AGV combined control method and device

Abstract

The invention discloses an AGV combined control method and device, and the method comprises the steps: firstly constructing an AGV pose state equation and a servo motor model which take the midpoint of a center connection line of a left driving wheel and a right driving wheel as a reference point, building the full-link quantization correlation of motor input to an AGV pose, building an AGV internal model, and configuring an IMC-PID control unit; a driving path image is collected, a path center line is extracted through processing, and the position and angle deviation of the AGV relative to the center line are calculated; the deviation is input into an IMC-PID control unit, and the speed adjustment amount of the left driving wheel and the right driving wheel is calculated; and inputting the speed adjustment amount into a driving system, correcting the pose of the AGV by adjusting the differential rotation speed, and collecting the real-time pose and the rotation speed of a driving wheel to form double-feedback closed-loop control. According to the invention, AGV pose dynamic correction can be realized, and trajectory tracking precision, response speed and system anti-interference capability are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of AGVs, and in particular to a method and apparatus for the combined control of AGVs. Background Technology

[0002] Automated Guided Vehicles (AGVs) are core automated equipment in fields such as intelligent manufacturing and logistics warehousing. The accuracy of their autonomous navigation trajectory tracking and the stability of their motion directly determine the operating efficiency and safety of the automated system. Their control core is to achieve precise posture correction by adjusting the speed of the drive wheels to ensure stable travel along the preset path.

[0003] Current AGV control solutions suffer from numerous technical problems: Modeling often involves constructing separate kinematic or drive motor models, lacking models adapted to actual motion characteristics, resulting in insufficient control basis; path image processing and deviation detection are easily affected by environmental interference, making accuracy difficult to guarantee and failing to provide reliable guidance information; the mainstream PID controllers often fail to meet requirements in actual operation due to environmental complexity, making it difficult to quickly eliminate distance and angle errors, easily leading to large curvatures in the driving trajectory; furthermore, various improved control algorithms either have high computational load and poor real-time performance, or weak anti-interference capabilities and poor parameter adaptability, resulting in poor overall trajectory tracking performance.

[0004] Meanwhile, AGVs exhibit time-varying and nonlinear dynamic characteristics, making them susceptible to uncertainties such as ground friction and load fluctuations. Furthermore, the existing navigation, perception, and control systems lack sufficient linkage, making it difficult to balance response speed, control accuracy, and anti-interference capabilities in trajectory tracking. Consequently, the system exhibits poor robustness and cannot meet the high-precision and high-stability operation requirements of AGVs under complex working conditions, becoming a pressing technical challenge for the industry. Summary of the Invention

[0005] The main objective of this invention is to overcome the shortcomings of existing technologies, such as the lack of quantitative correlation between motor input and posture in AGV modeling, insufficient accuracy in visual deviation detection, slow error elimination due to environmental influences in PID control and easy trajectory deviation, and poor linkage between perception and control. This invention proposes a joint control method and device for AGVs to achieve dynamic correction of AGV posture and improve the accuracy of trajectory tracking, response speed and anti-interference capability.

[0006] The present invention adopts the following technical solution:

[0007] An AGV joint control method includes the following steps:

[0008] For AGV, construct the AGV pose state equation with the midpoint of the line connecting the centers of the left and right drive wheels as the reference point; construct a servo motor model for the AGV drive motor; establish a full-link quantization association from motor input to AGV pose based on the AGV pose state equation and the servo motor model; and construct the AGV internal model and configure the IMC-PID control unit based on the full-link quantization association.

[0009] Images of the AGV's travel path are acquired, the images are processed to obtain the centerline of the path, and the deviation information of the AGV's position and angle relative to the centerline is calculated.

[0010] The deviation information is input into the IMC-PID control unit, which calculates the speed adjustment of the left and right drive wheels of the AGV in real time based on the magnitude and rate of change of the deviation.

[0011] The speed adjustment is transmitted to the AGV's drive system, and the AGV's posture is dynamically corrected by adjusting the differential speed of the left and right drive wheels. At the same time, the real-time posture of the AGV and the real-time speed of the left and right drive wheels are collected as feedback signals, and the feedback signals are input to the IMC-PID control unit to form a closed-loop control.

[0012] The AGV is a two-speed differential structure. The AGV pose state equation is obtained by discretizing the AGV kinematic model using discrete-time representation combined with Euler approximation. The pose state equation includes the quantitative correlation between the AGV heading angle, planar displacement and the angular velocities of the left and right drive wheels.

[0013] The servo motor model is a three-loop vector control model of position, speed and current of permanent magnet synchronous motor. The voltage, torque and motion equations of the drive motor are established using the dq axis method, and the regulation law of motor input voltage on drive wheel angular velocity is clarified, providing a motor-side quantitative basis for the establishment of full-link quantitative correlation.

[0014] The full-link quantization association is a two-level quantization association from motor input to drive wheel angular velocity and from drive wheel angular velocity to AGV pose, realizing the full-link motion characteristic mapping from motor input to AGV pose. The AGV internal model is a digital abstract model of this full-link quantization association, matching the actual motion dynamic characteristics of the AGV.

[0015] The processing of the path image is as follows: weighted average grayscale conversion, median filtering for noise reduction, Otsu's binarization, Canny algorithm edge extraction, and line-by-line scanning fitting. The weighted average method assigns different weights to the red, green, and blue components of the image according to the characteristics of human brightness perception. The line-by-line scanning fitting calculates the midpoint coordinates of the left and right edge points of the path and fits them to obtain a continuous path centerline.

[0016] The deviation information of position and angle is calculated by comparing the geometric position of the AGV's own positioning reference with the center line of the path and the AGV's driving direction with the extension direction of the center line of the path, respectively, providing a quantitative basis for position and posture correction.

[0017] The IMC-PID control unit is equipped with a first-order filter. Based on the AGV's internal model, the IMC-PID control unit predicts the AGV's pose change trend under different speed adjustment amounts. It combines proportional adjustment of the deviation magnitude and derivative adjustment of the deviation change rate to calculate the independent speed adjustment amounts of the left and right drive wheels, thereby achieving a combination of feedforward prediction and feedback adjustment.

[0018] The adjustment of the differential speed of the left and right drive wheels is specifically as follows: based on the independent speed adjustment of the left and right drive wheels, the independent speed of the two drive wheels is adjusted in a non-uniform manner. The heading angle of the AGV is corrected by the speed difference of the two drive wheels, and the position offset of the AGV is corrected by the synchronous increase and decrease of the speed of the two drive wheels, thus completing the coordinated dynamic correction of the posture.

[0019] The closed-loop control formed by the feedback signal is a dual-feedback closed-loop control. The real-time pose signal of the AGV is used to correct the trajectory deviation of the AGV, and the real-time speed signals of the left and right drive wheels are used to correct the execution deviation of the drive system. The signal acquisition frequency of the dual-feedback closed-loop control is matched with the calculation frequency of the IMC-PID control unit to realize the real-time correction of the deviation.

[0020] An AGV joint control device includes:

[0021] The model building module is used to build the AGV pose state equation with the midpoint of the line connecting the centers of the left and right drive wheels as the reference point, build a servo motor model for the AGV drive motor, establish a full-link quantization association from the motor input to the AGV pose based on the AGV pose state equation and the servo motor model, and build the AGV internal model and configure the IMC-PID control unit according to the full-link quantization association.

[0022] The vision processing module is used to acquire images of the AGV's travel path, process the images to obtain the centerline of the path, and calculate the deviation information of the AGV's position and angle relative to the centerline.

[0023] The control calculation module, connected to the vision processing module and the model building module, is used to receive the deviation information and transmit it to the IMC-PID control unit. The IMC-PID control unit calculates the speed adjustment of the left and right drive wheels of the AGV in real time according to the magnitude and rate of change of the deviation.

[0024] The drive correction module, connected to the control calculation module, is used to receive the speed adjustment amount and transmit it to the AGV's drive system. It dynamically corrects the AGV's posture by adjusting the differential speed of the left and right drive wheels, and at the same time, it collects the real-time posture of the AGV and the real-time speed of the left and right drive wheels as feedback signals and transmits them to the IMC-PID control unit to form a closed-loop control.

[0025] As can be seen from the above description of the present invention, compared with the prior art, the present invention has the following beneficial effects:

[0026] 1. This invention establishes a full-link quantitative association from motor input to AGV posture, and combines visual path recognition with IMC-PID closed-loop control to achieve accurate and stable tracking of AGV trajectory. Compared with traditional control methods, it has the advantages of faster response, smaller overshoot, and stronger anti-interference ability.

[0027] 2. This invention constructs a full-link quantitative association based on the AGV kinematic model and the servo motor model. The resulting AGV internal model can accurately reflect the actual dynamic characteristics of motion, providing a reliable model basis for the control algorithm and improving the rationality and engineering applicability of the control strategy.

[0028] 3. This invention extracts the path centerline through a multi-level image processing flow and uses a geometric comparison method to quantify position and angle deviations, thereby improving the reliability of visual navigation and the accuracy of deviation calculation, and providing an accurate basis for real-time pose correction.

[0029] 4. The IMC-PID control unit of the present invention combines feedforward prediction and feedback regulation, which can quickly calculate the speed adjustment of the drive wheel according to the magnitude and rate of change of the deviation, thereby improving the dynamic response performance and robustness of the system while ensuring control accuracy.

[0030] 5. This invention adopts dual feedback closed-loop control and achieves posture collaborative correction through differential speed, which can simultaneously correct trajectory deviation and actuator deviation, making AGV posture correction smoother and faster, and further improving operational stability and tracking accuracy. Attached Figure Description

[0031] Figure 1 This is the main flowchart of the method of the present invention;

[0032] Figure 2 A simplified diagram of the kinematic model of an AGV;

[0033] Figure 3 For servo motor three-loop adjustment system;

[0034] Figure 4 Equivalent model of permanent magnet synchronous motor;

[0035] Figure 5This is a general structural diagram of internal mold control;

[0036] Figure 6 This is a structural diagram of the equivalent internal model control system;

[0037] Figure 7 To convert the image to grayscale;

[0038] Figure 8 This is a screenshot showing the noise reduction effect.

[0039] Figure 9 This is the result of binarization;

[0040] Figure 10 This is an image showing the edge extraction result;

[0041] Figure 11 The result of extracting the center line;

[0042] Figure 12 Path tracking diagram under no-interference conditions;

[0043] Figure 13 The path tracking diagram is shown when the interference is 0.5.

[0044] Figure 14 The tracking error comparison chart is shown when the interference is 0.5.

[0045] Figure 15 The path tracking diagram is shown when the interference is 1.

[0046] Figure 16 The tracking error comparison chart is shown when the interference is 1.

[0047] Figure 17 The tracking error comparison chart is shown when the interference is 2.

[0048] Figure 18 The tracking error comparison chart is shown when the interference is 3.

[0049] Figure 19 Comparison chart of tracking errors when interference is 5;

[0050] Figure 20 Path tracking diagram when interference is 2;

[0051] Figure 21 Path tracking diagram when interference is 3;

[0052] Figure 22 Path tracking diagram when interference is 5;

[0053] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Detailed Implementation

[0054] The present invention will be further described below through specific embodiments.

[0055] This embodiment proposes a joint control method for AGVs; see [link / reference]. Figure 1 It includes the following steps:

[0056] 1) For AGV, construct the AGV pose state equation with the midpoint of the line connecting the centers of the left and right drive wheels as the reference point; construct a servo motor model for the AGV drive motor; establish the full-link quantization association from motor input to AGV pose based on the AGV pose state equation and servo motor model, and then construct the AGV internal model and configure the IMC-PID control unit according to the full-link quantization association.

[0057] The AGV is a two-speed differential structure. Its posture state equation is obtained by discretizing the AGV kinematic model using discrete time representation combined with Euler approximation. The posture state equation contains the quantitative correlation between the AGV heading angle, planar displacement and the angular velocity of the left and right drive wheels, which can accurately reflect the dynamic motion state of the AGV.

[0058] The above AGV posture state equation is constructed with the AGV kinematic model as the core, and combines the vehicle force constraints reflected by the chassis dynamics model, as well as the relationship between the torque and speed transmission of the drive wheels represented by the drive wheel dynamics model. It is finally determined based on the comprehensive kinematic geometry relationship and the actual force conditions, and can accurately reflect the overall motion state of the AGV with the midpoint of the line connecting the centers of the left and right drive wheels as the reference point.

[0059] Among them, the AGV kinematic model is used to characterize the relationship between the AGV's motion state and its own geometric parameters. When modeling, the core geometric parameters such as the AGV's wheel radius and wheelbase are first determined. Then, based on the motion law of the two-speed differential AGV's straight-line driving and turning driving, kinematic equations that can describe the relationship between its posture, speed and other motion states and geometric parameters are established.

[0060] In this embodiment, the AGV adopts a structure with a front omnidirectional wheel and a rear dual drive wheel, using an electric motor as the power source. Driving and steering actions are achieved by matching the rotational speeds of the left and right drive wheels. Because the actual operating conditions of the AGV are complex, it is difficult to directly establish an accurate kinematic mathematical model. Therefore, the following reasonable assumptions and simplifications are made during modeling: the wheels run at low speed on a two-dimensional horizontal surface, the influence of wheel slippage is ignored, and the two drive wheels remain on the same axis.

[0061] Based on the above simplified conditions, a kinematic model of a two-speed differential AGV is constructed. The midpoint C of the line connecting the centers of the left and right drive wheels is selected as the AGV pose reference point. Therefore, the AGV's pose... , The x and y coordinates (horizontal and vertical positions) of the AGV in the global coordinate system. Let be the heading angle of the AGV; and These represent the angular velocities of the left and right drive wheels, as shown in the simplified diagram below. Figure 2 As shown, the kinematic model is:

[0062] ;

[0063] ;

[0064] ;

[0065] in, This is the time derivative of the AGV in its lateral position. This is the time derivative of the AGV in its lateral position. Let be the first derivative of the AGV's heading angle; ω is the linear velocity, and ω is the angular velocity;

[0066] By using discrete-time representation and Euler approximation, the discrete-time model of the kinematic model, i.e., the AGV pose state equation, is obtained as follows:

[0067] ;

[0068] ;

[0069] ;

[0070] in, The heading angle of the AGV. The length between the left and right wheels. Let be the radius of the wheel. The sampling period is denoted as k. Therefore, controlling the wheel speeds of the two drive wheels can change the pose of the AGV, where k is the kth sampling time; x(k), y(k), and θ(k) are the x-coordinate, y-coordinate, and heading angle of the AGV's pose at the kth sampling time, respectively. , These are the velocity and angular velocity at time k, respectively; , , These represent the x-coordinate, y-coordinate, and heading angle of the AGV's pose at the (k+1)th sampling time, respectively.

[0071] The chassis dynamics model, based on d'Alembert's principle, establishes the correlation equations between the forces and motion states of the chassis during longitudinal translation, lateral translation, and yaw rotation, thereby clarifying the overall mechanical response law of the chassis. The drive wheel dynamics model, focusing on factors such as driving torque, rolling resistance torque, and vertical ground load, establishes the dynamic relationship between the drive wheel's angular velocity, angular acceleration, and external forces to characterize the local force-induced motion characteristics of the drive wheel. Together, the chassis dynamics model and the drive wheel dynamics model constitute a dynamic system where force transmission and motion response are coupled, verifying the rationality of the aforementioned kinematic models from a mechanical mechanism perspective. This provides a mechanical foundation that closely reflects actual working conditions for the subsequent construction of the servo motor model and the establishment of a full-link quantitative correlation between motor input and AGV posture.

[0072] In this embodiment, the AGV's front section adopts a driven omnidirectional wheel structure. To further improve the overall vehicle mechanics analysis system, a omnidirectional wheel dynamics model can be constructed to analyze the supporting force, rolling resistance, and lateral friction force acting on the omnidirectional wheel. This ensures the integrity of the overall vehicle force balance analysis and verifies the rationality and feasibility of the simplified assumptions in the kinematic model regarding the omnidirectional wheel's pure rolling motion without lateral sliding from a mechanical perspective. Since the omnidirectional wheel is a driven follower component, its dynamic characteristics have a relatively small impact on the AGV's active motion control. Therefore, when constructing the final pose state equations for control, reasonable simplification can be made based on the comprehensive kinematic geometry and actual force conditions. This ensures that the resulting equations accurately reflect the overall motion state of the AGV with the midpoint of the line connecting the centers of the left and right drive wheels as the reference point, while also possessing good engineering practicality. This provides a reliable model foundation for the subsequent establishment of the servo motor model and the full-link quantitative correlation of motor input to the AGV pose.

[0073] In this embodiment, the servo motor model is a three-loop vector control model of the position, speed and current of a permanent magnet synchronous motor. The voltage, torque and motion equations of the drive motor are established using the dq axis method, and the regulation law of the motor input voltage on the angular velocity of the drive wheel is clarified, providing a motor-side quantitative basis for the establishment of the full-link quantitative correlation.

[0074] Specifically, servo motor models primarily study the control characteristics of servo motors. When constructing a servo motor model, it's necessary to first determine the type of servo motor and its control method. Then, by analyzing the dynamic performance of the servo motor, such as response time and steady-state error, a mathematical model is established. These models describe the relationship between the control inputs (such as voltage and current) and outputs (such as speed and torque) of the servo motor. The core of AGV motion control is controlling the drive motor. The AGV drive system uses a permanent magnet synchronous motor (PMSM) with a three-loop adjustment system, forming a vector control servo system. That is, a position loop is added to the dual closed-loop speed control system. Figure 3 As shown, , These are the reference inputs for position and velocity, respectively. This refers to the rotor angular velocity; Output the position; For speed feedback coefficient; This is the transfer function for the speed reducer.

[0075] When using vector control, permanent magnet synchronous motors typically employ the dq-axis method. In the dq-axis method, the simplified model of the motor is as follows: Figure 4 As shown. Wherein: For motor control voltage, This refers to the rotor angular velocity; This is the equivalent resistance of the winding; This is the equivalent inductance torque coefficient; is the number of pole pairs; is the equivalent flux linkage of the magnetic field rotor; This is the load torque; For rotational inertia, It is a permanent magnet flux chain. This is the back electromotive force coefficient of the motor.

[0076] The mathematical model in the dq coordinate system consists of four parts: equations of motion, torque equations, voltage equations, and flux linkage equations.

[0077] The voltage equation is as follows:

[0078] ;

[0079] In the formula: , These are the voltages along the d-axis and q-axis, respectively. For armature resistance, , These are the currents along the d-axis and q-axis, respectively. , The flux linkages along the d-axis and q-axis are respectively. Electric angular velocity;

[0080] The flux linkage equations are as follows: Magnetic flux linkage equation: Chain equation:

[0081] ;

[0082] In the formula: , The inductances along the d-axis and q-axis are respectively. It is a permanent magnet flux linkage. Magnetic field components are used. With a control strategy where = 0, the torque equation is:

[0083] ;

[0084] In the formula: This represents the number of pole pairs of the motor.

[0085] The equation of motion is:

[0086] ;

[0087] In the formula: This refers to the output torque of the motor. This refers to the load torque; The viscous friction coefficient of the motor; The equivalent moment of inertia includes the motor's own moment of inertia and the load's equivalent moment of inertia; The mechanical angular velocity of the motor is calculated using the following formula:

[0088] ;

[0089] When viscous friction coefficient and servo motor magnetic saturation are ignored, when The transfer function between the speed of the permanent magnet synchronous motor and the input voltage is simplified as follows:

[0090] ;

[0091] In the formula: For the Laplace operator, This is the motor voltage coefficient. is the armature time constant of the motor.

[0092] A Hall current sensor is installed inside the driver, and a first-order low-pass filter is set in the loop. The transfer function of the current loop is:

[0093] ;

[0094] In the formula: The gain coefficient of the regulator; This is the time constant for the low-pass filter.

[0095] The transfer function of the PWM power amplifier circuit is:

[0096] ;

[0097] in: , These are: magnification factor and time constant. The mechanical transmission link is a speed reducer, and its transfer function is...

[0098] ;

[0099] Where: D is the wheel diameter, i represents the transmission ratio of the drive wheel, and s is the complex variable of the Laplace transform.

[0100] Therefore, based on the above, the transfer function of the motor input voltage and the linear displacement output by the AGV is:

[0101] ;

[0102] The above transfer function The quantitative mapping relationship between motor control voltage and AGV linear displacement is constructed through a series servo motor electromechanical model, PWM drive model, current loop model and AGV wheel-body kinematic model. It is the core component of the full-link quantitative association from motor input to AGV posture.

[0103] Based on this, and combining the dynamic relationship between the speed difference between the left and right drive wheels and the heading angle in the AGV kinematic model, the transfer function is extended to the complete dimension of the AGV pose state equation with the midpoint of the line connecting the centers of the left and right drive wheels as the reference point, and finally a full-link quantitative association is formed from the motor input voltage to the AGV position and heading angle full pose state.

[0104] In this embodiment, the end-to-end quantization association from motor input to AGV pose is specifically constructed as a two-level quantization association: from motor input to drive wheel angular velocity, and from drive wheel angular velocity to AGV pose. Through the cascading of these two levels of mapping, a complete motion characteristic mapping from motor control voltage input to AGV's full pose state output is achieved. The AGV's internal model is a digital abstract model of this end-to-end quantization association, whose parameters match the actual dynamic characteristics of the AGV's motion, providing a model basis for subsequent parameter tuning and control algorithm design of the IMC-PID control unit.

[0105] To achieve precise control of AGV motion and posture, an internal AGV model reflecting the actual dynamic characteristics of the AGV is constructed based on the end-to-end quantization correlation from motor input to AGV posture. An IMC-PID control unit is then configured based on this internal model. This control unit combines the internal model control principle with the traditional PID control structure, improving the robustness and trajectory tracking accuracy of the control system by introducing the AGV's internal model.

[0106] The IMC-PID control unit is equipped with a first-order filter. Based on the AGV's internal model, the IMC-PID control unit predicts the AGV's pose change trend under different speed adjustment amounts. It combines proportional adjustment of the deviation magnitude and derivative adjustment of the deviation change rate to calculate the independent speed adjustment amounts of the left and right drive wheels, thus achieving a combination of feedforward prediction and feedback adjustment.

[0107] The AGV internal model in this embodiment is a composite control strategy that integrates the internal model principle and PID (proportional-integral-derivative) control. It constructs an AGV internal model capable of characterizing the system's dynamic characteristics and embeds this model into the IMC-PID control unit to improve the robustness and trajectory tracking performance of the control system. Its typical structure is as follows: Figure 5 As shown in the figure, This is a predetermined value. External interference Output for objects; The controlled object; For internal models; Internal mold controller. Figure 6 For the equivalent internal model control system structure diagram, the closed-loop output of the IMC-PID control unit is expressed as:

[0108] ;

[0109] In the formula:

[0110] ;

[0111] Where T1 and T2 represent process time, and K represents process gain.

[0112] In practical applications, to improve system stability, a first-order filter is introduced to ensure system robustness by adding a filter stage.

[0113] ;

[0114] In the formula: This is the filtering time constant.

[0115] According to the IMC-PID control unit design method, we can obtain:

[0116] ;

[0117] From this formula, we can obtain the feedback controller in the IMC-PID control unit:

[0118] ;

[0119] PID control:

[0120] ;

[0121] In the formula, This is the proportionality coefficient. The integral time constant is... is the differential time constant.

[0122] 2) Collect images of the AGV's travel path, process the images to obtain the centerline of the path, and calculate the deviation information of the AGV's position and angle relative to the centerline.

[0123] In this embodiment, images of the AGV's driving path are acquired by a visual sensor (such as a camera), the color strip features on the path are identified and its center line is extracted, and then the positional deviation and angular deviation of the AGV relative to the center line are calculated.

[0124] In this embodiment, the path image is processed sequentially as follows: weighted average grayscale conversion, median filtering for noise reduction, Otsu's binarization, Canny algorithm edge extraction, and line-by-line scanning fitting. The weighted average method assigns different weights to the red, green, and blue components of the image based on the characteristics of human brightness perception. The line-by-line scanning fitting calculates the midpoint coordinates of the left and right edge points of the path and fits them to obtain a continuous path centerline.

[0125] These steps are used to extract useful information from the original image, enabling accurate target recognition, localization, and control. First, grayscale conversion transforms the color image into a grayscale image, eliminating color information and highlighting the image's brightness features. Next, image denoising eliminates noise interference, improving image clarity and accuracy. Then, image binarization converts the grayscale image into a binary image, containing only black and white. After image binarization, edge extraction becomes easier and more accurate. Finally, centerline extraction is performed based on the edge extraction results.

[0126] The raw images captured by the sensor camera not only show the AGV's movement path but also contain noise. Furthermore, the captured images are in color and cannot be used directly. They need to be processed to successfully obtain guidance information that the AGV can recognize.

[0127] Based on this, the paper uses the weighted average method to perform grayscale processing on the path image. The formula for the weighted average method is as follows:

[0128] ;

[0129] In the formula: i and j represent the row and column of the image, respectively. The red component, Green component, Blue component.

[0130] Based on parameter adjustments of the human brightness perception system, a weighted average is applied to the three components, with different weights. The resulting image after grayscale processing of the path image using the weighted average method is shown below. Figure 7 As shown.

[0131] Effective noise removal is crucial for ensuring image clarity. This embodiment employs median filtering for noise reduction, which not only removes noise but also preserves edge information. Its principle is to take the median value of each point within the image's neighborhood, thus addressing the issue of single noise points affecting the image. Furthermore, it effectively prevents noise interference. Digital images are generally represented using two-dimensional data. If the pixels within an image template are sorted according to their pixel values, a monotonically increasing (or decreasing) two-dimensional data sequence can be obtained. In this case, the output of the two-dimensional median filter can be:

[0132] ;

[0133] In the formula, , These are the original image and the processed image, respectively. For two-dimensional templates, a 3×3 or 5×5 rectangular area is usually selected as the two-dimensional template. Of course, there are also options for different shapes such as lines, circles, crosses, and rings. Figure 8 The effect of processing the grayscale image of the guidance path using median filtering is demonstrated.

[0134] After noise removal, the image underwent binarization processing using the Otsu method to determine the binarization segmentation threshold. This threshold was defined as follows:

[0135] ;

[0136] In the formula: For inter-class variance, , P1 and P2 are the mean values ​​of the foreground and background colors, respectively, and represent the proportions of foreground and background color pixels to the total number of pixels. The grayscale mean is shown in the image. The binarized image after processing using the Otsu method is shown below. Figure 9 As shown.

[0137] In image processing, the Canny algorithm first performs smoothing operations by applying a quasi-Gaussian function, and then finds the maximum value of the derivative in the directed first-order derivative. The Canny edge detection method achieves a good balance between edge detection and noise suppression by performing a first-order derivative of the Gaussian function.

[0138] The image is smoothed using a Gaussian function, then gradients are calculated using the finite difference method, followed by non-maximum suppression to process the gradients. Finally, a double thresholding algorithm is used to detect and connect edges. The resulting image edge extraction diagram is shown below. Figure 10 As shown.

[0139] To ensure accurate navigation of the AGV, the centerline of the guidance path must be obtained. Therefore, straight line fitting is needed to fit the acquired path edge lines to determine the positions of all center points. First, a line-by-line scan of the path edge is performed to obtain the coordinates of each pixel (the grayscale value of points on the edge line is 255). Let the coordinates of the left edge point be... The coordinates of the right edge point are The coordinates of the center point are Then, the grayscale value of the obtained center point is set to 255, and the grayscale values ​​of other pixels are set to 0. This extracts the centerline of the guide path. The centerline is curved, and after fitting using the least squares method, a straight line equation is obtained: Image centerline extraction result diagram as shown below Figure 11 As shown.

[0140] The deviation information in position and angle is calculated by comparing the geometric position of the AGV's own positioning reference with the path centerline, and by comparing the AGV's traveling direction with the extension direction of the path centerline.

[0141] Using the AGV's own positioning reference as a guide, the geometric position of the AGV's own positioning reference is compared with the path centerline to obtain the offset of the AGV's actual position from the path centerline, which is the position deviation. At the same time, the current travel direction of the AGV is compared with the extension direction of the path centerline, and the angle between the two is calculated, which is the angle deviation. Through the above two comparisons and calculations, the position deviation and angle deviation are obtained respectively, forming complete deviation information.

[0142] 3) Input the deviation information into the IMC-PID control unit. The IMC-PID control unit calculates the speed adjustment of the left and right drive wheels of the AGV in real time based on the magnitude and rate of change of the deviation.

[0143] Specifically, step 3) uses the positional and angular deviations of the AGV relative to the path centerline obtained in step 2) as the core inputs. Combined with the AGV internal model, full-link quantization association, and IMC-PID control unit structure design constructed above, the speed adjustment is calculated in real time in stages to ensure that the calculation results closely match the actual motion characteristics of the AGV. The specific implementation process is as follows:

[0144] The deviation information is input to the IMC-PID control unit. This control unit first predicts the AGV's motion trend under the current deviation based on the embedded AGV internal model and the previously established full-link quantization correlation between motor input and AGV posture. Specifically, it predicts the rate of change of the AGV's distance from the path centerline based on the position deviation, and predicts the trend of change of the AGV's heading angle based on the angle deviation. This allows it to predict the increase in posture deviation that will occur if the AGV does not adjust its speed, thus achieving feedforward predictive control and proactively mitigating the risk of deviation amplification.

[0145] Building upon this, the IMC-PID control unit, combined with its built-in first-order filter, further filters the deviation signal to eliminate high-frequency interference, ensuring the stability of the control unit's output and preventing frequent adjustments to the drive wheel speed due to minor fluctuations in the deviation signal, which could affect the smoothness of the AGV's operation. Simultaneously, the control unit weights and fuses position and angle deviations according to a preset control strategy. The weighting coefficients are set based on the actual operating conditions of the AGV, prioritizing rapid correction of heading angle deviations before gradually eliminating position deviations. This ensures the AGV corrects its heading before adjusting its position, preventing further trajectory deviations.

[0146] Subsequently, the IMC-PID control unit, based on the fused deviation signal and combined with the PID control formula derived earlier, calculates in real time the proportional adjustment of the deviation magnitude, the integral adjustment of the deviation integral, and the derivative adjustment of the deviation change rate. The proportional coefficient, integral time constant, and derivative time constant are all tuned based on the AGV's internal model, maintaining consistency with the parameters associated with the servo motor model and the full-link quantization, ensuring that the control parameters match the actual dynamic characteristics of the AGV and balancing the system's response speed and stability.

[0147] The IMC-PID control unit comprehensively calculates the proportional, integral, and derivative control values, outputting independent speed adjustments for the left and right drive wheels, including the speed adjustments for the left and right drive wheels. The specific calculation process combines the discrete-time equations of the AGV's kinematic model, determining the speed difference between the left and right drive wheels based on the angular deviation, and determining the reference speed adjustment amplitude for the left and right drive wheels based on the positional deviation. This ensures that the speed adjustment accurately corresponds to the correction requirements for the posture deviation: when the AGV has a positive positional deviation, the speed adjustment of the left and right drive wheels is increased simultaneously, while the speed difference between the left and right wheels is adjusted according to the heading angle deviation, causing the AGV to move closer to the path centerline; as the deviation gradually decreases to within a preset threshold range, the speed adjustment is gradually decreased to avoid overshoot.

[0148] The IMC-PID control unit limits the calculated speed adjustment of the left and right drive wheels to ensure that the adjustment does not exceed the maximum speed range of the servo motor. At the same time, it matches the control input requirements of the servo motor model mentioned above and converts the speed adjustment into the corresponding motor control voltage adjustment signal, providing precise control commands for the subsequent speed control of the drive motor.

[0149] 4) The speed adjustment is transmitted to the AGV's drive system. The AGV's posture is dynamically corrected by adjusting the differential speed of the left and right drive wheels. At the same time, the real-time posture of the AGV and the real-time speed of the left and right drive wheels are collected as feedback signals. The feedback signals are input to the IMC-PID control unit to form a closed-loop control.

[0150] Specifically, step 4), as the execution and closed-loop feedback link of the entire AGV joint control method, follows the independent speed adjustment of the left and right drive wheels calculated in step 3), realizing the coordinated operation of dynamic posture correction and closed-loop control, ensuring that the AGV travels stably along the path centerline. The specific implementation process is explained as follows:

[0151] Adjusting the differential speed of the left and right drive wheels is the core execution method for achieving dynamic posture correction of the AGV. Specifically, it is based on the AGV's rear dual drive wheels, front omnidirectional wheel structure, and two-speed differential motion characteristics, combined with the independent speed adjustment output in step 3). According to the independent speed adjustment of the left and right drive wheels, the speeds of the two drive wheels are adjusted independently and non-uniformly. For example, the left drive wheel adjusts its speed according to a base speed plus a corresponding independent speed adjustment, and the right drive wheel adjusts its speed according to the same pattern. The AGV's heading angle is corrected through the speed difference between the two drive wheels, and the AGV's position offset is corrected by the synchronous increase and decrease of the speeds of the two drive wheels, thus completing the coordinated dynamic posture correction.

[0152] The reference speed is determined based on the AGV's preset travel speed and is consistent with the speed control logic in the servo motor model mentioned earlier. This ensures that the speed adjustment process matches the dynamic characteristics of the motor and avoids instability caused by sudden speed changes. The heading angle correction process matches the quantitative correlation between the speed difference of the left and right drive wheels and the yaw motion in the AGV's kinematic model. The larger the angle deviation, the larger the speed difference is set, ensuring that the heading angle is quickly corrected to the preset range. The position offset correction adjusts the overall travel speed of the AGV by synchronously increasing or decreasing the speed of the drive wheels on both sides, allowing the AGV to quickly approach the center line of the path. At the same time, it works in conjunction with the heading angle correction to achieve coordinated correction of position and heading, avoiding trajectory deviation caused by single correction.

[0153] The closed-loop control formed by the feedback signal is a dual-feedback closed-loop control, and its specific implementation process is as follows:

[0154] The feedback signals consist of two core types. One type is the AGV's real-time pose signal, calculated in real-time using path images acquired by a vision sensor and the aforementioned deviation calculation method. This signal includes the AGV's real-time position and heading angle information. This signal is used to correct the AGV's trajectory deviation by comparing the real-time pose with the target pose (the pose corresponding to the path centerline) and updating the deviation information, providing the latest deviation basis for the IMC-PID control unit's subsequent speed adjustment calculation. The other type is the real-time speed signal of the left and right drive wheels, acquired by the servo motor's built-in speed sensor, corresponding to the motor's mechanical angular velocity in the servo motor model mentioned earlier. This signal is used to correct the drive system's execution deviation by monitoring in real-time whether the actual speed of the drive wheels matches the target speed corresponding to the speed adjustment. If a deviation exists, it is fed back to the IMC-PID control unit for real-time correction of the speed adjustment, ensuring the drive system's execution accuracy. To ensure the real-time performance and synchronization of the closed-loop control, the signal acquisition frequency of the dual-feedback closed-loop control is matched with the calculation frequency of the IMC-PID control unit, achieving real-time deviation correction.

[0155] The entire closed-loop control process forms a complete control link: the IMC-PID control unit outputs the speed adjustment amount → the drive system executes the speed adjustment → real-time posture and drive wheel speed feedback are collected → the control unit corrects the speed adjustment amount based on the feedback, and so on, until the position deviation and angle deviation of the AGV are stable within the preset allowable error range, so as to realize the precise trajectory tracking and stable driving of the AGV.

[0156] It should be noted that the design of the dual feedback closed-loop control strictly follows the full-link quantization association from motor input to AGV posture constructed above. The real-time posture feedback of AGV corresponds to the output end (posture state) of the entire link, and the real-time speed feedback of the drive wheel corresponds to the intermediate link (drive wheel speed) of the entire link. Through feedback at both ends, the precise control of the entire link is achieved, which ensures both posture tracking accuracy and the execution stability of the drive system.

[0157] Based on this, this embodiment also proposes an AGV joint control device to implement the above-mentioned AGV joint control method, including:

[0158] The model building module is used to construct the AGV pose state equation with the midpoint of the line connecting the centers of the left and right drive wheels as the reference point, construct the servo motor model for the AGV drive motor, establish the full-link quantization association from motor input to AGV pose based on the AGV pose state equation and servo motor model, and construct the AGV internal model and configure the IMC-PID control unit according to the full-link quantization association.

[0159] The vision processing module is used to acquire images of the AGV's travel path, process the images to obtain the centerline of the path, and calculate the deviation information of the AGV's position and angle relative to the centerline.

[0160] The control calculation module, connected to the vision processing module and the model building module, is used to receive deviation information and transmit it to the IMC-PID control unit. The IMC-PID control unit calculates the speed adjustment of the left and right drive wheels of the AGV in real time according to the magnitude and rate of change of the deviation.

[0161] The drive correction module, connected to the control calculation module, is used to receive speed adjustment quantities and transmit them to the AGV's drive system. It dynamically corrects the AGV's posture by adjusting the differential speed of the left and right drive wheels. At the same time, it collects the real-time posture of the AGV and the real-time speed of the left and right drive wheels as feedback signals and transmits them to the IMC-PID control unit to form a closed-loop control.

[0162] In this embodiment of the apparatus, the model building module is used to execute step S1 of the above-described AGV joint control method, the vision processing module is used to execute step S2 of the above method, the control calculation module is used to execute step S3 of the above method, and the drive correction module is used to execute step S4 of the method.

[0163] To further verify the effectiveness and superiority of the proposed internal model PID and vision technology combined control method, a corresponding AGV trajectory tracking simulation model was constructed on the Matlab simulation platform. In the simulation experiment, a step signal was used as the system input command; the traditional PID controller parameters were tuned as follows: proportional coefficient K... P =1, Differential time constant K d =0.5, integral time constant Ki = 0.2; for the IMC-PID control unit, a first-order filter is configured in the simulation to improve the system robustness, filter constant... Simultaneously, set the initial value of the IMC-PID control unit. , , Weighted adjustment parameters for weighted fusion of positional and angular deviations. .

[0164] The simulation experiment was conducted based on a path centerline localization scheme using visual recognition, with the path centerline accurately extracted through a visual processing module. Using the path centerline identified by the visual processing module, tracking simulations were performed for comparison with no interference and interference levels of 0.5, 1.0, 2.0, 3.0, and 5.0.

[0165] like Figure 12 As shown, under undisturbed operating conditions, traditional PID control systems suffer from slow response and large overshoot; however, when using IMC-PID control, the system response is significantly faster, and the overshoot is effectively controlled. (Comparison) Figure 13 and Figure 12 , Figure 14and Figure 12 It is evident that under conditions of external interference, the IMC-PID control scheme has stronger anti-interference capability, faster system response, and smaller overall tracking error compared to the traditional PID.

[0166] against Figures 17-22 The simulation data for the large disturbance conditions shown (disturbance intensities of 2, 3, and 5) demonstrate that under these conditions, the traditional PID control system takes approximately 8-10 seconds to gradually approach the desired value; while the IMC-PID control scheme achieves the desired value quickly in only 3-6 seconds, with overshoot controlled to a minimal range throughout the transition process, enabling faster and smoother dynamic recovery to the desired trajectory. Further comparison of the simulation results between vision technology and the pure IMC-PID technology shows that the combined vision navigation and IMC-PID control scheme exhibits the best performance in tracking accuracy, response speed, and operational stability, effectively enabling AGVs to achieve high-precision real-time tracking of complex paths.

[0167] The comprehensive simulation results lead to the conclusion that the IMC-PID control method adopted in this application outperforms traditional PID control. The reason for this is that internal model control, by introducing an internal model that accurately reflects the dynamic characteristics of the system, can effectively predict future system behavior trends and make control decisions in advance, thereby improving the system's control accuracy and response speed. Simulation results show that internal model PID control achieves fast response while maintaining a small steady-state error, and the system output can more accurately follow changes in the setpoint. Furthermore, the internal model control structure has strong robustness to changes in system parameters and external disturbances. In simulation verifications introducing different types of disturbances and parameter changes, it can better maintain the overall stability of the system, ensuring the trajectory tracking accuracy and operational reliability of the AGV under complex working conditions.

[0168] The present invention also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist alone and not assembled into the electronic device.

[0169] The aforementioned computer-readable medium carries one or more programs that, when executed by an electronic device, cause the electronic device to implement the methods described in the above embodiments.

[0170] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0171] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the method according to the embodiments of this disclosure.

[0172] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein.

[0173] The above are merely specific embodiments of the present invention, but the design concept of the present invention is not limited thereto. Any non-substantial modifications made to the present invention using this concept shall be considered as infringing upon the protection scope of the present invention.

Claims

1. A method for joint control of AGVs, characterized in that, Includes the following steps: For AGV, construct the AGV pose state equation with the midpoint of the line connecting the centers of the left and right drive wheels as the reference point; construct a servo motor model for the AGV drive motor; establish a full-link quantization association from motor input to AGV pose based on the AGV pose state equation and the servo motor model; and construct the AGV internal model and configure the IMC-PID control unit based on the full-link quantization association. Images of the AGV's travel path are acquired, the images are processed to obtain the centerline of the path, and the deviation information of the AGV's position and angle relative to the centerline is calculated. The deviation information is input into the IMC-PID control unit, which calculates the speed adjustment of the left and right drive wheels of the AGV in real time based on the magnitude and rate of change of the deviation. The speed adjustment is transmitted to the AGV's drive system, and the AGV's posture is dynamically corrected by adjusting the differential speed of the left and right drive wheels. At the same time, the real-time posture of the AGV and the real-time speed of the left and right drive wheels are collected as feedback signals, and the feedback signals are input to the IMC-PID control unit to form a closed-loop control.

2. The AGV joint control method as described in claim 1, characterized in that, The AGV is a two-speed differential structure. The AGV pose state equation is obtained by discretizing the AGV kinematic model using discrete-time representation combined with Euler approximation. The pose state equation includes the quantitative correlation between the AGV heading angle, planar displacement and the angular velocities of the left and right drive wheels.

3. The AGV joint control method as described in claim 1, characterized in that, The servo motor model is a three-loop vector control model of position, speed and current of permanent magnet synchronous motor. The voltage, torque and motion equations of the drive motor are established using the dq axis method, and the regulation law of motor input voltage on drive wheel angular velocity is clarified, providing a motor-side quantitative basis for the establishment of full-link quantitative correlation.

4. The AGV joint control method according to claim 1, characterized in that, The full-link quantization association is a two-level quantization association from motor input to drive wheel angular velocity and from drive wheel angular velocity to AGV pose, realizing the full-link motion characteristic mapping from motor input to AGV pose. The AGV internal model is a digital abstract model of this full-link quantization association, matching the actual motion dynamic characteristics of the AGV.

5. The AGV joint control method according to claim 1, characterized in that, The processing of the path image is as follows: weighted average grayscale conversion, median filtering for noise reduction, Otsu's binarization, Canny algorithm edge extraction, and line-by-line scanning fitting. The weighted average method assigns different weights to the red, green, and blue components of the image according to the characteristics of human brightness perception. The line-by-line scanning fitting calculates the midpoint coordinates of the left and right edge points of the path and fits them to obtain a continuous path centerline.

6. The AGV joint control method according to claim 1, characterized in that, The deviation information of position and angle is calculated by comparing the geometric position of the AGV's own positioning reference with the center line of the path and the AGV's driving direction with the extension direction of the center line of the path, respectively, providing a quantitative basis for position and posture correction.

7. The AGV joint control method according to claim 1, characterized in that, The IMC-PID control unit is equipped with a first-order filter. Based on the AGV's internal model, the IMC-PID control unit predicts the AGV's pose change trend under different speed adjustment amounts. It combines proportional adjustment of the deviation magnitude and derivative adjustment of the deviation change rate to calculate the independent speed adjustment amounts of the left and right drive wheels, thereby achieving a combination of feedforward prediction and feedback adjustment.

8. The AGV joint control method according to claim 1, characterized in that, The adjustment of the differential speed of the left and right drive wheels is specifically as follows: based on the independent speed adjustment of the left and right drive wheels, the independent speed of the two drive wheels is adjusted in a non-uniform manner. The heading angle of the AGV is corrected by the speed difference of the two drive wheels, and the position offset of the AGV is corrected by the synchronous increase and decrease of the speed of the two drive wheels, thus completing the coordinated dynamic correction of the posture.

9. The AGV joint control method according to claim 1, characterized in that, The closed-loop control formed by the feedback signal is a dual-feedback closed-loop control. The real-time pose signal of the AGV is used to correct the trajectory deviation of the AGV, and the real-time speed signals of the left and right drive wheels are used to correct the execution deviation of the drive system. The signal acquisition frequency of the dual-feedback closed-loop control is matched with the calculation frequency of the IMC-PID control unit to realize the real-time correction of the deviation.

10. An AGV joint control device, characterized in that, include: The model building module is used to build the AGV pose state equation with the midpoint of the line connecting the centers of the left and right drive wheels as the reference point, build a servo motor model for the AGV drive motor, establish a full-link quantization association from the motor input to the AGV pose based on the AGV pose state equation and the servo motor model, and build the AGV internal model and configure the IMC-PID control unit according to the full-link quantization association. The vision processing module is used to acquire images of the AGV's travel path, process the images to obtain the centerline of the path, and calculate the deviation information of the AGV's position and angle relative to the centerline. The control calculation module, connected to the vision processing module and the model building module, is used to receive the deviation information and transmit it to the IMC-PID control unit. The IMC-PID control unit calculates the speed adjustment of the left and right drive wheels of the AGV in real time according to the magnitude and rate of change of the deviation. The drive correction module, connected to the control calculation module, is used to receive the speed adjustment amount and transmit it to the AGV's drive system. It dynamically corrects the AGV's posture by adjusting the differential speed of the left and right drive wheels, and at the same time, it collects the real-time posture of the AGV and the real-time speed of the left and right drive wheels as feedback signals and transmits them to the IMC-PID control unit to form a closed-loop control.

Images