CN-121989253-A - Mechanical arm positioning error prediction and path planning method based on multi-algorithm cooperation

CN121989253ACN 121989253 ACN121989253 ACN 121989253ACN-121989253-A

Abstract

The invention discloses a mechanical arm positioning error prediction and path planning method based on multi-algorithm cooperation, and belongs to the technical field of industrial robot control and intelligent manufacturing. The method comprises the steps of mechanical arm data acquisition and preprocessing, forward kinematics modeling based on a neural network, positioning error prediction based on a random forest regression model, and iterative optimization solving of paths by adopting a multi-algorithm fusion method to obtain the mechanical arm movement path which meets the precision requirement and is optimal in target. The method realizes high-precision error prediction through random forests, realizes path collaborative optimization under precision constraint through multi-algorithm fusion strategy, ensures micron-level positioning precision and simultaneously gives consideration to motion efficiency, and can be widely applied to the high-end manufacturing fields of aerospace precision assembly, on-machine measurement and the like.

Inventors

YANG ZHIHAO
DING DAWEI
ZHANG KAIYAN

Assignees

南京邮电大学

Dates

Publication Date: 20260508
Application Date: 20260330

Claims (7)

1. The method for predicting the positioning error of the mechanical arm and planning the path based on the cooperation of multiple algorithms is characterized by comprising the following steps: s1, mechanical arm data acquisition and pretreatment; s2, modeling forward kinematics based on a neural network; s3, carrying out positioning error prediction based on a random forest regression model; And S4, carrying out iterative optimization solution on the path by adopting a multi-algorithm fusion method to obtain the mechanical arm movement path meeting the precision requirement and with the optimal target.
2. The method for predicting positioning errors and planning paths of mechanical arms based on multi-algorithm cooperation according to claim 1, wherein the step S1 specifically comprises: Planning a plurality of sampling points in a working space of the mechanical arm, controlling the mechanical arm to sequentially move to each sampling point, synchronously recording the actual three-dimensional position of the tail end of the mechanical arm, and simultaneously reading the theoretical calculation position from a mechanical arm controller; The acquired data includes a joint angle vector Q, an actual position P actual , a theoretical position P theoretical , and a positioning error vector Δp=p actual - P theoretical .
3. The method for predicting positioning errors and planning paths of mechanical arms based on cooperation of multiple algorithms according to claim 2, wherein the step S2 specifically comprises: s21, constructing a multilayer feedforward neural network, wherein the structure is as follows: The input layer is used for carrying out normalization processing when the number of the neurons is equal to the number of the joints of the mechanical arm and inputting data, so that the angles of all joints are in similar magnitude; a first hidden layer comprising 128 neurons, the activation function being a ReLU; A second hidden layer containing 64 neurons, the activation function is also ReLU; The output layer comprises 3 neurons which respectively correspond to the three-dimensional coordinates of the end effector and adopt a linear activation function; s22, randomly dividing the data set acquired in the step S1 into three subsets according to a proportion, wherein the three subsets comprise a training set, a verification set and a test set; S23, after training by using a training set, evaluating the performance of the model by using a test set which does not participate in training, wherein indexes comprise root mean square errors and decision coefficients, and S24, finally obtaining a neural network model NN forward , wherein the function of the neural network model NN is P pred = NN forward (Q), and P pred is the predicted terminal position.
4. The method for predicting positioning errors and planning paths of mechanical arms based on cooperation of multiple algorithms according to claim 3, wherein the step S3 specifically comprises: Taking the joint angle vector Q obtained in the step S1 as input and the corresponding positioning error vector delta P as output, constructing a random forest regression model RF error , and carrying out positioning error prediction, wherein the positioning error prediction is as follows: e pred =RF error (Q); Where e pred is the predicted positioning error value.
5. The method for predicting positioning errors and planning paths of mechanical arms based on cooperation of multiple algorithms according to claim 4, wherein the step S4 specifically comprises: s41, initializing a population, and randomly generating a plurality of groups of path parameters to serve as an initial solution; S42, running a plurality of intelligent optimization algorithms in parallel, and independently searching candidate paths meeting precision constraint in respective search spaces; s43, periodically collecting optimal solutions of the algorithms, and carrying out result fusion to generate new candidate solutions; S44, repeating the step S43 until the candidate solution meets the termination condition, and obtaining a mechanical arm movement track which meets all position precision requirements and is optimal in time.
6. The method for predicting positioning errors and planning paths of a manipulator based on multi-algorithm coordination according to claim 5, wherein the intelligent optimization algorithm adopted in the step S42 comprises GWO algorithm, PSO algorithm, GWO-PSO enhancement algorithm and adaptive selector algorithm.
7. The method of claim 5, wherein the fusing of the results in step S43 includes using any of direct selection, weighted averaging, elite crossing, and local search enhancement.

Description

Mechanical arm positioning error prediction and path planning method based on multi-algorithm cooperation Technical Field The invention relates to the technical field of industrial robot control and intelligent manufacturing, in particular to a mechanical arm positioning error prediction and path planning method based on multi-algorithm cooperation, which is particularly suitable for scenes such as automatic manufacturing, precise assembly, on-machine measurement and the like with strict requirements on positioning accuracy. Background In current industrial robot applications, the positioning accuracy of the robotic arm directly determines product quality and process consistency. With the development of intelligent manufacturing to a high-precision tip direction, the fields of aerospace precision assembly, semiconductor chip packaging, medical instrument processing and the like have provided sub-millimeter-level and even micrometer-level requirements for absolute positioning accuracy of industrial robots. However, due to multi-source error factors such as geometric parameter errors, joint clearances, flexible deformation, thermal errors and the like, the absolute positioning accuracy of the industrial robot is often far lower than the repeated positioning accuracy, and the industrial robot becomes a main bottleneck for limiting the application of the industrial robot in the high-accuracy field. The traditional path planning method mainly focuses on kinematic constraint (such as obstacle avoidance and track smoothing) and kinematic inverse solution calculation, and the control on positioning errors often depends on the following two ways: 1. And compensating errors based on a physical model by establishing a physical model such as geometric parameter errors, joint clearances, thermal deformation and the like. This method requires accurate measurement of a large number of parameters and the model can fail over time as a result of wear and temperature changes. For example, DH parameter calibration methods, while capable of compensating for partial geometric errors, are difficult to handle nonlinear, time-varying error factors. 2. Real-time feedback based on external sensor, real-time measuring terminal position by laser tracker, binocular vision system, etc. and on-line adjusting. This approach has control delays, is difficult to ensure accuracy at high speeds, increases system costs, and is limited in industrial field applications. The common problem in the prior art is that error control is passive and delayed, and path planning and precision guarantee mutually fracture, resulting in rejection either to preserve precision, efficiency, or lack of precision. Specifically, the existing method lacks accurate prediction capability of positioning errors of the mechanical arm in a full working space, and cannot actively avoid areas with larger errors in a planning stage, and meanwhile, the traditional path planning algorithm (such as A, RRT and the like) only considers obstacle avoidance and kinematic feasibility, and does not integrate accuracy constraint into a planning process as an optimization target. In recent years, some researches begin to try to apply a machine learning method to robot error compensation, but most of the research remains in an offline calibration stage, and deep fusion of error prediction and real-time path planning cannot be achieved. In addition, when solving the path planning problem of complex constraint, the single optimization algorithm (such as genetic algorithm and particle swarm optimization) often has the defects of low convergence speed, easy trapping in local optimum, poor robustness and the like. Therefore, a method for controlling a robot arm is needed that actively considers the error effect during the planning stage and cooperatively optimizes the precision and efficiency. Disclosure of Invention The invention aims to provide a mechanical arm control method which can actively consider error influence in a planning stage and cooperatively optimize precision and efficiency. In order to achieve the above purpose, the method adopted by the invention is as follows: a mechanical arm positioning error prediction and path planning method based on multi-algorithm cooperation comprises the following steps: s1, mechanical arm data acquisition and pretreatment; s2, modeling forward kinematics based on a neural network; s3, carrying out positioning error prediction based on a random forest regression model; And S4, carrying out iterative optimization solution on the path by adopting a multi-algorithm fusion method to obtain the mechanical arm movement path meeting the precision requirement and with the optimal target. Further, the step S1 specifically includes: Planning a plurality of sampling points in a working space of the mechanical arm, controlling the mechanical arm to sequentially move to each sampling point, synchronously recording the actual three-dimensional position of the ta