A motion trajectory tracking control method and system for AGV warehouse robot

Abstract

The present invention relates to the technical field of warehouse robots, and in particular to a motion trajectory tracking control method and system for an AGV warehouse robot. The predicted dynamic characteristic data of the target AGV warehouse robot within a preset time period is obtained according to the real-time load mass and the real-time center of gravity position and in combination with a motion state prediction model; the operating state of the target AGV warehouse robot within the preset time period is determined and analyzed to obtain the operating state of the target AGV warehouse robot within the preset time period; if the operating state of the target AGV warehouse robot within the preset time period is a trajectory out-of-control state, a correction control scheme is generated, and correction control processing is performed on the target AGV warehouse robot based on the correction control scheme. The present invention can effectively improve the trajectory tracking control accuracy and operation safety of the AGV in a complex operating environment by real-time evaluation and adjustment of the trajectory state of the AGV, and ensure that the AGV can stably and efficiently complete the warehousing operation tasks.

Description

Technical Field

[0001] The present invention relates to the technical field of warehouse robots, and in particular to a motion trajectory tracking control method and system for an AGV warehouse robot. Background Art

[0002] As an important material handling equipment, AGV warehouse robots are widely used in automated warehousing, smart factories and other fields. AGV trajectory tracking control is one of its core technologies, and its goal is to enable AGV to accurately travel along the predetermined path and accurately reach the target location.

[0003] At present, certain research results have been achieved in trajectory tracking control methods for AGV warehouse robots, but most methods are mainly designed for empty or constant load situations. However, in actual warehouse operations, intelligent AGV warehouse robots often face situations where the load size and center of gravity position are constantly changing. When the load mass and center of gravity position of the AGV warehouse robot change during operation, its kinematic and dynamic models will change accordingly, resulting in a decrease in the performance of the trajectory tracking control algorithm designed based on fixed parameters, which is specifically manifested in the reduction of trajectory tracking accuracy and positioning accuracy. In severe cases, it may even destroy the stability of the AGV warehouse robot's motion. In order to ensure the stability of the motion posture of the AGV warehouse robot under load changes, it is necessary to make corresponding adjustments to the motion control and error compensation.

[0004] Existing studies have shown that changes in load mass have a significant impact on the motion trajectory of AGV warehouse robots. As the load mass increases, the over-steering tendency of the AGV warehouse robot gradually increases; when the load mass increases to a certain extent, the over-steering tendency will first weaken and then strengthen, and tend to stabilize when approaching the maximum load. In addition, the position change of the load in the forward direction of the AGV warehouse robot in the body coordinate system has little effect on the steering trend, while the change in the lateral position of the load has a significant effect on the steering trend. As the lateral position increases, the over-steering tendency gradually weakens; but when the lateral position increases to a certain value, the degree of weakening of the over-steering tendency gradually decreases.

[0005] In summary, the existing AGV warehouse robot trajectory tracking control method rarely considers the impact of load changes, and it is difficult to ensure the trajectory tracking accuracy and motion stability of the AGV warehouse robot under load change conditions. Therefore, it is of great practical significance to develop an AGV warehouse robot trajectory tracking control method and system that can adapt to load changes. Summary of the invention

[0006] The present invention overcomes the problem in the prior art that load changes affect the motion trajectory accuracy of an AGV storage robot, and provides a motion trajectory tracking control method and system for an AGV storage robot.

[0007] To achieve the above-mentioned purpose, the technical solution adopted by the present invention is:

[0008] The first aspect of the present invention discloses a motion trajectory tracking control method for an AGV warehouse robot, comprising the following steps:

[0009] Acquire historical dynamic feature data of the target AGV warehouse robot in the future under working conditions of different load masses and center of gravity positions, and construct a motion state prediction model of the target AGV warehouse robot according to the historical dynamic feature data;

[0010] Acquire the real-time load mass and real-time center of gravity position of the target AGV warehouse robot at a preset time node, and acquire the predicted dynamic feature data of the target AGV warehouse robot within a preset time period based on the real-time load mass and real-time center of gravity position and in combination with the motion state prediction model;

[0011] According to the predicted dynamic feature data and in combination with the three-dimensional grid coordinate system, the operating state of the target AGV warehouse robot within the preset time period is determined and analyzed to obtain the operating state of the target AGV warehouse robot within the preset time period; wherein the operating state includes a trajectory out of control state and a trajectory stable state;

[0012] If the operating state of the target AGV warehouse robot within the preset time period is a trajectory out-of-control state, a correction control scheme is generated, and correction control processing is performed on the target AGV warehouse robot based on the correction control scheme.

[0013] Preferably, the historical dynamic feature data of the target AGV warehouse robot in the future under the working conditions of different load masses and center of gravity positions are obtained, and the motion state prediction model of the target AGV warehouse robot is constructed according to the historical dynamic feature data, specifically:

[0014] Introduce the long short-term memory network and initialize the input layer, hidden layer and output layer of the long short-term memory network; import the load mass and center of gravity position as input features into the input layer, and use the historical dynamic feature data in the future time as the output target of the output layer;

[0015] By using the time series processing capability of the long short-term memory network, the correlation between input features and output targets is automatically learned, and the network parameters are optimized through the back-propagation algorithm to obtain the weight matrix from the input layer to the hidden layer.

[0016] The weight matrix from the input layer to the hidden layer is normalized based on the correlation between the input features and the output targets, and the normalized weight matrix is ​​transposed to obtain a transposed matrix of the normalized weight matrix;

[0017] The normalized weight matrix is ​​multiplied by its transposed matrix to construct the final correlation matrix, thereby quantitatively expressing the correlation between the input features and the output targets.

[0018] Construct a prediction model, and import the association matrix into the prediction model for coding training until the model prediction accuracy meets the requirements, save the final training parameters of the model, and obtain the motion state prediction model of the target AGV warehouse robot;

[0019] The historical dynamic feature data includes the historical motion speed, historical angular velocity, historical acceleration and historical trajectory nodes of the target AGV storage robot at each time stamp in the future.

[0020] Preferably, the real-time load mass and real-time center of gravity position of the target AGV warehouse robot are obtained at a preset time node, and the predicted dynamic feature data of the target AGV warehouse robot within a preset time period is obtained based on the real-time load mass and real-time center of gravity position and combined with the motion state prediction model, specifically:

[0021] During the operation of the target AGV warehouse robot, the real-time load mass and real-time center of gravity position of the target AGV warehouse robot are obtained at a preset time node;

[0022] Importing the real-time load mass and real-time center of gravity position of the target AGV warehouse robot into the motion state prediction model for prediction;

[0023] Through prediction, the predicted dynamic feature data of the target AGV storage robot within a preset time period is obtained; the predicted dynamic feature data includes the predicted motion speed, predicted angular velocity, predicted acceleration and predicted trajectory nodes of the target AGV storage machine at each time stamp within the preset time period.

[0024] Preferably, the operating state of the target AGV warehouse robot within a preset time period is determined and analyzed based on the predicted dynamic feature data and combined with the three-dimensional grid coordinate system to obtain the operating state of the target AGV warehouse robot within the preset time period, specifically:

[0025] With the X-axis coordinate as the operating time dimension of the target AGV warehouse robot, the Y-axis coordinate as the horizontal position dimension of the target AGV warehouse robot on the operating plane, and the Z-axis coordinate as the vertical position dimension of the target AGV warehouse robot on the operating plane, a three-dimensional grid coordinate system is constructed, and the grid unit size of the three-dimensional grid coordinate system is initialized;

[0026] Obtain a preset operation task of the target AGV warehouse robot, obtain preset trajectory nodes of the target AGV warehouse robot at each time stamp within a preset time period in the preset operation task, and generate a preset operation trajectory curve of the target AGV warehouse robot within the preset time period in the three-dimensional grid coordinate system according to the preset trajectory nodes at each time stamp;

[0027] Obtaining predicted trajectory nodes of the target AGV warehouse robot at each time stamp within a preset time period from the predicted dynamic feature data, and generating a predicted running trajectory curve of the target AGV warehouse robot within the preset time period in the three-dimensional grid coordinate system according to the predicted trajectory nodes at each time stamp;

[0028] Performing search and analysis on each grid unit in the three-dimensional grid coordinate system;

[0029] If a certain grid unit contains both a predicted running trajectory curve and a preset running trajectory curve, the grid unit is marked as a type of grid unit;

[0030] If only a predicted running trajectory curve or only a preset running trajectory curve exists in a certain grid unit, the grid unit is marked as a second-class grid unit;

[0031] If neither the predicted running trajectory curve nor the preset running trajectory curve exists in a certain grid unit, the grid unit is marked as a third type of grid unit.

[0032] Preferably, the operating state of the target AGV warehouse robot within a preset time period is determined and analyzed according to the predicted dynamic feature data and combined with the three-dimensional grid coordinate system to obtain the operating state of the target AGV warehouse robot within the preset time period, and the following steps are also included:

[0033] Count the total number of grid cells marked as type one and count the total number of grid cells marked as type two;

[0034] The total number of grid cells marked as the second category is divided by the total number of grid cells marked as the first category to obtain the trajectory error ratio of the target AGV warehouse robot within a preset time period;

[0035] Compare the trajectory error ratio of the target AGV warehouse robot within a preset time period with a preset ratio threshold;

[0036] When the trajectory error ratio of the target AGV warehouse robot within the preset time period is greater than the preset ratio threshold, the operation state of the target AGV warehouse robot within the preset time period is determined as a trajectory out-of-control state;

[0037] When the trajectory error ratio of the target AGV warehouse robot within the preset time period is not greater than the preset ratio threshold, the operating state of the target AGV warehouse robot within the preset time period is determined to be a trajectory stable state.

[0038] Preferably, if the operating state of the target AGV warehouse robot within the preset time period is a trajectory stable state, the current control strategy of the target AGV warehouse robot is maintained unchanged, and the motion trajectory of the target AGV warehouse robot continues to be tracked and monitored at the next preset time node.

[0039] Preferably, if the operating state of the target AGV warehouse robot within the preset time period is a trajectory out-of-control state, a correction control scheme is generated, and correction control processing is performed on the target AGV warehouse robot based on the correction control scheme, specifically:

[0040] If the operating state of the target AGV warehouse robot within the preset time period is a trajectory out-of-control state, then the X-coordinate nodes corresponding to each grid unit marked as the second type are obtained, and the X-coordinate nodes corresponding to each grid unit marked as the second type are determined as the trajectory out-of-control time nodes of the target AGV warehouse robot;

[0041] At the same time, the predicted dynamic feature data of the target AGV warehouse robot within a preset time period is obtained, and the predicted motion control parameters corresponding to each trajectory out-of-control time node of the target AGV warehouse robot are extracted from the predicted dynamic feature data; wherein the predicted motion control parameters include predicted motion speed, predicted angular velocity and predicted acceleration;

[0042] Obtain a preset control scheme of the target AGV warehouse robot, and obtain preset motion control parameters corresponding to each trajectory out-of-control time node of the target AGV warehouse robot in the preset control scheme; wherein the preset motion control parameters include a preset motion speed, a preset angular velocity, and a preset acceleration;

[0043] Calculate the difference between the same predicted motion control parameter and the preset motion control parameter of the target AGV warehouse robot at each trajectory out-of-control time node, and obtain the drift amplitude of each motion control parameter of the target AGV warehouse robot at each trajectory out-of-control time node;

[0044] Generate a deviation correction control scheme for the target AGV warehouse robot at each trajectory out-of-control time node according to the drift amplitude of each motion control parameter of the target AGV warehouse robot at each trajectory out-of-control time node;

[0045] At the corresponding trajectory out-of-control time node, the target AGV warehouse robot is corrected and regulated based on the corresponding correction control scheme.

[0046] The second aspect of the present invention discloses an AGV warehouse robot motion trajectory tracking and control system, the AGV warehouse robot motion trajectory tracking and control system includes a memory and a processor, the memory stores an AGV warehouse robot motion trajectory tracking control method program, when the AGV warehouse robot motion trajectory tracking control method program is executed by the processor, any one of the steps of the AGV warehouse robot motion trajectory tracking control method is implemented.

[0047] The third aspect of the present invention discloses a computer-readable storage medium, which includes an AGV warehouse robot motion trajectory tracking control method program. When the AGV warehouse robot motion trajectory tracking control method program is executed by a processor, any one of the steps of the AGV warehouse robot motion trajectory tracking control method is implemented.

[0048] The present invention solves the technical defects existing in the background technology, and has the following beneficial effects: obtaining the historical dynamic feature data of the target AGV warehouse robot in the future under the working conditions of different load masses and center of gravity positions, and constructing the motion state prediction model of the target AGV warehouse robot according to the historical dynamic feature data; obtaining the real-time load mass and real-time center of gravity position of the target AGV warehouse robot at a preset time node, and obtaining the predicted dynamic feature data of the target AGV warehouse robot in a preset time period according to the real-time load mass and real-time center of gravity position and in combination with the motion state prediction model; judging and analyzing the working state of the target AGV warehouse robot in a preset time period according to the predicted dynamic feature data and in combination with the three-dimensional grid coordinate system, and obtaining the working state of the target AGV warehouse robot in a preset time period; if the working state of the target AGV warehouse robot in the preset time period is a trajectory out-of-control state, then generating a correction control scheme, and performing correction control processing on the target AGV warehouse robot based on the correction control scheme. The present invention can effectively improve the trajectory tracking control accuracy and operation safety of the AGV in a complex operating environment by real-time evaluation and adjustment of the trajectory state of the AGV, and ensure that the AGV can stably and efficiently complete the warehousing operation task. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, drawings of other embodiments can be obtained based on these drawings without paying creative work.

[0050] Figure 1 A first method flow chart of a motion trajectory tracking control method for an AGV warehouse robot;

[0051] Figure 2 A second method flow chart of a motion trajectory tracking control method for an AGV warehouse robot;

[0052] Figure 3 This is a system block diagram of an AGV warehouse robot motion trajectory tracking control system. DETAILED DESCRIPTION

[0053] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0054] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0055] like Figure 1 As shown, the first aspect of the present invention discloses a motion trajectory tracking control method for an AGV warehouse robot, comprising the following steps:

[0056] S102, obtaining historical dynamic feature data of the target AGV warehouse robot in the future under working conditions of different load masses and center of gravity positions, and constructing a motion state prediction model of the target AGV warehouse robot according to the historical dynamic feature data;

[0057] S104, obtaining the real-time load mass and real-time center of gravity position of the target AGV warehouse robot at a preset time node, and obtaining the predicted dynamic feature data of the target AGV warehouse robot within a preset time period according to the real-time load mass and real-time center of gravity position and in combination with a motion state prediction model;

[0058] S106, determining and analyzing the operating state of the target AGV warehouse robot within a preset time period according to the predicted dynamic feature data and in combination with the three-dimensional grid coordinate system, and obtaining the operating state of the target AGV warehouse robot within the preset time period; wherein the operating state includes a trajectory out-of-control state and a trajectory stable state;

[0059] S108, if the operating state of the target AGV warehouse robot within the preset time period is a trajectory stable state, the current control strategy of the target AGV warehouse robot is maintained unchanged, and the motion trajectory of the target AGV warehouse robot is continuously tracked and monitored at the next preset time node;

[0060] S110: If the operating state of the target AGV warehouse robot within a preset time period is a trajectory out-of-control state, a correction control scheme is generated, and correction control processing is performed on the target AGV warehouse robot based on the correction control scheme.

[0061] It should be noted that the motion state prediction model of AGV is constructed by using the historical dynamic characteristic data under different load masses and center of gravity positions, and the model can predict the motion state of AGV in the future time period. At the preset time node, the real-time load mass and center of gravity position of AGV are obtained, and combined with the prediction model, the dynamic characteristic data of AGV in the future time period is predicted. Then, the operating state of AGV in the preset time period is judged and analyzed to identify the track out-of-control state and track stable state. If the AGV is in the track stable state, the current control strategy is maintained unchanged; if it is in the track out-of-control state, a correction control scheme is generated, and the AGV is corrected based on the scheme. Control processing. The present invention can effectively improve the trajectory tracking accuracy and operation safety of AGV in complex operating environments by real-time evaluation and adjustment of the trajectory state of AGV, ensuring that AGV can stably and efficiently complete warehousing operations.

[0062] Preferably, the historical dynamic feature data of the target AGV warehouse robot in the future under the working conditions of different load masses and center of gravity positions are obtained, and the motion state prediction model of the target AGV warehouse robot is constructed according to the historical dynamic feature data, specifically:

[0063] Obtain the service log of the target AGV warehouse robot, and obtain the historical dynamic feature data of the target AGV warehouse robot in the future under the working conditions of different load masses and center of gravity positions according to the service log; wherein the historical dynamic feature data includes the historical motion speed, historical angular velocity, historical acceleration and historical trajectory nodes of the target AGV warehouse robot at each time stamp in the future;

[0064] Introduce the long short-term memory network and initialize the input layer, hidden layer and output layer of the long short-term memory network; Import the load mass and center of gravity position as input features into the input layer, and use the historical dynamic feature data in the future as the output target of the output layer;

[0065] By using the time series processing capability of the long short-term memory network, the correlation between input features and output targets is automatically learned, and the network parameters are optimized through the back-propagation algorithm to obtain the weight matrix from the input layer to the hidden layer.

[0066] Among them, the long short-term memory network (LSTM) structure is first initialized, including setting the input layer, hidden layer and output layer, and then the input feature sequence such as load mass and center of gravity position is input into the network, and the corresponding future dynamic feature data is used as the output target sequence. In the network training stage, the time series correlation between the input feature sequence and the output target sequence is automatically learned through the internal mechanism of LSTM to capture the time dependency. At the same time, the back propagation algorithm is used to optimize the network parameters and adjust the weight connection inside the hidden layer to minimize the error between the predicted output and the actual output. In this process, the algorithm continuously iterates and updates the network parameters until convergence, and finally obtains an optimized weight matrix from the input layer to the hidden layer, which can effectively express the complex correlation between the input features and the output targets;

[0067] The weight matrix from the input layer to the hidden layer is normalized based on the correlation between the input features and the output targets, and the normalized weight matrix is ​​transposed to obtain a transposed matrix of the normalized weight matrix;

[0068] Among them, when entering the normalization processing stage, each weight value is divided by the sum of the weights of its row to achieve row normalization of the weight matrix, ensuring that the sum of the weights of each row is 1. This step is to adjust the influence of each input feature on the hidden layer neurons to make it more balanced. Subsequently, in order to further strengthen the correlation between the input features and the output targets, the normalized weight matrix is ​​transposed to generate the transposed matrix of the normalized weight matrix;

[0069] The normalized weight matrix is ​​multiplied by its transposed matrix to construct the final correlation matrix, thereby quantitatively expressing the correlation between the input features and the output targets.

[0070] A prediction model is constructed, and the association matrix is ​​imported into the prediction model for coding training until the model prediction accuracy meets the requirements, and the final training parameters of the model are saved to obtain the motion state prediction model of the target AGV warehouse robot.

[0071] Among them, the service log refers to the detailed operation record generated by the target AGV warehouse robot during the actual operation process, including the operating data under different load masses and center of gravity positions, such as movement speed, angular velocity, acceleration and trajectory nodes. These data reflect the historical dynamic characteristics of the AGV under various operating conditions and provide valuable basic data resources for building a motion state prediction model.

[0072] It should be noted that this method collects historical dynamic feature data of AGV and uses the powerful time series processing capabilities of the long short-term memory network (LSTM network) to build a prediction model that can accurately predict the motion state of AGV. Through normalization and transposition operations, not only the correlation between input features and output targets is quantified, but also the generalization ability and prediction accuracy of the model are improved. The final motion state prediction model can effectively predict the motion state of AGV under different loads and center of gravity positions, providing important data support for the trajectory tracking control of AGV.

[0073] Preferably, the real-time load mass and real-time center of gravity position of the target AGV warehouse robot are obtained at a preset time node, and the predicted dynamic feature data of the target AGV warehouse robot within a preset time period is obtained based on the real-time load mass and real-time center of gravity position and combined with the motion state prediction model, such as Figure 2 As shown, specifically:

[0074] S202, during the operation of the target AGV storage robot, obtaining the real-time load mass and the real-time center of gravity position of the target AGV storage robot at a preset time node;

[0075] S204, importing the real-time load mass and real-time center of gravity position of the target AGV warehouse robot into the motion state prediction model for prediction;

[0076] S206. Obtain predicted dynamic feature data of the target AGV storage robot within a preset time period through prediction; the predicted dynamic feature data includes the predicted motion speed, predicted angular velocity, predicted acceleration and predicted trajectory nodes of the target AGV storage machine at each time stamp within the preset time period.

[0077] It should be noted that by obtaining the load mass and center of gravity of the AGV in real time during the operation, and combining the motion state prediction model to predict dynamic feature data, this method can monitor and accurately predict the motion state of the AGV in a complex operating environment in real time. This real-time prediction capability can improve the trajectory tracking accuracy and operation safety of the AGV, and also provide data support for the dynamic adjustment and real-time control of the AGV, providing a reliable basis for subsequent adjustment and optimization of the AGV's motion trajectory.

[0078] Preferably, the operating state of the target AGV warehouse robot within a preset time period is determined and analyzed based on the predicted dynamic feature data and combined with the three-dimensional grid coordinate system to obtain the operating state of the target AGV warehouse robot within the preset time period, specifically:

[0079] With the X-axis coordinate as the operating time dimension of the target AGV warehouse robot, the Y-axis coordinate as the horizontal position dimension of the target AGV warehouse robot on the operating plane, and the Z-axis coordinate as the vertical position dimension of the target AGV warehouse robot on the operating plane, a three-dimensional grid coordinate system is constructed, and the grid unit size of the three-dimensional grid coordinate system is initialized;

[0080] Obtain a preset operation task of the target AGV warehouse robot, obtain preset trajectory nodes of the target AGV warehouse robot at each time stamp within a preset time period in the preset operation task, and generate a preset operation trajectory curve of the target AGV warehouse robot within the preset time period in the three-dimensional grid coordinate system according to the preset trajectory nodes at each time stamp;

[0081] Obtaining predicted trajectory nodes of the target AGV warehouse robot at each time stamp within a preset time period from the predicted dynamic feature data, and generating a predicted running trajectory curve of the target AGV warehouse robot within the preset time period in the three-dimensional grid coordinate system according to the predicted trajectory nodes at each time stamp;

[0082] Performing search and analysis on each grid unit in the three-dimensional grid coordinate system;

[0083] If a certain grid unit contains both a predicted running trajectory curve and a preset running trajectory curve, the grid unit is marked as a type of grid unit;

[0084] If only a predicted running trajectory curve or only a preset running trajectory curve exists in a certain grid unit, the grid unit is marked as a second-class grid unit;

[0085] If neither the predicted running trajectory curve nor the preset running trajectory curve exists in a certain grid unit, the grid unit is marked as a third-class grid unit;

[0086] Count the total number of grid cells marked as type one and count the total number of grid cells marked as type two;

[0087] The total number of grid cells marked as the second category is divided by the total number of grid cells marked as the first category to obtain the trajectory error ratio of the target AGV warehouse robot within a preset time period;

[0088] Compare the trajectory error ratio of the target AGV warehouse robot within a preset time period with a preset ratio threshold;

[0089] When the trajectory error ratio of the target AGV warehouse robot within the preset time period is greater than the preset ratio threshold, the operation state of the target AGV warehouse robot within the preset time period is determined as a trajectory out-of-control state;

[0090] When the trajectory error ratio of the target AGV warehouse robot within the preset time period is not greater than the preset ratio threshold, the operating state of the target AGV warehouse robot within the preset time period is determined to be a trajectory stable state.

[0091] It should be noted that, first, a three-dimensional grid coordinate system is constructed, in which the X-axis represents the time dimension, the Y-axis represents the horizontal position dimension on the working plane, and the Z-axis represents the longitudinal position dimension on the working plane, and the size of the grid unit is set, wherein the size of the grid unit is set according to the control tracking accuracy requirements of the AGV. The higher the control tracking accuracy requirements, the smaller the grid unit. According to the preset working tasks of the AGV, the preset trajectory nodes of each time stamp in the preset time period are obtained, and the preset running trajectory curve is generated in the three-dimensional grid coordinate system. At the same time, the predicted dynamic feature data is used to obtain the predicted trajectory nodes of the AGV at each time stamp in the preset time period, and the predicted running trajectory curve is generated in the three-dimensional grid coordinate system. Each grid unit in the three-dimensional grid coordinate system is searched and analyzed, and the grid units are divided into one category, two categories, and three categories according to whether there are predicted trajectory curves and preset trajectory curves in the unit. The number of grid units in category one and category two is counted, and the ratio of the number of grid units in category two to the number of grid units in category one is calculated to obtain the trajectory error ratio. The trajectory error ratio refers to the ratio of the number of grid cells where the predicted trajectory curve overlaps with the preset trajectory curve in the three-dimensional grid coordinate system to the number of grid cells that only contain the predicted trajectory curve or the preset trajectory curve, which is used to evaluate the trajectory tracking accuracy of the AGV within a preset time period. The trajectory error ratio is compared with the preset ratio threshold. If the trajectory error ratio is greater than the threshold, the AGV is judged to be in a trajectory out-of-control state; if it is not greater than the threshold, it is judged to be in a trajectory stable state.

[0092] In summary, by constructing a three-dimensional grid coordinate system and using it to compare and analyze the predicted trajectory and preset trajectory of the AGV, this method can quantitatively evaluate the trajectory tracking accuracy of the AGV within a preset time period. By calculating the trajectory error ratio, the system can accurately determine the operating status of the AGV, so as to take corresponding control measures in time. This determination method not only improves the safety and reliability of AGV operations, but also provides technical support for the realization of real-time monitoring and intelligent decision-making of AGV.

[0093] Preferably, if the operating state of the target AGV warehouse robot within the preset time period is a trajectory out-of-control state, a correction control scheme is generated, and correction control processing is performed on the target AGV warehouse robot based on the correction control scheme, specifically:

[0094] If the operating state of the target AGV warehouse robot within the preset time period is a trajectory out-of-control state, then the X-coordinate nodes corresponding to each grid unit marked as the second type are obtained, and the X-coordinate nodes corresponding to each grid unit marked as the second type are determined as the trajectory out-of-control time nodes of the target AGV warehouse robot;

[0095] At the same time, the predicted dynamic feature data of the target AGV warehouse robot within a preset time period is obtained, and the predicted motion control parameters corresponding to each trajectory out-of-control time node of the target AGV warehouse robot are extracted from the predicted dynamic feature data; wherein the predicted motion control parameters include predicted motion speed, predicted angular velocity and predicted acceleration;

[0096] Obtain a preset control scheme of the target AGV warehouse robot, and obtain preset motion control parameters corresponding to each trajectory out-of-control time node of the target AGV warehouse robot in the preset control scheme; wherein the preset motion control parameters include a preset motion speed, a preset angular velocity, and a preset acceleration;

[0097] Calculate the difference between the same predicted motion control parameter and the preset motion control parameter of the target AGV warehouse robot at each trajectory out-of-control time node, and obtain the drift amplitude of each motion control parameter of the target AGV warehouse robot at each trajectory out-of-control time node;

[0098] Generate a deviation correction control scheme for the target AGV warehouse robot at each trajectory out-of-control time node according to the drift amplitude of each motion control parameter of the target AGV warehouse robot at each trajectory out-of-control time node;

[0099] At the corresponding trajectory out-of-control time node, the target AGV warehouse robot is corrected and regulated based on the corresponding correction control scheme.

[0100] Among them, the above-mentioned preset control scheme refers to a set of motion control strategies pre-formulated by relevant technical personnel under normal working conditions based on the operating tasks, design parameters and operating experience of the target AGV warehouse robot. The scheme includes the ideal motion control parameters of the target AGV warehouse robot at each time node during the operation process, such as motion speed, angular velocity and acceleration. In addition, the method steps for obtaining the preset motion control parameters are: first, extract the information corresponding to each trajectory out-of-control time node from the preset control scheme; then, determine the expected ideal motion control parameter values ​​at these time nodes according to the provisions in the scheme. The purpose of this is to compare the predicted motion control parameters with the actual operation in the future, so as to evaluate and adjust the deviation and ensure that the AGV warehouse robot can operate stably according to the predetermined path and speed.

[0101] It should be noted that if the target AGV warehouse robot is in a state of trajectory out of control during the preset time period, the system will identify all X-coordinate nodes marked as second-class grid cells, which correspond to the time points when the AGV trajectory is out of control, and extract the predicted motion control parameters corresponding to the time nodes of the trajectory out of control from the predicted dynamic feature data, and obtain the corresponding preset motion control parameters from the preset control scheme. Then, the difference between the predicted motion control parameters and the preset motion control parameters (such as the difference between the predicted motion speed and the preset motion speed) is calculated to obtain the drift amplitude of each motion control parameter, which reflects the deviation between the actual motion of the AGV and the expected motion. Then, the drift amplitude is analyzed to identify the motion control parameters that cause the trajectory to be out of control; then, according to the size and direction of the drift amplitude, a corresponding correction strategy is formulated, and the motion control parameters of the AGV are adjusted to correct the AGV and restore its trajectory tracking accuracy. In summary, by identifying the time nodes of trajectory out-of-control, calculating the drift amplitude of motion control parameters and generating a correction control scheme, this method can effectively correct the trajectory out-of-control state of the AGV warehouse robot in real time, which not only improves the trajectory tracking accuracy of the AGV, but also enhances its adaptability and stability in complex working environments, thereby improving the overall operating performance and safety of the AGV.

[0102] In this embodiment, the AGV warehouse robot motion trajectory tracking control method may further include the following steps:

[0103] Obtain a preset control scheme of the target AGV warehouse robot, and draw a preset posture three-dimensional model diagram corresponding to the normal operation of the target AGV warehouse robot at each operation time node according to the preset control scheme;

[0104] Construct a decision tree, and initialize the splitting node of the decision tree according to the operation time node of the target AGV storage robot, iteratively split a number of tree forks in the decision tree according to the splitting node, and create storage space in the tree fork;

[0105] The preset posture three-dimensional model diagrams corresponding to the normal operation of the target AGV warehouse robot at each operation time node are stored in the storage space of the corresponding tree fork to form a posture discrimination decision tree;

[0106] Acquire the real-time pose image information of the target AGV warehouse robot, construct a real-time pose three-dimensional model diagram of the target AGV warehouse robot according to the real-time pose image information, and obtain the time node of collecting the real-time pose image information;

[0107] Importing the time node of collecting the real-time posture image information into the posture discrimination decision tree for pairing, obtaining the tree fork of the time node of collecting the real-time posture image information, and marking the matched tree fork;

[0108] Extracting a corresponding preset posture three-dimensional model graph from the storage space of the marked tree fork, introducing a Euclidean distance algorithm, and calculating the degree of overlap between the real-time posture three-dimensional model graph and the extracted preset posture three-dimensional model graph based on the Euclidean distance algorithm;

[0109] If the overlap is greater than the preset overlap, the current control strategy of the target AGV warehouse robot is maintained unchanged, and the position and posture of the target AGV warehouse robot is monitored in real time;

[0110] If the overlap is not greater than the preset overlap, the real-time pose 3D model image and the extracted preset pose 3D model image are registered based on an iterative closest point matching algorithm. After the registration is completed, the model area where the real-time pose 3D model image overlaps with the extracted preset pose 3D model image is deleted, and the area where the real-time pose 3D model image does not overlap with the extracted preset pose 3D model image is retained to obtain the offset state model image of the target AGV warehouse robot;

[0111] The offset state model diagram of the target AGV storage robot is subjected to feature extraction processing to obtain the position offset direction and offset amount of the target AGV storage robot; and the target AGV storage robot is subjected to deviation correction control processing according to the position offset direction and offset amount.

[0112] Among them, the algorithm steps for extracting features from the offset state model graph of the target AGV warehouse robot and obtaining the offset direction and offset are as follows: First, the offset state model graph is preprocessed, including denoising and edge detection, to highlight the key features in the model graph. Next, a feature extraction algorithm (such as principal component analysis or contour extraction) is used to identify significant feature points in the model graph, which represent the main areas of AGV posture offset. Then, the offset direction is determined by calculating the geometric relationship between the feature points, for example, by calculating the direction of the line connecting the center point of the feature point and the center point of the preset model graph. Next, the distance measurement method (such as Euclidean distance) is used to calculate the distance between the feature point and the corresponding point of the preset model graph to obtain the offset. Finally, the offset direction and offset are integrated to generate a posture offset feature vector for subsequent correction control processing.

[0113] It should be noted that this method can accurately monitor the real-time posture state of the AGV and quickly identify posture deviation. In the case of posture deviation, the system can automatically generate a correction control plan and adjust the motion state of the AGV in time to ensure its trajectory tracking accuracy and operation stability, effectively improving the adaptive ability and correction efficiency of the AGV in complex operating environments.

[0114] In this embodiment, the AGV warehouse robot motion trajectory tracking control method may further include the following steps:

[0115] Obtain the fault log of the target AGV warehouse robot, perform feature extraction on the fault log, and obtain the state transition information corresponding to various fault types of the target AGV warehouse robot;

[0116] The Markov chain is introduced, and the fault state of the target AGV warehouse robot is used as the state node of the Markov chain. According to the fault type and state transition information in the fault log, the state transition graph is constructed, and the state transition probability matrix is ​​initialized.

[0117] Based on the fault occurrence sequence in the fault log, the number of transitions between each fault type is counted, the state transition probability is calculated, and the calculated state transition probability is filled into the state transition probability matrix to generate a fault state probability matrix;

[0118] Constructing a Markov model, and importing the fault state probability matrix into the Markov model for encoding learning;

[0119] After the target AGV warehouse robot is subjected to the correction control process based on the correction control scheme, the motion trajectory of the target AGV warehouse robot is continuously tracked and monitored at the next preset time node;

[0120] If the operating state determined at the next preset time node of the target AGV warehouse robot is still a trajectory out-of-control state, the real-time characteristic parameters of the target AGV warehouse robot are obtained, wherein the real-time characteristic parameters include operating voltage, operating current, motor temperature and noise level;

[0121] Importing the real-time characteristic parameters of the target AGV warehouse robot into the Markov model to perform fault probability deduction, and obtaining the fault state probability value of the target AGV warehouse robot;

[0122] If the fault state probability value of the target AGV storage robot is greater than the preset probability value, the target AGV storage robot is controlled to stop operating.

[0123] Among them, the real-time characteristic parameters of the target AGV warehouse robot are imported into the Markov model to perform fault probability deduction, and the fault state probability value of the target AGV warehouse robot is obtained. Specifically, the real-time collected working voltage, current, motor temperature and noise level are standardized to generate a multidimensional feature vector; then the feature vector is mapped to the state node code of the Markov chain to determine the current fault state category; based on the state transition probability matrix, the transition probability of the state node to each fault state is extracted, and the dynamic weight is calculated in combination with the historical transition frequency; then, a joint probability distribution is generated according to the correlation between the current state and the previous state; finally, the transition probability of each fault path is synthesized through a weighted sum algorithm, and a probability value set of fault types such as mechanical failure, circuit abnormality, motor overheating and transmission failure is output, and its maximum a posteriori probability value is obtained as the final fault state probability value.

[0124] It should be noted that the fault log of AGV is first collected and key features are extracted from it. These features reflect the state changes of AGV when different fault types occur. Using the extracted fault log features, the fault state of AGV is regarded as the state node in the Markov chain, the state transition diagram is constructed, and the state transition probability matrix is ​​initialized. By analyzing the fault occurrence sequence in the fault log, the number of transitions and state transition probabilities between different fault types are counted and calculated to form a fault state probability matrix. The calculated fault state probability matrix is ​​imported into the Markov model for encoding learning so that the model can predict future fault states based on historical fault data. After the deviation correction control, the motion trajectory of AGV continues to be monitored. If the operating state of AGV is still out of control at the next preset time node, its real-time feature parameters are obtained. The real-time feature parameters are input into the Markov model for fault probability deduction. If the deduced fault state probability value exceeds the preset threshold, the AGV is controlled to stop working. This method realizes real-time tracking and fault prediction of the motion trajectory of AGV warehouse robots through fault log analysis and Markov model. By extracting features from fault logs and calculating state transition probabilities, a Markov model that can reflect the fault status of AGV is constructed, which can monitor the motion trajectory of AGV in real time and perform fault probability deduction in a timely manner when the trajectory is out of control. It can effectively predict the potential faults of AGV, take measures such as stopping operations in advance, prevent faults from occurring, improve the safety and reliability of the system, and optimize the operating efficiency of AGV.

[0125] like Figure 3As shown, the second aspect of the present invention discloses an AGV warehouse robot motion trajectory tracking and control system 6, the AGV warehouse robot motion trajectory tracking and control system includes a memory 41 and a processor 52, the memory 41 stores an AGV warehouse robot motion trajectory tracking control method program, when the AGV warehouse robot motion trajectory tracking control method program is executed by the processor 52, any one of the steps of the AGV warehouse robot motion trajectory tracking control method is implemented.

[0126] The third aspect of the present invention discloses a computer-readable storage medium, which includes an AGV warehouse robot motion trajectory tracking control method program. When the AGV warehouse robot motion trajectory tracking control method program is executed by a processor, any one of the steps of the AGV warehouse robot motion trajectory tracking control method is implemented.

[0127] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0128] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0129] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0130] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0131] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0132] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A motion trajectory tracking control method for an AGV warehouse robot, characterized in that: The following steps are involved: Acquire historical dynamic feature data of the target AGV warehouse robot in the future under working conditions of different load masses and center of gravity positions, and construct a motion state prediction model of the target AGV warehouse robot according to the historical dynamic feature data; Acquire the real-time load mass and real-time center of gravity position of the target AGV warehouse robot at a preset time node, and acquire the predicted dynamic feature data of the target AGV warehouse robot within a preset time period based on the real-time load mass and real-time center of gravity position and in combination with the motion state prediction model; According to the predicted dynamic feature data and in combination with the three-dimensional grid coordinate system, the operating state of the target AGV warehouse robot within the preset time period is determined and analyzed to obtain the operating state of the target AGV warehouse robot within the preset time period; wherein the operating state includes a trajectory out of control state and a trajectory stable state; If the target AGV warehouse robot is in an out-of-control state within a preset time period, a deviation correction control scheme is generated, and a deviation correction control process is performed on the target AGV warehouse robot based on the deviation correction control scheme; Among them, according to the predicted dynamic feature data and combined with the three-dimensional grid coordinate system, the operating state of the target AGV warehouse robot within the preset time period is determined and analyzed to obtain the operating state of the target AGV warehouse robot within the preset time period, specifically: With the X-axis coordinate as the operating time dimension of the target AGV warehouse robot, the Y-axis coordinate as the horizontal position dimension of the target AGV warehouse robot on the operating plane, and the Z-axis coordinate as the vertical position dimension of the target AGV warehouse robot on the operating plane, a three-dimensional grid coordinate system is constructed, and the grid unit size of the three-dimensional grid coordinate system is initialized; Obtain a preset operation task of the target AGV warehouse robot, obtain preset trajectory nodes of the target AGV warehouse robot at each time stamp within a preset time period in the preset operation task, and generate a preset operation trajectory curve of the target AGV warehouse robot within the preset time period in the three-dimensional grid coordinate system according to the preset trajectory nodes at each time stamp; Obtaining predicted trajectory nodes of the target AGV warehouse robot at each time stamp within a preset time period from the predicted dynamic feature data, and generating a predicted running trajectory curve of the target AGV warehouse robot within the preset time period in the three-dimensional grid coordinate system according to the predicted trajectory nodes at each time stamp; Performing search and analysis on each grid unit in the three-dimensional grid coordinate system; If a certain grid unit contains both a predicted running trajectory curve and a preset running trajectory curve, the grid unit is marked as a type of grid unit; If only a predicted running trajectory curve or only a preset running trajectory curve exists in a certain grid unit, the grid unit is marked as a second-class grid unit; If neither the predicted running trajectory curve nor the preset running trajectory curve exists in a certain grid unit, the grid unit is marked as a third-class grid unit; Count the total number of grid cells marked as type one and count the total number of grid cells marked as type two; The total number of grid cells marked as the second category is divided by the total number of grid cells marked as the first category to obtain the trajectory error ratio of the target AGV warehouse robot within a preset time period; Compare the trajectory error ratio of the target AGV warehouse robot within a preset time period with a preset ratio threshold; When the trajectory error ratio of the target AGV warehouse robot within the preset time period is greater than the preset ratio threshold, the operating state of the target AGV warehouse robot within the preset time period is determined to be a trajectory out of control state; otherwise, the operating state of the target AGV warehouse robot within the preset time period is determined to be a trajectory stable state.

2. The motion trajectory tracking control method of an AGV storage robot according to claim 1 is characterized in that: The historical dynamic feature data of the target AGV warehouse robot in the future under the working conditions of different load masses and center of gravity positions are obtained, and the motion state prediction model of the target AGV warehouse robot is constructed according to the historical dynamic feature data, specifically: Introduce the long short-term memory network and initialize the input layer, hidden layer and output layer of the long short-term memory network; import the load mass and center of gravity position as input features into the input layer, and use the historical dynamic feature data in the future time as the output target of the output layer; By using the time series processing capability of the long short-term memory network, the correlation between input features and output targets is automatically learned, and the network parameters are optimized through the back-propagation algorithm to obtain the weight matrix from the input layer to the hidden layer. The weight matrix from the input layer to the hidden layer is normalized based on the correlation between the input features and the output targets, and the normalized weight matrix is ​​transposed to obtain a transposed matrix of the normalized weight matrix; The normalized weight matrix is ​​multiplied by its transposed matrix to construct the final correlation matrix, thereby quantitatively expressing the correlation between the input features and the output targets. Construct a prediction model, and import the association matrix into the prediction model for coding training until the model prediction accuracy meets the requirements, save the final training parameters of the model, and obtain the motion state prediction model of the target AGV warehouse robot; The historical dynamic feature data includes the historical motion speed, historical angular velocity, historical acceleration and historical trajectory nodes of the target AGV storage robot at each time stamp in the future.

3. The motion trajectory tracking control method of an AGV storage robot according to claim 1, characterized in that: The real-time load mass and real-time center of gravity position of the target AGV warehouse robot are obtained at a preset time node, and the predicted dynamic feature data of the target AGV warehouse robot within a preset time period is obtained based on the real-time load mass and real-time center of gravity position and combined with the motion state prediction model, specifically: During the operation of the target AGV warehouse robot, the real-time load mass and real-time center of gravity position of the target AGV warehouse robot are obtained at a preset time node; Importing the real-time load mass and real-time center of gravity position of the target AGV warehouse robot into the motion state prediction model for prediction; Through prediction, the predicted dynamic feature data of the target AGV storage robot within a preset time period is obtained; the predicted dynamic feature data includes the predicted motion speed, predicted angular velocity, predicted acceleration and predicted trajectory nodes of the target AGV storage machine at each time stamp within the preset time period.

4. The motion trajectory tracking control method of an AGV warehouse robot according to claim 1, characterized in that: If the operating state of the target AGV warehouse robot within the preset time period is a trajectory stable state, the current control strategy of the target AGV warehouse robot is maintained unchanged, and the motion trajectory of the target AGV warehouse robot continues to be tracked and monitored at the next preset time node.

5. The motion trajectory tracking control method of an AGV storage robot according to claim 1, characterized in that: If the target AGV warehouse robot is in a trajectory out-of-control state during the preset time period, a correction control scheme is generated, and correction control processing is performed on the target AGV warehouse robot based on the correction control scheme, specifically: If the operating state of the target AGV warehouse robot within the preset time period is a trajectory out-of-control state, then the X-coordinate nodes corresponding to each grid unit marked as the second type are obtained, and the X-coordinate nodes corresponding to each grid unit marked as the second type are determined as the trajectory out-of-control time nodes of the target AGV warehouse robot; At the same time, the predicted dynamic feature data of the target AGV warehouse robot within a preset time period is obtained, and the predicted motion control parameters corresponding to each trajectory out-of-control time node of the target AGV warehouse robot are extracted from the predicted dynamic feature data; wherein the predicted motion control parameters include predicted motion speed, predicted angular velocity and predicted acceleration; Obtain a preset control scheme of the target AGV warehouse robot, and obtain preset motion control parameters corresponding to each trajectory out-of-control time node of the target AGV warehouse robot in the preset control scheme; wherein the preset motion control parameters include a preset motion speed, a preset angular velocity, and a preset acceleration; Calculate the difference between the same predicted motion control parameter and the preset motion control parameter of the target AGV warehouse robot at each trajectory out-of-control time node, and obtain the drift amplitude of each motion control parameter of the target AGV warehouse robot at each trajectory out-of-control time node; Generate a deviation correction control scheme for the target AGV warehouse robot at each trajectory out-of-control time node according to the drift amplitude of each motion control parameter of the target AGV warehouse robot at each trajectory out-of-control time node; At the corresponding trajectory out-of-control time node, the target AGV warehouse robot is corrected and regulated based on the corresponding correction control scheme.

6. An AGV warehouse robot motion trajectory tracking control system, characterized in that: The AGV warehouse robot motion trajectory tracking control system includes a memory and a processor. The memory stores an AGV warehouse robot motion trajectory tracking control method program. When the AGV warehouse robot motion trajectory tracking control method program is executed by the processor, the AGV warehouse robot motion trajectory tracking control method steps as described in any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes an AGV warehouse robot motion trajectory tracking control method program. When the AGV warehouse robot motion trajectory tracking control method program is executed by a processor, the AGV warehouse robot motion trajectory tracking control method as described in any one of claims 1 to 5 is implemented.

Images