristic optimization algorithms; For each heuristic optimization algorithm, independently running the heuristic optimization algorithm for a plurality of times, and calculating a corresponding obstacle motion parameter optimization evaluation index vector based on a group of rotation parameters and non-orthogonal parameters output by each running, wherein the obstacle motion parameter optimization evaluation index vector comprises the number of non-zero elements in a corresponding sparse parameter matrix, an average reconstruction error between a reconstruction motion track and an initial motion track, a maximum local error between the reconstruction motion track and the initial motion track, and optimization time consumption of single running; optimizing the evaluation index vector based on all barrier motion parameters of each heuristic optimization algorithm, and constructing an original evaluation data set; Generating a plurality of resampled subsets based on the original evaluation data set by a put-back sampling method; For each resampling subset, respectively calculating a comprehensive score of each heuristic optimization algorithm under the resampling subset, wherein the comprehensive score is calculated based on the obstacle motion parameter optimization evaluation index vector; For each resampling subset, determining a heuristic optimization algorithm with the highest comprehensive score as an optimal candidate algorithm of the resampling subset; And counting the times of each heuristic optimization algorithm which is determined to be an optimal candidate algorithm in all resampling subsets, and taking the heuristic optimization algorithm with the highest times as a target optimization algorithm for reducing the motion parameters of the obstacle, so as to carry out multi-round assignment on the rotation parameter variable and the non-orthogonal parameter variable based on the target optimization algorithm.
7. The method of claim 6, wherein for each resampling subset, calculating a composite score for each heuristic optimization algorithm under the resampling subset, respectively, comprises: for each resampling subset, determining each heuristic optimization algorithm contained in the resampling subset, determining all obstacle motion parameter optimization evaluation index vectors corresponding to the heuristic optimization algorithm from the resampling subset, and calculating the average value of the heuristic optimization algorithm on each evaluation index according to all the obstacle motion parameter optimization evaluation index vectors, wherein the average value is used for representing the average performance of the heuristic optimization algorithm under a robot obstacle avoidance scene; optimizing an evaluation index vector based on all obstacle motion parameters in the resampling subset for each evaluation index, and determining a maximum value and a minimum value of the evaluation index; And for each heuristic optimization algorithm, calculating the normalized score of each evaluation index of the heuristic optimization algorithm in the resampling subset according to the average value of each evaluation index of the heuristic optimization algorithm and the maximum value and the minimum value, and carrying out weighted summation on the normalized score of each evaluation index according to a preset weight coefficient to obtain the comprehensive score of the heuristic optimization algorithm in the resampling subset.
8. Robot keeps away barrier device based on dimension reduction motion parameter, its characterized in that includes: The fitting module is used for monitoring a target working area in real time through a vision sensor carried by the robot, recording an initial motion track of an obstacle in a three-dimensional space when the obstacle is monitored to appear in the target working area, and fitting the initial motion track into an initial motion parameter matrix under a working coordinate system of the robot; The matrix acquisition module is used for acquiring a local base coordinate transformation matrix containing a rotation parameter variable and a non-orthogonal parameter variable, wherein the rotation parameter variable is used for representing the rotation gesture from the working coordinate system to the local base coordinate system, and the non-orthogonal parameter variable is used for representing the included angle between two coordinate axes in the local base coordinate system; The sparse processing module is used for carrying out multi-round assignment on the rotation parameter variable and the non-orthogonal parameter variable, after each round of assignment, utilizing the obtained local base coordinate transformation matrix to transform the initial motion parameter matrix into a transformed parameter matrix under the local base coordinate system, carrying out sparsification processing on the transformed parameter matrix to obtain a sparse parameter matrix, calculating a reconstruction error between a reconstruction motion track corresponding to the sparse parameter matrix and the initial motion track, and taking an assignment corresponding to the minimum reconstruction error in the multi-round assignment as a target rotation parameter and a target non-orthogonal parameter; The dimension reduction parameter determining module is used for taking the target rotation parameter, the target non-orthogonal parameter and the corresponding sparse parameter matrix as dimension reduction parameter representation of the initial motion parameter matrix; The obstacle avoidance module is used for predicting the motion trail of the obstacle according to the dimension reduction parameter representation, generating an avoidance path of the robot based on the predicted motion trail and controlling the robot to move along the avoidance path.
9. A storage medium having stored thereon a computer program, which when executed by a processor, implements the method of any of claims 1 to 7.

Description

Robot obstacle avoidance method and device based on dimension reduction motion parameters and storage medium Technical Field The application relates to the technical field of robot obstacle avoidance, in particular to a robot obstacle avoidance method and device based on dimension reduction motion parameters and a storage medium. Background With the development of industrial robots to high-speed, high-precision and intelligent directions, the robot working area gradually expands from a structured closed space to a semi-structured open space. In such environments, obstructions (e.g., personnel, moving tools, logistics equipment, etc.) often dynamically intrude into the robot work area in an unintended manner. In order to realize safe and reliable man-machine cooperation and dynamic obstacle avoidance, the robot needs to sense and predict the track state of the movement obstacle in real time and plan the avoidance path of the robot based on the prediction result. However, the motion of dynamic obstacles usually presents complex features such as non-stationary, non-linear, multi-modal, abrupt changes in direction, and local disturbances, which pose serious challenges for real-time sensing and prediction of robots. The track modeling and representing method in the existing robot obstacle avoidance system is mainly divided into two types, namely, one type is to perform high-dimensional function fitting on the three-dimensional track of the obstacle under the working coordinate system of the robot, the track form is described by adopting fixed basis functions such as polynomials, trigonometric functions, convolution kernels and the like, and the other type is based on the track expression method occupying a grid or high-dimensional potential space, and complex motion is represented by constructing a high-dimensional feature space. The method has the advantages that the fitting parameter redundancy is high, the geometric meaning is undefined, the high-frequency disturbance and the direction jump are difficult to express under limited parameters, the accuracy and the reliability of the robot obstacle avoidance path planning are directly affected, and the method can describe complex motions, but has extremely high characteristic dimension and heavy calculation load, and is difficult to meet the severe requirements of the robot obstacle avoidance on real-time response. In addition, the existing method generally directly transmits high-dimensional parameters to a robot motion planning module after track modeling is completed, so that the robot calculates resources to be tense when processing a multi-obstacle scene, and high-efficiency and safe dynamic obstacle avoidance is difficult to realize. Disclosure of Invention In view of the above, the application provides a robot obstacle avoidance method, a device and a storage medium based on dimension reduction motion parameters, which enable a coordinate system to be adaptively aligned with a main motion direction of an obstacle track by introducing a dynamic local base coordinate system containing rotation parameters and non-orthogonal parameters, so that a complex motion track which is difficult to align under a fixed robot working coordinate system is converted into a sparse parameter matrix with high concentrated energy under the local base coordinate system, the problems of high fitting parameter redundancy and undefined geometric meaning caused by dislocation of the coordinate system and the track direction in the traditional method are effectively solved, a large number of secondary components which contribute to track description weakly are abandoned through sparsification processing, the dimension of the motion parameter is greatly reduced, and on the basis, a target rotation parameter, a target non-orthogonal parameter and the sparse parameter matrix are jointly used as dimension reduction parameter representation, so that a robot motion planning module does not need to directly process the high dimension redundancy parameter, but performs track prediction and avoidance path generation based on the reduced low dimension characteristics, the calculation load under a multi-obstacle scene is remarkably reduced, the requirement of the robot obstacle avoidance on strict response is met, and the obstacle avoidance is high in real time, and the safety and the reliability are realized. According to one aspect of the application, there is provided a robot obstacle avoidance method based on dimension reduction motion parameters, comprising: the method comprises the steps that a target working area is monitored in real time through a vision sensor carried by a robot, when an obstacle appears in the target working area, an initial motion track of the obstacle in a three-dimensional space is recorded, and the initial motion track is fitted into an initial motion parameter matrix of the robot under a working coordinate system; Obtaining a local base coordinate transformation