CN107443369A - A calibration-free servo control method for robotic arms based on inverse identification of visual measurement models - Google Patents

Abstract

The invention discloses a non-calibration servo control method of a robot arm based on inverse identification of a visual measurement model, belonging to the field of robot arm control. This method adopts the inverse identification of the measurement model based on binocular camera vision, and the obtained measurement feedback value is the position information of the manipulator, which is convenient for the construction of the control system and the optimization of the motion trajectory. In this way, the calibration-free visual servo control of the robot arm is realized.

Description

A calibration-free servo control method for robotic arms based on inverse identification of visual measurement models

technical field

The invention belongs to the field of robot systems, and relates to a calibration-free servo motion control of a robot arm based on inverse model identification.

Background technique

The mechanical arm generally refers to the multi-joint automatic mechanical equipment in the field of industrial applications. It is one of the earliest modern robots. It can imitate human actions, efficiently complete certain complex tasks, and liberate from tedious processing or dangerous situations. Human hands. The research of manipulator is a compound subject involving many fields, including control theory, mechanical structure, information engineering, computer technology, microcomputer principle and other mechanical and information knowledge. It has a wide range of applications, such as agricultural production, energy mining, industrial loading, automated production, service industry, military industry, etc., and will have a relatively long-term development in the future.

Visual servoing refers to a method of collecting image information of objects through optical devices or other non-contact sensors, and then using the image information as the feedback signal of the servo system to adjust the state of the controlled object or the output of the controller to complete servo control.

Through the image feedback of the camera to the working space, the controller is designed to guide the control method of the robotic arm to complete the task autonomously, and the internal and external parameters of the camera are not calibrated. The research on uncalibrated visual servo control of manipulators began in the 1990s. According to whether it converts the image space to Cartesian space, it can be divided into image-based and Cartesian space-based uncalibrated visual servo control. The image-based uncalibrated visual servo controller is more direct and conforms to the human visual system. It is generally realized by special layout or intelligent algorithm. Its requirements for the environment are lower than those of the controller based on Cartesian space, but it has the robustness of the control method. Non-optimal problems of sex and motion trajectories.

Through calibration, although visual servo control can be performed accurately, there are technical requirements in the calibration process, which not only makes it difficult to guarantee accuracy, but also has poor mobility. Calibration-free visual servoing refers to the control method that completes the control under the premise that the internal and external parameters of the camera are not accurately calculated or their exact values are not known during the entire control process. Usually, the uncalibrated visual servo is more versatile, and the requirements for installation parameters and control methods are not as strict as the calibrated visual servo. In the working space with complex external environment or large interference, the application of calibration-free visual servoing is more extensive.

Calibration-free visual servoing is the development direction of current research. The research on uncalibrated visual servoing of manipulators mainly focuses on two aspects: one is to improve the image Jacobian matrix or improve the anti-interference ability by adding auxiliary algorithms or data processing methods; the other is to use other intelligent algorithms to replace the image Jacobian matrix.

It is controlled by an intelligent algorithm, and because the Jacobian matrix is not used, the problems of inaccurate Jacobian matrix recursion and large amount of calculation are avoided. The intelligent control of uncalibrated visual servoing of manipulators currently researched is mostly realized by directly changing the control parameters through image deviation. This method lacks direct control over the motion parameters of the manipulator, so it is difficult to directly optimize its trajectory.

Therefore, there is currently a lack of a calibration-free visual servo direct control method that can overcome the above defects.

Contents of the invention

Using a binocular camera, because the image space is a two-dimensional plane, and the Cartesian space is a three-dimensional space, the monocular camera lacks depth information, so the description of the three-dimensional space using a monocular camera is not complete. The binocular camera can adopt a certain layout method, so as to make up for the missing depth information and describe the pose in the Cartesian space more accurately. The invention adopts the inverse identification of the measurement model based on the binocular camera vision to realize the uncalibrated visual servo control.

In order to achieve the goal, the present invention proposes the following technical solutions:

A calibration-free visual servo direct control method for a robotic arm based on inverse identification of a measurement model: comprising the following steps:

Step 1: The binocular camera is installed at a fixed position, and its visual layout mode is an orthogonal layout;

Step 2: set the motion position of mechanical arm, and based on binocular camera, obtain corresponding image feature parameters based on imaging plane;

Step 3: Based on the above data, carry out inverse identification modeling on the visual measurement model, and establish a model inverse identification based on the amount of image feature extraction and the movement position of the manipulator;

Step 4: Based on the above measurement inverse model, build a servo control system to realize calibration-free servo control.

The technical effect of the present invention is: based on the binocular camera vision, the inverse identification of the position measurement model of the manipulator is carried out, and the obtained measurement feedback value is the position information of the manipulator, which facilitates the construction of the control system and the optimization of the motion trajectory.

Description of drawings

In order to make the purpose, technical scheme and beneficial effect of the present invention clearer, the present invention provides the following drawings for illustration:

Figure 1 Binocular camera fixed position installation

Fig. 2 Robot arm servo control system based on inverse identification of visual measurement model

Figure 3 Feature extraction process of binocular vision measurement

detailed description

Preferred examples of the present invention are carried out below in conjunction with accompanying drawing

Because the present invention is aimed at the inverse identification of the binocular vision measurement model, to construct the servo control system of the manipulator. The specific implementation in step 1 is as shown in Figure 1, the binocular camera is installed at a fixed position, and its visual layout mode is an orthogonal layout;

In this example, step 2 is specifically implemented as shown in FIG. 2 . Set a set of motion reference positions x _set , y _set , z _set , Rx _set , Ry _set , Rz _set , and obtain the image characteristic parameters u ₁ , v ₁ , θ ₁ , u ₂ , v of the imaging plane based on binocular measurement ₂ , θ ₂ . The measurement principle in the process of binocular measurement is shown in Figure 3.

In this paper, two recognition points P ₁ and P ₂ are marked on the end of the manipulator, and the pose of the end of the manipulator in Cartesian space is reflected by the projection of these two points in the two cameras. They project four points P ₁₁ , P ₁₂ , P ₂₁ , P ₂₂ in total on the image planes of the two cameras. Theoretically, the selection of image features s=(P ₁₁ , P ₂₁ , P ₂₁ , P ₂₂ ) can meet the requirements of representing the end pose of the manipulator, but the image features are eight-dimensional and have a high dimension. The following methods can be used Perform dimensionality reduction.

Let θ ₁ be the angle between the vector (P ₁₁ , P ₁₂ ) and the u-axis of the camera-image plane coordinate, and θ ₂ be the angle between the vector (P ₂₁ , P ₂₂ ) and the camera-two image plane coordinate u-axis . The extracted image features can be expressed as s=(P11, P21, θ1, θ2). Wherein P11 and P21 are coordinate points, which can be expressed as P11=(u1, v1), P21=(u2, v2). That is, the selected 6 image features are u1, v1, u2, v2, θ1, θ2.

x _set , y _set , z _set , Rx _set , Ry _set , Rz _set are respectively the set position and speed of the robot arm in the Cartesian coordinate system;

In this example, step 3 is implemented based on image feature extraction quantities u ₁ , v ₁ , θ ₁ , u ₂ , v ₂ , θ ₂ and robot arm movement position information x, y, z, Rx, Ry, Rz. Inverse identification of visual measurement models. Due to the nonlinearity of the visual measurement process based on dual cameras, the identification methods with nonlinear models are applicable. Considering the strong fitting ability of neural network in nonlinear model identification, it is the mainstream method mainly adopted. In this example, the specific implementation of step 4 is shown in Figure 2. After the inverse identification of the visual measurement model is established, the feedback control quantities of the control system are x, y, z, Rx, Ry, Rz, thus realizing the Cartesian coordinate Servo control of position measurement information.

Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that it can be described in terms of form and Various changes may be made in the details without departing from the scope of the invention defined by the claims.

Claims (3)

1. A non-calibration servo control method of a robotic arm based on inverse identification of a visual measurement model, characterized in that, Step 1: The binocular camera is installed at a fixed position, and its visual layout mode is an orthogonal layout; Step 2: set the motion position of mechanical arm, and based on binocular camera, obtain corresponding image feature parameters based on imaging plane; Step 3: Based on the above data, carry out inverse identification modeling on the visual measurement model, and establish a model inverse identification based on the amount of image feature extraction and the movement position of the manipulator; Step 4: Based on the above measurement inverse model, build a servo control system to realize calibration-free servo control. 2. The inverse model based on visual features and binocular vision measurement are adopted, which simplifies the positioning process based on binocular vision and realizes the integrated measurement design of the position parameters of the manipulator. 3. The neural network and other methods are used to establish the inverse model of the relationship between visual features and the position parameters of the manipulator, thereby realizing the indirect measurement of the position parameters of the manipulator, thus simplifying the nonlinear modeling process of visual features and position parameters of the manipulator.