Robot operation self-correction method and system based on multi-dimensional data space

By importing robot parameters and modeling systems in multi-dimensional data space, performing rehearsal and feedback steps, and adjusting instruction parameters, the problems of collision and dumping when robots grab targets are solved, improving the crawling success rate and accuracy, and reducing costs , enhances the adaptability of the robot.

CN118254190BActive Publication Date: 2024-11-19SHANDONG HUMAN-MASCH XIEHE INTELLIGENT TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410682996.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-30
Publication Date
2024-11-19
Estimated Expiration
2044-05-30

AI Technical Summary

Technical Problem

When existing robots capture targets, especially when target positions change or obstacles exist, they are prone to collisions or dumping, resulting in failure in grabbing. The training cost of big data algorithms is high, which limits the popularity and widespread practicality of robots.

Method used

Using a robot operation self-correction method based on multidimensional data space, by establishing a multidimensional data space, importing robot parameters and modeling systems, obtaining task instructions, performing rehearsal and feedback steps, judging and adjusting instruction parameters or paths, performing operation actions, and Issue a warning if necessary to improve the success rate and accuracy of crawling.

Benefits of technology

It improves the success rate and accuracy of the robot in a changing environment, reduces the damage and collision of the target, reduces the cost of training and reasoning, and enhances the adaptability and flexibility of the robot.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118254190B_ABST
    Figure CN118254190B_ABST
Patent Text Reader

Abstract

The present application relates to the technical field of robot operation correction, and in particular to a robot operation self-correction method and system based on multidimensional data space, wherein the method includes establishing a multidimensional data space, obtaining a task instruction step, a target acquisition step, a rehearsal step, a feedback step, an instruction execution step and an alarm step; wherein the system includes: a multidimensional data space, a task instruction acquisition module, a target acquisition module, a rehearsal module, a feedback module, an instruction execution module and an alarm module. The present application has the effect of improving the success rate of target capture.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of robot operation correction, and in particular to a robot operation self-correction method and system based on multi-dimensional data space. Background Art

[0002] At present, with the widespread use of machine learning and big data model algorithms in the field of robot research and development, robot behavior control has become more simple and efficient.

[0003] However, at the same time, the inherent characteristics of big data algorithms are gradually restricting the further development of robots. The huge amount of data makes the training cost remain high, and the dependence on machine performance becomes more and more serious, which makes it difficult for robot-related products to be popularized and widely used.

[0004] However, if the grasping target changes or the grasping target position changes, when the pre-set instruction set is executed again, there is a possibility that the robot touches the target, causing damage to the target or causing it to fall, resulting in failure to grasp the target. Summary of the invention

[0005] In order to improve the success rate of target grasping, the present application provides a robot operation self-correction method and system based on multi-dimensional data space.

[0006] In a first aspect, the present application provides a robot operation self-correction method based on multidimensional data space, which adopts the following technical solution:

[0007] A robot operation self-correction method based on multi-dimensional data space comprises the following steps:

[0008] Establish multidimensional data space: Establish multidimensional data space through multidimensional data space system, form multidimensional space coordinate system, and import the robot's own parameters and modeling system into the multidimensional data space;

[0009] Get task instructions: transmit the robot operation instruction set task package to the robot;

[0010] Target acquisition: Get the position of the target to be grasped or the robot's arrival position, and calculate the corresponding spatial distance;

[0011] Preview: simulate the process of running the robot's operation according to the preset operation instruction set, and judge whether there is an error instruction in the execution result of each step of the instruction during the execution of the operation. If so, execute the feedback step, otherwise, execute the instruction execution step;

[0012] Feedback: Determine whether the error instruction is controllable. If so, fine-tune the instruction parameters or the path to be followed and execute the instruction execution step; otherwise, execute the warning step;

[0013] Instruction execution: The robot performs operations according to the operation instruction set;

[0014] Alert: Feedback to developers and issue warnings.

[0015] By adopting the above technical solution, a multidimensional data space is established, and the robot parameters are imported into the multidimensional data space. In combination with the multidimensional coordinate system, the coordinates of the robot are obtained, and then the task to be performed is obtained. Then, the position of the target in the task is obtained, and then the coordinate position of the target is calculated in combination with the robot's own coordinate position, and the corresponding spatial distance is calculated. Then, a simulation operation is performed through a preset instruction set to determine whether there is an erroneous instruction. When there is an erroneous instruction, it is determined whether it is a controllable erroneous instruction. If so, the instruction parameters or the travel path are fine-tuned, and then the robot's operation action is executed. If not, If there is an erroneous instruction, the robot will execute the operation action. If the erroneous instruction is uncontrollable, an alarm will be issued and feedback will be given to the developer. Through the setting of the rehearsal step and the feedback step, before the robot executes the operation action, it will first conduct simulation training according to the location of the target. According to whether the simulation result is good, it will be determined whether the operation action will be executed. If the operation action cannot be executed, it will be determined whether it can be corrected, thereby improving the success rate of grasping the target. At the same time, the grasping action can be modified and adjusted by correction, thereby improving the accuracy of the operation action, reducing collision with the target, and reducing the tipping and damage of the target.

[0016] Moreover, compared with the existing robot control systems, whether it is the pure mathematical algorithm of traditional manual control or the intelligent algorithm combining the emerging machine learning and big data model training, the algorithms they run are either already solidified and lack real-time performance without exception. An effective mechanism for the robot to adjust and think by itself can be set through the settings of rehearsal steps and feedback steps, so that the robot can make appropriate adjustments to an unknown situation at any time, even if it is only a slight difference, thereby improving the success rate of existing operating strategies. Moreover, through the above settings, the robot can adjust and think by itself, and can meet the requirements and accuracy from 90 to 100, facilitate the execution of adaptive high-precision operating actions, reduce the degree of human participation, and reduce costs.

[0017] Optionally, after the instruction execution step, a training step is also provided;

[0018] Training: Record the thinking and execution process of fine-tuning instruction parameters or travel paths as training data, and conduct enhanced training at the appropriate time.

[0019] By adopting the above technical solution, through the modified instruction parameters, after execution is completed, in idle time, the robot can be controlled through the rehearsal steps and instruction execution steps to conduct simulation training, and during the execution process, the instructions can be modified to enhance the robot's grasping success rate.

[0020] Optionally, an obstacle intervention step is further provided after the target acquisition step;

[0021] Obstacle intervention: including obstacle acquisition steps, intervention blocking steps and adjustment simulation steps;

[0022] Obstacle acquisition: retrieve and analyze the obstacle position through sensor equipment, and determine the obstacle coordinates through its own position;

[0023] Intervention obstacle judgment: judge whether there is a collision with an obstacle. If so, execute the adjustment simulation step; otherwise, execute the preview step;

[0024] Adjust simulation: simulate and adjust the instruction set to see if it can reach the target coordinate position. If so, execute the preview step with the adjusted instruction set. Otherwise, execute the alarm step.

[0025] By adopting the above technical solution, when there is an obstacle between the target and the robot, the position of the obstacle is retrieved and analyzed by the sensor equipment, and then the obstacle coordinates are determined by the robot's own coordinates. Then, it is determined whether a collision with the obstacle will occur during the grasping process executed according to the set instruction set. If there is a collision, the adjustment simulation step is executed to simulate the adjustment of the instruction set to determine whether the target coordinate position can be reached. If so, the rehearsal step is executed to determine again whether the execution of the instruction can be completed. Otherwise, the alarm step is executed. By setting the obstacle intervention step, it is determined whether the existence of the obstacle affects the execution of the operation action, and if it does, it is determined whether the operation action can be performed to complete the task after correction and adjustment. By setting the above two steps, the grasping success rate is improved, and at the same time, adaptive adjustments can be made according to the position of the obstacle, making the robot more flexible and real-time.

[0026] Optionally, after the target acquisition step, a difficulty determination step and a simplification step are further provided;

[0027] Difficulty judgment: Get the complexity of the task instructions and judge whether it is simple. If so, execute the simplified steps; otherwise, execute the preview steps;

[0028] Simplification: Virtualize the robot itself as a coordinate point in the multidimensional data space, calculate the motion parameters of the coordinate point when it reaches the specified coordinate, and then perform the rehearsal steps.

[0029] By adopting the above technical solution, the complexity of the task is judged according to the target position and the task instructions. If the complexity is simple, the robot itself is virtualized as a coordinate point in the data space, and the motion parameters of the coordinate point when it reaches the specified coordinate are calculated. Then, the rehearsal step is executed, and after there are no erroneous instructions, the instruction execution step is executed. By simplifying the setting of steps, the computing resources required for training and reasoning can be significantly reduced, the training and reasoning speed of the model can be accelerated, the efficiency of task operations can be greatly improved, and computing power can be saved, reducing data storage costs.

[0030] Optionally, a segmented execution step is further provided between the difficulty determination step and the preview step;

[0031] Segmented execution: including task instruction decomposition steps, segmented simplification steps and precision execution steps;

[0032] Instruction decomposition: divide the task instructions into simple stages and complex stages according to the order of execution;

[0033] Segmented simplification: For simple stage task instructions, execute simplification steps, and for complex stage task instructions, execute precision execution steps;

[0034] Precision execution: The operations in the complex stage are refined, the motion parameters of the robot when it reaches the specified coordinates are calculated, and then the rehearsal steps are performed.

[0035] By adopting the above technical solution, when the difficulty of executing the instructions is high, the task instructions are carried out in stages. For example, in the initial moving stage, the environment is relatively simple, the complexity is low, and the precision requirement is not high. At this time, the robot can virtualize itself as a coordinate point in the multidimensional data space, and calculate the motion parameters when the coordinate point reaches the specified coordinate; when the robot approaches the target, the task complexity becomes higher and the precision requirement increases. The operation at this time needs to be refined, and the robot can improve the simulation accuracy of the local area; through the above settings, the combination of simplicity and complexity can greatly improve the task calculation efficiency while ensuring the task execution accuracy, and at the same time save computing power and storage costs.

[0036] Optionally, in the instruction execution step, when the robot executes the instructions in the operation instruction set task package, unoccupied coordinate points are marked as 0, and occupied coordinate points are marked as 1.

[0037] By adopting the above technical solution, in the process of executing task instructions, occupied coordinate points are marked as 1 and unoccupied coordinate points are marked as 0 to distinguish them, and the coordinates occupied by obstacles and targets can also be marked to improve the accuracy of task execution and reduce collisions.

[0038] Optionally, after the instruction execution step, an execution statistics step is also provided;

[0039] Execution statistics: Collect statistics on the data set of robot task execution and perform model simulation training when the robot is not performing grasping work.

[0040] By adopting the above technical solution, after the grasping is successful, the data set of the robot task execution is statistically analyzed, and simulation training is performed when the robot is not performing grasping work. In the simulation process, corrections and optimizations are continuously performed to improve the grasping accuracy.

[0041] In a second aspect, the present application provides a robot operation self-correction system based on multi-dimensional data space, which adopts the following technical solution:

[0042] A robot operation self-correction system based on multi-dimensional data space, including the following modules:

[0043] Multidimensional data space: Through the multidimensional data space system, a multidimensional data space is established to form a multidimensional space coordinate system, which is used to generate the three-dimensional coordinates of the robot by combining the robot parameters with the modeling system;

[0044] Obtaining task instruction module: The input end is connected to the output end of the multidimensional data space, and is used to obtain the task instructions manually or directly issued by the operator, and transmit the robot operation instruction set task package to the robot;

[0045] Target acquisition module: The input end is connected to the output end of the multidimensional data space and the output end of the task instruction acquisition module, and is used to obtain the position of the target to be grasped or the robot's arrival position, and calculate the corresponding spatial distance;

[0046] Rehearsal module: the input end is connected to the output end of the task instruction acquisition module and the output end of the target acquisition module, and is used to simulate the operation according to the preset instruction set;

[0047] Feedback module: The input end is connected to the output end of the preview module. When an erroneous instruction occurs in the preview module, it is used to determine whether the erroneous instruction is controllable;

[0048] Instruction execution module: The input end is connected to the output end of the preview module and the output end of the feedback module. When the simulation of the preview module is normal and the erroneous instructions output by the feedback module are controllable and corrected, the robot executes the action according to the instruction;

[0049] Alarm module: The input end is connected to the output end of the feedback module, and is used to provide feedback to the developer and issue a warning when the feedback module outputs an error instruction that is uncontrollable.

[0050] By adopting the above technical solution, a spatial coordinate system is established through a multi-dimensional data space, and then combined with the coordinate position of the robot, the task instruction is sent to the robot through the task instruction acquisition module, and then the target acquisition module is executed to acquire the position of the target and calculate the corresponding spatial distance. Then the rehearsal module is executed according to the task instruction. The rehearsal module simulates the operation of the robot, and then when it can run correctly, the instruction execution module is executed. If there is an erroneous instruction, the feedback module is executed to determine whether it can be corrected. If it can be corrected, the instruction execution module is executed after the correction is completed. If the execution of the task instruction cannot be completed after the correction, the alarm module is executed to issue a warning; through the setting of the above modules, the robot's action execution is achieved through rehearsal, instruction correction and alarm. After thinking, the execution of unfinished actions is reduced, thereby reducing damage to the target; and through instruction correction and simulation, the adaptability in unfamiliar environments or unknown situations is improved, thereby improving the robot's grasping rate.

[0051] In summary, the present application includes at least one of the following beneficial technical effects:

[0052] 1. Through the setting of the rehearsal step and the feedback step, before the robot performs the operation, it first conducts simulation training according to the location of the target. It determines whether to perform the operation according to whether the simulation result is good. If the operation cannot be performed, it determines whether it can be corrected, thereby improving the success rate of grasping the target. At the same time, the grasping action is modified and adjusted by correction, thereby improving the accuracy of grasping, reducing collision with the target, and reducing the tipping and damage of the target;

[0053] 2. By simplifying task instructions in stages and combining simplified and traditional methods, we can significantly improve task computing efficiency while ensuring task execution accuracy, while saving computing power and storage costs;

[0054] 3. In the process of executing task instructions, occupied coordinate points are marked as 1 and unoccupied coordinate points are marked as 0 to distinguish them. The coordinates occupied by obstacles and targets can also be marked to improve the accuracy of task execution and reduce collisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a flow chart of the robot operation self-correction method in Example 1 of the present application;

[0056] Figure 2 is a flowchart of the training steps in Example 2 of the present application;

[0057] Figure 3 It is a flowchart of the instruction decomposition step to the precision execution step in Example 2 of the present application;

[0058] Figure 4 is a system diagram of a robot-operated self-correcting system in an embodiment of the present application. DETAILED DESCRIPTION

[0059] The following combination Figures 1 to 4 This application is described in further detail.

[0060] This embodiment discloses a robot operation self-correction method based on multi-dimensional data space.

[0061] Example 1: Reference Figure 1 , a robot operation self-correction method based on multi-dimensional data space, comprising the following steps:

[0062] S1: Establish multidimensional data space: Establish multidimensional data space through multidimensional data space system, form multidimensional space coordinate system, and import the robot's own parameters and modeling system into the multidimensional data space;

[0063] Specifically, the multi-dimensional space coordinate system can use three-dimensional coordinates to import the robot's own parameters and modeling system into the multi-dimensional data space to form a robot model and the robot's spatial coordinate occupancy, where the robot's own parameters include occupied space, size, position of each robot component, and outer edge dimensions.

[0064] S2: Get task instructions: transmit the robot operation instruction set task package to the robot;

[0065] Specifically, the task instructions can be sent manually by the lower computer, or the sending time or conditions can be automatically set through the network, including the type of instruction (including: grabbing, touching, moving, etc.), the goal of the task, the parameters of the operation of related components, etc.

[0066] S3: Target acquisition: Get the position of the target to be grasped or the robot's arrival position, and calculate the corresponding spatial distance;

[0067] Specifically, the position of the target to be grasped is determined according to the task instructions, and then the coordinate position of the target is determined according to its own coordinates, or the position to be reached by the robot is determined according to the task instructions, and the corresponding spatial distance is calculated. The spatial distance can be calculated using Euclidean distance, Manhattan distance, Chebyshev distance, Minkowski distance, etc.

[0068] S4: Preview: simulate the process of running the robot's operation according to the preset operation instruction set, and judge whether there is an error instruction in the execution result of each step of the instruction during the execution of the operation action. If so, execute the feedback step, otherwise, execute the instruction execution step;

[0069] S5: Feedback: Determine whether the error instruction is controllable. If so, fine-tune the instruction parameters or the path to be traveled, and execute the instruction execution step; otherwise, execute the alarm step;

[0070] S6: Instruction execution: The robot performs operation actions according to the operation instruction set;

[0071] S7: Warning: Feedback to developers and issue warnings.

[0072] Specifically, during the rehearsal, the execution of a single instruction is simulated through the instruction set to determine whether the robot can reach or complete the instruction. When there is an error in the instruction or the robot cannot reach the point or cannot perform a certain operation at the point, the parameters of the instruction or the path of travel are fine-tuned and simulated again. If it can be executed, the instruction execution step is executed. If the grasping to achieve the target cannot be performed after fine-tuning, the alarm step is executed and feedback is given to the developer for adjustment training. Among them, when the width of the target object is greater than the grasping width of the robot or is exactly equal to the grasping width or slightly smaller than the grasping width, the robot's accuracy requirements are very high, and after an erroneous instruction occurs, it needs to be accurately corrected, otherwise the operation cannot be performed, and there is intervention from other situations after the correction, and the grasping still cannot be completed. At this time, only the alarm step can be performed.

[0073] The implementation principle of Example 1 of the present application is as follows: after establishing a multidimensional data space, an operation instruction set task package is issued to the robot. The robot obtains the position of the target through a sensor device, and combines its own coordinates to obtain the coordinate position of the target, and then obtains the spatial distance through calculation. Then, a rehearsal simulation is performed according to the instruction set task package to determine whether there are erroneous instructions. When there are erroneous instructions, a feedback step is executed to modify the instructions and perform simulations. If the target can be grasped, the instruction execution step is executed to perform the grasping operation. If the modified instructions still cannot complete the grasping of the target, the alarm step is executed to stop the grasping and provide feedback to the developer.

[0074] Example 2: Reference Figure 2 , the difference between this embodiment and embodiment 1 is that after the instruction execution step, a training step S8 is also provided;

[0075] S8: Training: Record the thinking and execution process of fine-tuning instruction parameters or travel paths as training data, and conduct enhanced training at the appropriate time.

[0076] Specifically, after the instruction is modified, the execution record after thinking is used as training data. When the robot is not performing grasping work, enhanced training simulation is carried out. By simulating the executable training data, the robot is trained so that the next time it grasps the target of the same type and position, it can be executed accurately and reduce collisions with the target object. At the same time, after training the robot, it is convenient to improve the robot's subsequent operating efficiency.

[0077] Reference Figure 3 In other embodiments, an obstacle intervention step S9 is further provided after the target acquisition step;

[0078] S9: obstacle intervention: including obstacle acquisition step S9-1, intervention and obstruction step S9-2 and adjustment simulation step S9-3;

[0079] S9-1: Obstacle acquisition: retrieve and analyze the obstacle position through sensor equipment, and determine the obstacle coordinates through its own position;

[0080] S9-2: Intervention obstacle judgment: judge whether there is a collision with an obstacle, if so, execute the adjustment simulation step, otherwise, execute the preview step;

[0081] S9-3: Adjustment simulation: simulate and adjust the instruction set to see if it can reach the target coordinate position. If so, execute the preview step with the adjusted instruction set. Otherwise, execute the alarm step.

[0082] Specifically, when there are obstacles, the obstacles include other target objects to be grasped, racks or other robots. When there are obstacles, the position of the obstacle is retrieved and analyzed through the sensor equipment, and then the position of the obstacle and the position of the obstacle relative to the target object are determined through its own coordinates. Then, the instruction set is used to determine whether the robot will collide with the obstacle when executing the instruction. If there is a possibility of a collision, the instruction set is adjusted to simulate whether it can reach the target coordinate position and execute the grasping of the task target based on the adjusted instruction set. If so, the instruction execution step is performed according to the adjusted instruction set. If not, the alarm step is performed to allow the developer to train the robot or adjust the parameters.

[0083] In other embodiments, after the target acquisition step, a difficulty determination step S10 and a simplification step S11 are further provided;

[0084] S10: Difficulty judgment: Obtain the complexity of the task instruction and judge whether it is simple. If so, execute the simplification step; otherwise, execute the preview step;

[0085] S11: Simplification: Virtualize the robot itself as a coordinate point in the multidimensional data space, calculate the motion parameters of the coordinate point when it reaches the specified coordinate, and then perform the preview step.

[0086] Specifically, the complexity of the task instructions is obtained, where the complexity needs to take into account information such as environmental complexity, precision requirements, and task difficulty. When the environmental complexity is simple, the precision requirements are low, and the task difficulty is low, it belongs to the simple type. The simplification step is executed, and the robot itself is virtualized as a coordinate point in the multidimensional data space. Then, the motion parameters of the coordinate point when it reaches the execution coordinate are calculated. After simulation of the rehearsal step, the instruction execution step is executed after it is correct.

[0087] In other embodiments, a segmented execution step S12 is further provided between the difficulty determination step and the preview step;

[0088] S12: Segment execution: including task instruction decomposition step S12-1, segment simplification step S12-2 and precision execution step S12-3;

[0089] S12-1: Instruction decomposition: Divide the task instructions into simple stages and complex stages according to the order of execution;

[0090] S12-2: Segmented simplification: For simple stage task instructions, perform simplification steps, and for complex stage task instructions, perform precision execution steps;

[0091] S12-3: Precision execution: Refine the operations in the complex stage, calculate the motion parameters when the robot coordinates reach the specified coordinates, and then execute the preview steps.

[0092] Specifically, when the difficulty is judged to be not simple, the task instructions can be decomposed into simple stages and complex stages. The simple stage is a process step with a relatively simple environment and low precision requirements, and the complex stage can be a stage for executing crawling or a stage with a relatively complex environment. At this time, the segmented simplification steps are executed, and the task instructions of the simple stage are executed in the simplified steps. For the task instructions of the complex stage, they are refined, and then the rehearsal steps are executed. After the rehearsal simulation is correct, the instruction execution steps are executed again. The difference from the traditional simulation operation program is that the model in the multidimensional data space is not a fixed size, but can be simplified in proportion to the accuracy of the simulation to save resources.

[0093] In other embodiments, in the instruction execution step, when the robot executes the instructions in the operation instruction set task package, the unoccupied coordinate points are marked as 0, and the occupied coordinate points are marked as 1.

[0094] Specifically, when the instruction is executed and when the robot coordinates, obstacle coordinates, and target coordinates are obtained, the occupied points can be marked as 1 and the unoccupied points can be marked as 0, which is convenient for judging whether there are overlapping points, whether the robot has moved to the right position, or whether the robot has moved according to the instruction.

[0095] After the instruction execution step, an execution statistics step S13 is also provided;

[0096] S13: Execution statistics: Count the data set of robot task execution and perform model simulation training when the robot is not performing grasping work.

[0097] Specifically, all data sets of the robot's task execution are counted, and then model simulation training is performed when the robot is not performing grasping work. Especially for grasping actions with higher complexity, adjustments can be run multiple times to facilitate accurate grasping.

[0098] Specifically, taking the operation of a robotic arm as an example, assuming that the upper half of the robotic arm is 20 cm, the lower half is 30 cm, and the diameters are 5 cm and 6 cm respectively, if the traditional modeling method is used, it will consume a lot of resources to simulate the motion trajectory of the robotic arm. However, in multidimensional data space, these redundant data can be ignored, and only the upper half of the robotic arm needs to be virtualized as 2 coordinate points, and the lower half as 3. In this way, most of the redundant calculations are reduced, and the simulation process is turned into an interactive movement between two line segments. Therefore, this method allows the robot to complete the corresponding calculations on equipment with very limited performance, thereby greatly reducing the cost of robot production and energy consumption during use.

[0099] The implementation principle of Example 2 of the present application is: when the modified instructions can complete the grasping of the target through simulation, and can complete the grasping of the target through execution, the thinking records and execution parameters can be saved, and the training work can be performed when the robot is idle to improve the robot's action execution efficiency.

[0100] When there is an obstacle between the grasping target and the robot, the position of the obstacle is obtained and simulated to determine whether there is overlap in coordinate points. If so, an adjustment simulation is performed to determine whether the target can be grasped after the parameters are adjusted. If so, the instruction execution step is executed, otherwise, the alarm step is executed.

[0101] Before executing the rehearsal, it is necessary to judge the difficulty. When the difficulty is judged to be simple, the simplification step is executed to simplify the robot's coordinates into coordinate points, and then the rehearsal is performed. When the grasping execution can be completed after the rehearsal, the instruction execution step is executed. When it is judged to be not simple, the task instructions can be decomposed into simple stages and complex stages. The simple stage is a process step with a relatively simple environment and low precision requirements. The complex stage can be a stage for executing grasping or a stage with a relatively complex environment. At this time, the segmented simplification step is executed to execute the simplification step for the task instructions of the simple stage, and the task instructions of the complex stage are refined, and then the rehearsal step is executed. After the rehearsal simulation is correct, the instruction execution step is executed again.

[0102] In the multidimensional data space, each coordinate point is marked as 1 when occupied, and marked as 0 otherwise, which makes it easier to determine the position of the robot.

[0103] The embodiment of the present application also discloses a robot operation self-correction system based on multi-dimensional data space.

[0104] Reference Figure 4 , a robot operation self-correction system based on multi-dimensional data space, including the following modules:

[0105] Multidimensional data space: Through the multidimensional data space system, a multidimensional data space is established to form a multidimensional space coordinate system, which is used to generate the three-dimensional coordinates of the robot by combining the robot parameters with the modeling system;

[0106] Specifically, the construction of the multidimensional data space relies on the multidimensional data space system and the modeling system. By establishing a multidimensional coordinate system in the multidimensional data space and then pouring the robot parameters into the modeling system, the multidimensional coordinates of the robot are generated.

[0107] Obtaining task instruction module: The input end is connected to the output end of the multidimensional data space, and is used to obtain the task instructions manually or directly issued by the operator, and transmit the robot operation instruction set task package to the robot;

[0108] Target acquisition module: The input end is connected to the output end of the multidimensional data space and the output end of the task instruction acquisition module, and is used to obtain the position of the target to be grasped or the robot's arrival position, and calculate the corresponding spatial distance;

[0109] Specifically, the grasping instruction set task package is transmitted to the robot by manual delivery and automatic delivery through the network, and then the target's position information is obtained through the task instructions, and the spatial distance between the two is calculated through its own coordinates.

[0110] Rehearsal module: the input end is connected to the output end of the task instruction acquisition module and the output end of the target acquisition module, and is used to simulate the operation according to the preset instruction set;

[0111] Feedback module: The input end is connected to the output end of the preview module. When an erroneous instruction occurs in the preview module, it is used to determine whether the erroneous instruction is controllable;

[0112] Instruction execution module: The input end is connected to the output end of the preview module and the output end of the feedback module. When the simulation of the preview module is normal and the erroneous instructions output by the feedback module are controllable and corrected, the robot executes the action according to the instruction;

[0113] Alarm module: The input end is connected to the output end of the feedback module, and is used to provide feedback to the developer and issue a warning when the feedback module outputs an error instruction that is uncontrollable.

[0114] Specifically, a simulation calculation of robot grasping is performed according to the instruction set to determine whether there are erroneous instructions. If there are erroneous instructions, it is determined whether the erroneous instructions can be modified by executing parameter modification, and then it is determined whether the modified instruction set can complete the grasping of the target. When the grasping of the target can be completed, the instruction execution module is executed to allow the robot to grasp and execute according to the current modified instruction set.

[0115] Obstacle intervention module: The input end is connected to the output end of the target acquisition module to determine whether there is an obstacle and whether the obstacle interferes with the robot's grasping. If there is interference, it determines whether the target can be grasped by correcting the parameters.

[0116] Training module: The input end is connected to the output end of the feedback module and is used to train the robot using the modified data when the robot is not working.

[0117] Difficulty judgment module: The input end is connected to the output end of the task instruction acquisition module, and is used to obtain the difficulty of the task instruction and judge the complexity of the task instruction.

[0118] Simplification module: The input end is connected to the output end of the difficulty judgment module, and is used to simplify the robot coordinates into coordinate points after the difficulty judgment module determines that the robot is simple, and then calculate the spatial distance between the robot and the target.

[0119] Segmented execution module: The input end is connected to the output end of the difficulty judgment module. It is used to segment the task instructions after the difficulty judgment module determines that it is not simple. The simplified module is executed in the simple stage, and the refined operation is executed in the complex stage.

[0120] Execution statistics module: The input end is connected to the output end of the instruction execution module, which is used to count the number of times the instruction is executed and to train the robot model during the robot's idle time.

[0121] The implementation principle of a robot operation self-correction system based on multi-dimensional data space in an embodiment of the present application is as follows: a robot operation instruction set task package is sent to the robot manually or through the network, the robot obtains its own coordinates through the multi-dimensional data space, and then obtains the coordinates of the target through the target acquisition module, and calculates the spatial distance, and then executes the rehearsal module to perform a rehearsal simulation on the execution in the instruction set task package to determine whether there are erroneous instructions. When there are erroneous instructions, it is determined whether the robot's grasping task can be completed after correction by the feedback module. If so, the instruction execution module is executed to control the robot to perform the grasping work. Otherwise, the alarm module is executed to provide feedback to the developer and issue a warning.

[0122] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A robot operation self-correction method based on multidimensional data space, characterized in that: The following steps are involved: Establish multidimensional data space: Establish multidimensional data space through multidimensional data space system, form multidimensional space coordinate system, and import the robot's own parameters and modeling system into the multidimensional data space; Get task instructions: transmit the robot operation instruction set task package to the robot; Target acquisition: Get the position of the target to be grasped or the robot's arrival position, and calculate the corresponding spatial distance; Preview: simulate the process of running the robot's operation according to the preset operation instruction set, and judge whether there is an error instruction in the execution result of each step of the instruction during the execution of the operation. If so, execute the feedback step, otherwise, execute the instruction execution step; Feedback: Determine whether the error instruction is controllable. If so, fine-tune the instruction parameters or the path to be followed and execute the instruction execution step; otherwise, execute the warning step; Instruction execution: The robot performs operations according to the operation instruction set; Alert: Feedback to developers and issue warnings; An obstacle intervention step is also provided after the target acquisition step; Obstacle intervention: including obstacle acquisition steps, intervention blocking steps and adjustment simulation steps; Obstacle acquisition: retrieve and analyze the obstacle position through sensor equipment, and determine the obstacle coordinates through its own position; Intervention obstacle judgment: judge whether there is a collision with an obstacle. If so, execute the adjustment simulation step; otherwise, execute the preview step; Adjustment simulation: simulate and adjust the instruction set to determine whether the target coordinate position can be reached. If so, the rehearsal step is executed with the adjusted instruction set. Otherwise, the alarm step is executed. After the target acquisition step, there are also a difficulty judgment step and a simplification step; Difficulty judgment: Get the complexity of the task instructions and judge whether it is simple. If so, execute the simplified steps; otherwise, execute the preview steps; Simplification: Virtualize the robot itself as a coordinate point in the multidimensional data space, calculate the motion parameters of the coordinate point when it reaches the specified coordinate, and then perform the rehearsal steps; A segmented execution step is also provided between the difficulty determination step and the preview step; Segmented execution: including task instruction decomposition steps, segmented simplification steps and precision execution steps; Instruction decomposition: divide the task instructions into simple stages and complex stages according to the order of execution; Segmented simplification: For simple stage task instructions, execute simplification steps, and for complex stage task instructions, execute precision execution steps; Precision execution: The operations in the complex stage are refined, the motion parameters of the robot when it reaches the specified coordinates are calculated, and then the rehearsal steps are performed.

2. The robot operation self-correction method based on multidimensional data space according to claim 1, characterized in that: After the instruction execution step, a training step is also provided; Training: The thinking and execution process of fine-tuning instruction parameters or travel paths is recorded as training data, and enhanced training is performed when the robot is not performing grasping work.

3. The robot operation self-correction method based on multidimensional data space according to any one of claims 1 to 2, characterized in that: In the instruction execution step, when the robot executes the instructions in the operation instruction set task package, the unoccupied coordinate points are marked as 0, and the occupied coordinate points are marked as 1.

4. The robot operation self-correction method based on multidimensional data space according to claim 1, characterized in that: After the instruction execution step, an execution statistics step is also provided; Execution statistics: Collect statistics on the data set of robot task execution and perform model simulation training when the robot is not performing grasping work.

5. A system for implementing the robot operation self-correction method based on multidimensional data space as described in any one of claims 1 to 4, characterized in that: Includes the following modules: Multidimensional data space: Through the multidimensional data space system, a multidimensional data space is established to form a multidimensional space coordinate system, which is used to generate the three-dimensional coordinates of the robot by combining the robot parameters with the modeling system; Obtaining task instruction module: The input end is connected to the output end of the multidimensional data space, and is used to obtain the task instructions manually or directly issued by the operator, and transmit the robot operation instruction set task package to the robot; Target acquisition module: The input end is connected to the output end of the multidimensional data space and the output end of the task instruction acquisition module, and is used to obtain the position of the target to be grasped or the robot's arrival position, and calculate the corresponding spatial distance; Obstacle intervention module: The input end is connected to the output end of the target acquisition module to determine whether there is an obstacle and whether the obstacle interferes with the robot's grasping. If there is interference, it determines whether the target can be grasped by correcting the parameters. Rehearsal module: the input end is connected to the output end of the task instruction acquisition module, the target acquisition module and the output end of the obstacle intervention module, and is used to simulate operation according to a preset instruction set or a revised instruction set; Feedback module: The input end is connected to the output end of the preview module. When an erroneous instruction occurs in the preview module, it is used to determine whether the erroneous instruction is controllable; Instruction execution module: The input end is connected to the output end of the preview module and the output end of the feedback module. When the simulation of the preview module is normal and the erroneous instructions output by the feedback module are controllable and corrected, the robot executes the action according to the instruction; Alarm module: The input end is connected to the output end of the feedback module, and is used to provide feedback to the developer and issue a warning when the feedback module outputs an error instruction that is uncontrollable.