Robot repeated positioning system and positioning method thereof

Abstract

The invention provides a robot repeated positioning system and a positioning method thereof, and relates to the field of automatic positioning systems. The robot repeated positioning system comprises an intelligent positioning system, the intelligent positioning system comprises a data acquisition unit, a data analysis unit, a decision model unit, a robot control unit, a database unit and an estimation correction unit, and the data acquisition unit is connected with the data analysis unit and the robot control unit. Through the use of the positioning method based on the repeated positioning system, the robot can autonomously learn, update, configure and select a used positioning data source according to the condition in a specific scene, formulate a positioning point suitable for the scene, and deduce and decide cooperative positioning auxiliary information suitable for the scene. The problem of repeated deployment of the personnel positioning system in various scenes is reduced, the complexity of system deployment and debugging is reduced to a certain extent, and the cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of automated positioning systems, specifically to a robot repetitive positioning system and its positioning method. Background Technology

[0002] In automated production, robots are the main equipment used to replace manual labor for workpiece transportation and processing. During robot movement, high repeatability is required to ensure the accuracy of workpiece processing. In existing technologies, robot repeatability typically uses reference points for positioning.

[0003] Patent (CN201510988678.X) discloses a robot repetitive positioning system, comprising: an encoder mounted on each rotating joint of the robot for measuring the rotation angle of the joint; a three-axis accelerometer mounted on the robot's operating end for measuring the acceleration of the robot's operating end in three dimensions; a CCD lens mounted on the robot's operating end for capturing marker points located on a positioning reference object, the marker points being a plurality of concentrically arranged ring marks, with at least three marker points on the positioning reference object; a control module communicatively connected to the encoder, three-axis accelerometer, and CCD lens for processing the detected positioning parameters; and a database module communicatively connected to the control module for storing historical data of the robot positioning process. This invention also provides a positioning method for the above-mentioned robot repetitive positioning system. This invention overcomes the shortcomings of the prior art and improves the accuracy of robot repetitive positioning.

[0004] However, current repetitive positioning systems still rely mainly on manual selection by operators for the selection of positioning points. The quality of manual selection depends on the operator's experience, making it difficult to achieve optimal selection in real time. Furthermore, during multi-point positioning tests, the positioning points will continuously change in a short period of time, which is detrimental to the stability of the control system and cannot guarantee the optimality of the positioning point selection. This increases the power consumption of the robot's operating end, resulting in higher operating costs. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a robot repetitive positioning system and method, which solves the problems.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: a robot repetitive positioning system, comprising an intelligent positioning system, wherein the intelligent positioning system includes a data acquisition unit, a data analysis unit, a decision model unit, a robot control unit, a database unit, and an estimation and correction unit; the data acquisition unit is connected to the data analysis unit and the robot control unit; the data analysis unit and the robot control unit are connected to the estimation and correction unit; the robot control unit is connected to the decision model unit; and the decision model unit is connected to the database unit.

[0009] Preferably, the data acquisition unit includes an information acquisition module, a positioning acquisition module, a data conversion module, and a data transmission module. The information acquisition module is connected to a robot operation information acquisition module, which includes robot operation power information, robot operation point information, and robot inertial amplitude information. The positioning acquisition module includes a node information acquisition module and a fixed-point information acquisition module.

[0010] Preferably, the data analysis unit includes an intelligent processing module, a data analysis module, a data classification module, and a data judgment module. The intelligent processing module performs intelligent processing based on an artificial intelligence processing chip, and the data judgment module is used to judge the analyzed data and determine the optimal data value according to preset rules.

[0011] Preferably, the decision model unit includes a deep learning decision model and an optimal positioning point determination module, wherein the deep learning decision model is a decision model constructed based on a reinforcement learning model of a Markov decision model, and the optimal positioning point determination module uses the optimal power and minimum amplitude values ​​of the robot operation as the basis for decision-making.

[0012] Preferably, the robot control unit includes an initial running point calibration module, a mechanical operation module, and a mechanical control module, wherein the initial running point calibration module determines the initial point based on the target position given by the estimation and correction unit.

[0013] Preferably, the estimation correction unit includes a target position estimation module, an estimation position correction module, and a corrected position determination module. The target position estimation module estimates the current target position based on the target positioning point obtained from the historical robot output power, while the estimation position correction module and the corrected position determination module filter and correct the actual target position data based on the current target position estimation information to obtain the corrected target position value.

[0014] Preferably, a positioning method for a robot repetitive positioning system specifically includes the following steps:

[0015] S1. Model Building

[0016] The model space is defined, the Actor-Critic framework is used, convolutional neural networks and recurrent neural networks are selected for fitting and optimizing the Q function, the model is built based on the stochastic gradient descent algorithm, the model is trained with a large amount of training data and evaluated on the validation set, the model that has completed the evaluation test is used, and the optimal decision strategy is preset, with the decision-making basis and decision points based on the optimal power and minimum amplitude values.

[0017] S2. Data Acquisition

[0018] The data acquisition unit collects robot operation data, including the three-dimensional acceleration of the robot's operating end, the output power of the robot's operating end, the position of the target on the robot's operating end, and the operating amplitude value of the robot's operating end.

[0019] S3. Data and Information Analysis

[0020] Based on the data values ​​collected in step S2, the power output value and inertial amplitude value of the current robot operating end are determined. According to the running data of the operating end repeatedly, the data is introduced into the decision model. The optimal positioning point is determined by the optimal function when the power consumption of the robot operating end in the positioning system is the minimum and the amplitude value is the minimum.

[0021] S4. Final Output

[0022] Based on the decision-making of the optimal positioning point, the robot control unit corrects and calibrates the initial running point, and adjusts the optimal output power of the robot's operating end based on the mechanical operation module and the mechanical control module.

[0023] Preferably, the initial running point is initially estimated and marked based on historical running positioning points, and the second positioning point obtained during the first run based on this positioning point is filtered by establishing a Kalman filter equation to obtain the corresponding target corrected position.

[0024] (III) Beneficial Effects

[0025] This invention provides a robot repetitive localization system and method. It has the following beneficial effects:

[0026] 1. This invention provides a robot repetitive localization system and its localization method. By using historical localization points for preliminary estimation and determining the initial running point, filtering and correction are performed on the secondary localization points of the first run to obtain an optimized estimate of the current running value. This overcomes the shortcomings of using fixed historical parameters to locate the target localization point incorrectly, thereby improving the overall positioning accuracy of the target. At the same time, a decision model is constructed. By pre-setting the optimal decision strategy, the optimal power and minimum amplitude values ​​are used as the decision basis. Based on the calculation and analysis of multiple running parameters, and the autonomous learning of the decision model, the learning process does not require any human intervention and can automatically adjust the learning results according to changes in environmental conditions. This completes a more comprehensive and accurate calculation and analysis of the target localization error, determines the optimal localization point, and ensures the determination of the localization point with minimum power and minimum amplitude.

[0027] 2. This invention provides a robot repetitive positioning system and its positioning method. By using a positioning method based on the repetitive positioning system, the robot can autonomously learn and update the configuration of the selected positioning data source according to the specific scenario, formulate positioning points suitable for the scenario, and reason and decide on collaborative positioning auxiliary information suitable for the scenario. This reduces the problem of repeated deployment of personnel positioning systems in various scenarios, alleviates the complexity of system deployment and debugging to a certain extent, and reduces costs. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0029] Figure 2 This is a schematic diagram of the structural framework of each unit of the system of the present invention;

[0030] Figure 3 This is a schematic diagram of the information acquisition module of the present invention;

[0031] Figure 4 This is a schematic diagram of the positioning and acquisition module of the present invention.

[0032] The system comprises: 1. Intelligent positioning system; 2. Node information acquisition module; 3. Fixed-point information acquisition module; 4. Robot operating power information; 5. Robot operating point information; 6. Robot inertial amplitude information; 101. Data acquisition unit; 102. Data analysis unit; 103. Decision model unit; 104. Robot control unit; 105. Database unit; 106. Estimation and correction unit; 10101. Information acquisition module; 10102. Positioning acquisition module; 10103. Data conversion module; 1010 4. Data transmission module; 10201. Intelligent processing module; 10202. Data analysis module; 10203. Data classification module; 10204. Data judgment module; 10301. Deep learning decision model; 10302. Optimal positioning point determination module; 10401. Initial running point calibration module; 10402. Mechanical operation module; 10403. Mechanical control module; 10601. Target position estimation module; 10602. Estimated position correction module; 10603. Corrected position determination module. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Example:

[0035] like Figure 1-4 As shown, this embodiment of the invention provides a robot repetitive positioning system, including an intelligent positioning system 1. The intelligent positioning system 1 includes a data acquisition unit 101, a data analysis unit 102, a decision model unit 103, a robot control unit 104, a database unit 105, and an estimation and correction unit 106. The data acquisition unit 101 is connected to the data analysis unit 102 and the robot control unit 104. The data analysis unit 102 and the robot control unit 104 are connected to the estimation and correction unit 106. The robot control unit 104 is connected to the decision model unit 103. The decision model unit 103 is connected to the database unit 105.

[0036] In this embodiment, as Figure 2 , 3As shown in Figure 4, the data acquisition unit 101 includes an information acquisition module 10101, a positioning acquisition module 10102, a data conversion module 10103, and a data transmission module 10104. The information acquisition module 10101 is connected to a robot operation information acquisition module, which includes robot operation power information 4, robot operation point information 5, and robot inertial amplitude information 6. The positioning acquisition module 10102 includes a node information acquisition module 2 and a fixed-point information acquisition module 3. The information acquisition module 10101 collects the power, amplitude, and positioning point information of the operating end during the robot's operation, quickly obtaining various indicator data to provide a data basis for subsequent decision-making. The positioning acquisition module 10102 collects the operating status and data value information of each operating node to determine the position information of the operating node.

[0037] In this embodiment, the data analysis unit 102 includes an intelligent processing module 10201, a data analysis module 10202, a data classification module 10203, and a data judgment module 10204. The intelligent processing module 10201 performs intelligent processing based on an artificial intelligence processing chip. The data judgment module 10204 is used to process and judge the analyzed data and determine the optimal data value according to a preset rule, which is the minimum output power combination value based on the minimum amplitude value. The decision model unit 103 includes a deep learning decision model 10301 and an optimal positioning point determination module 10302. The deep learning decision model 10301 is a decision model constructed based on a reinforcement learning model of a Markov decision model.

[0038] The optimal positioning point determination module 10302 makes decisions based on the optimal power and minimum amplitude values ​​of the robot operation. The robot control unit 104 includes an initial running point calibration module 10401, a mechanical operation module 10402, and a mechanical control module 10403. The initial running point calibration module 10401 determines the initial point based on the target position given by the estimation correction unit 106. The estimation correction unit 106 includes a target position estimation module 10601, an estimation position correction module 10602, and a corrected position determination module 10603. The target position estimation module 10601 estimates the current target position based on the target positioning point obtained from the historical robot output power. The estimation position correction module 10602 and the corrected position determination module 10603 filter and correct the actual target position data based on the current target position estimation information to obtain the corrected target position value. The filtering correction uses the Kalman filtering equation for filtering.

[0039] A positioning method for a robot repetitive positioning system specifically includes the following steps:

[0040] S1. Model Building

[0041] The model space is defined, the Actor-Critic framework is used, convolutional neural networks and recurrent neural networks are selected for fitting and optimizing the Q function, the model is built based on the stochastic gradient descent algorithm, the model is trained with a large amount of training data and evaluated on the validation set, the model that has completed the evaluation test is used, and the optimal decision strategy is preset, with the decision-making basis and decision points based on the optimal power and minimum amplitude values.

[0042] The formula for fitting and optimizing the Q function using a neural network is as follows:

[0043]

[0044] Where, r t Let γ be the reward function obtained after each time period T, and γ be the decay factor. State s is the input of the neural network, and Q(s,a;θ) represents the output of the neural network, where θ is the weight in the neural network.

[0045] S2. Data Acquisition

[0046] The data acquisition unit 101 collects robot operation data, including the three-dimensional acceleration of the robot's operating end, the output power of the robot's operating end, the position of the target on the robot's operating end, and the operating amplitude value of the robot's operating end.

[0047] An encoder is installed on each rotating joint of the robot to measure the rotation angle of the joint. A three-axis accelerometer is installed on the robot's operating end to measure the acceleration of the robot's operating end in three dimensions. The amplitude value is collected by an amplitude sensor. Both the encoder and the three-axis accelerometer are connected to the acquisition model, and the database unit 105 is connected to the acquisition module to store historical data of the robot's positioning process.

[0048] S3. Data and Information Analysis

[0049] Based on the data values ​​collected in step S2, the power output value and inertial amplitude value of the current robot operating end are determined. According to the running data of the operating end repeatedly, the data is introduced into the decision model. The optimal positioning point is determined by the optimal function when the power consumption of the robot operating end in the positioning system is the minimum and the amplitude value is the minimum.

[0050] S4. Final Output

[0051] Based on the decision-making of the optimal positioning point, the robot control unit 104 corrects and calibrates the initial running point, and adjusts the output power of the robot operation end based on the mechanical operation module 10402 and the mechanical control module 10403.

[0052] The initial running point is initially estimated and marked based on historical running positioning points. The second positioning point obtained during the first run based on this positioning point is filtered by establishing a Kalman filter equation to obtain the corresponding target correction position.

[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A robot repetitive positioning system, comprising an intelligent positioning system (1), characterized in that: The intelligent positioning system (1) includes a data acquisition unit (101), a data analysis unit (102), a decision model unit (103), a robot control unit (104), a database unit (105), and an estimation correction unit (106). The data acquisition unit (101) is connected to the data analysis unit (102) and the robot control unit (104). The data analysis unit (102) and the robot control unit (104) are connected to the estimation correction unit (106). The robot control unit (104) is connected to the decision model unit (103). The decision model unit (103) is connected to the database unit (105).

2. The robot repetitive positioning system according to claim 1, characterized in that: The data acquisition unit (101) includes an information acquisition module (10101), a positioning acquisition module (10102), a data conversion module (10103), and a data transmission module (10104). The information acquisition module (10101) is connected to a robot operation information acquisition module. The robot operation information acquisition module includes robot operation power information (4), robot operation point information (5), and robot inertial amplitude information (6). The positioning acquisition module (10102) includes a node information acquisition module (2) and a fixed-point information acquisition module (3).

3. The robot repetitive positioning system according to claim 1, characterized in that: The data analysis unit (102) includes an intelligent processing module (10201), a data analysis module (10202), a data classification module (10203), and a data judgment module (10204). The intelligent processing module (10201) performs intelligent processing based on an artificial intelligence processing chip. The data judgment module (10204) is used to judge the data after processing and analysis, and determine the optimal data value according to preset rules.

4. A robot repetitive positioning system according to claim 1, characterized in that: The decision model unit (103) includes a deep learning decision model (10301) and an optimal positioning point determination module (10302). The deep learning decision model (10301) is a decision model constructed based on a reinforcement learning model of a Markov decision model. The optimal positioning point determination module (10302) uses the optimal power and minimum amplitude values ​​of the robot operation as the decision basis decision points.

5. A robot repetitive positioning system according to claim 1, characterized in that: The robot control unit (104) includes an initial running point calibration module (10401), a mechanical operation module (10402), and a mechanical control module (10403), wherein the initial running point calibration module (10401) determines the initial point based on the target position given by the estimation correction unit (106).

6. A robot repetitive positioning system according to claim 1, characterized in that: The estimation correction unit (106) includes a target position estimation module (10601), an estimation position correction module (10602), and a corrected position determination module (10603). The target position estimation module (10601) estimates the current target position based on the target positioning point obtained from the historical robot output power, while the estimation position correction module (10602) and the corrected position determination module (10603) obtain the corrected target position value by filtering and correcting the actual target position data based on the current target position estimation information.

7. A positioning method for a robot repetitive positioning system, characterized in that, Specifically, the following steps are included: S1. Model Building The model space is defined, the Actor-Critic framework is used, convolutional neural networks and recurrent neural networks are selected for fitting and optimizing the Q function, the model is built based on the stochastic gradient descent algorithm, the model is trained with a large amount of training data and evaluated on the validation set, the model that has completed the evaluation test is used, and the optimal decision strategy is preset, with the decision-making basis and decision points based on the optimal power and minimum amplitude values. S2. Data Acquisition The data acquisition unit collects robot operation data, including the three-dimensional acceleration of the robot's operating end, the output power of the robot's operating end, the position of the target on the robot's operating end, and the operating amplitude value of the robot's operating end. S3. Data and Information Analysis Based on the data values ​​collected in step S2, the power output value and inertial amplitude value of the current robot operating end are determined. According to the running data of the operating end repeatedly, the data is introduced into the decision model. The optimal positioning point is determined by the optimal function when the power consumption of the robot operating end in the positioning system is the minimum and the amplitude value is the minimum. S4. Final Output Based on the decision-making of the optimal positioning point, the robot control unit corrects and calibrates the initial running point, and adjusts the optimal output power of the robot's operating end based on the mechanical operation module and the mechanical control module.

8. The positioning method of a robot repetitive positioning system according to claim 7, characterized in that: The initial running point is initially estimated and marked based on historical running positioning points. The second positioning point obtained during the first run based on this positioning point is filtered by establishing a Kalman filter equation to obtain the corresponding target correction position.

Images