Substation oil sample collection robot intelligent pre-performance decision method based on digital twinning

Abstract

The application discloses a substation oil sample collection robot intelligent pre-performance decision method based on digital twinning, and belongs to the technical field of substation robot control. The method comprises the following steps: pre-performing a task according to an oil sample collection robot digital twinning system to obtain an optimal strategy of the robot; executing the task according to the optimal strategy by means of an entity robot and performing closed-loop reconstruction strategy; and evaluating after the task executed by the entity robot is completed. The application performs competitive pre-performance of a task by means of the oil sample collection robot digital twinning system, and filters out an optimal strategy before the task is executed, thereby eliminating risks such as collision, equipment damage and oil leakage caused by algorithm inadaptation from the source, realizing "zero risk" debugging, reducing debugging risk and operation and maintenance cost, and solving the problems of lack of feasibility decision before task execution and lack of closed-loop reconstruction decision based on virtual-real feedback in task execution in the prior art.

Description

Technical Field

[0001] This invention belongs to the field of substation robot control technology, specifically relating to an intelligent pre-simulation decision-making method for substation oil sample collection robots based on digital twins. Background Technology

[0002] Transformers are core equipment in power systems, and their insulating oil needs to be sampled and analyzed regularly to monitor their operating status. Traditional manual sampling methods pose significant safety hazards, such as the potential for explosive gas mixtures to be generated under abnormal transformer operating conditions, threatening the lives of operators. Furthermore, manual operation can easily lead to oil sample contamination, affecting the accuracy of testing.

[0003] To improve operational automation and safety, oil sampling robots have emerged. However, existing robot systems still face significant challenges in deployment and application: the development and verification of their algorithms heavily rely on real substation environments, resulting in long commissioning cycles and uncontrollable safety risks. More critically, current technology lacks an intelligent decision-making mechanism based on digital twins. Specifically: On the one hand, existing technologies lack the ability to make feasibility decisions before task execution: existing technologies typically employ a fixed set of navigation or planning algorithms, which are difficult to adapt to the complex and ever-changing operating conditions within substations (such as changing obstacle distribution and narrow passages). This invention aims to address how to utilize digital twins to verify multiple different algorithms in parallel during the pre-simulation phase, based on the characteristics of the current scenario, thereby determining the most suitable, efficient, or safest algorithm strategy and avoiding the inefficiency or risks associated with a single algorithm.

[0004] On the other hand, existing technologies lack closed-loop reconstruction decisions based on virtual and real feedback during task execution: when existing robots encounter difficulties in real-world execution (such as path planning failure or failure to reach the destination for a long time), they can often only report errors and stop, lacking the ability to recover autonomously. Summary of the Invention

[0005] To address the aforementioned shortcomings in existing technologies, the intelligent pre-simulation decision-making method for substation oil sample collection robots based on digital twins provided by this invention solves the problems of lack of feasibility decisions before task execution and lack of closed-loop reconstruction decisions based on virtual and real feedback during task execution.

[0006] To achieve the aforementioned objectives, the technical solution adopted by this invention is: an intelligent pre-simulation decision-making method for substation oil sample collection robots based on digital twins, comprising the following steps: S1. Construct a digital twin system for an oil sample collection robot; S2. Based on the digital twin system of the oil sample collection robot, the task execution is rehearsed to obtain the optimal strategy of the robot; S3. The physical robot executes the task according to the optimal strategy and performs closed-loop reconfiguration of the strategy; S4. Conduct an evaluation after the physical robot has completed its task.

[0007] Furthermore, S1 includes the following sub-steps: S11. Based on the measured dimensions in the station, establish a high-fidelity digital twin including the scene, sampling points, obstacles and robots, and complete the scale calibration and kinematic link configuration; S12. Configure a communication module on the digital twin to enable two-way data interaction between the digital twin and the physical robot; S13. Preset a security rule base and initialize a knowledge base. The knowledge base stores the mapping relationship between different scenario features and the optimal algorithm combination.

[0008] Furthermore: In S11, the specific method for establishing a high-fidelity digital twin is as follows: Using 3ds Max and SolidWorks 3D modeling software, 3D geometric modeling of the scene, sampling points, obstacles, and robot was performed. The 3D model was exported in .stl format. 3ds Max was used to convert the 3D model format, unnecessary parts were removed for lightweight processing, and abnormal coordinate axes were corrected to obtain a 3D model in .fbx format. This model was then imported into the Unity 3D digital twin platform to complete the creation of a high-fidelity digital twin.

[0009] Furthermore, S2 includes the following sub-steps: S21. Based on the constructed digital twin system of the oil sample collection robot, in the digital twin environment, the scene feature vector of the current task is identified according to the data of the digital twin and virtual sensors, and a preset number of algorithm combinations are selected from the knowledge base according to the scene feature vector. S22. Use a combination of algorithms in parallel or serial mode to perform task rehearsals. Each rehearsal drives the digital twin to execute the task from the starting point to the sampling point and records the performance indicators of the rehearsal. The performance indicators include the total task time, trajectory smoothness, minimum safe distance, peak joint torque, and whether safety rules are triggered. S23. Based on the preset performance index weights, perform weighted scoring on each performance index, and take the algorithm combination corresponding to the highest score of the pre-performance as the robot's optimal strategy.

[0010] Furthermore, in S22, the method for rehearsing the task includes the following steps: The specific method for task pre-simulation is as follows: a hierarchical planning strategy is adopted to plan feasible trajectories for the chassis motion and robotic arm control of the oil sample robot, and competitive pre-simulation and evaluation are carried out based on scene characteristics to obtain the optimal feasible trajectory with no collision. The competitive pre-simulation method is as follows: the selected algorithm combination is added to the competition pool, and the algorithm combination is switched according to scene characteristics to perform task pre-simulation, generating feasible trajectories and recording the performance indicators of each pre-simulation.

[0011] The beneficial effects of the above-mentioned further solutions are as follows: The planning method of the present invention, through a strict hierarchical optimization strategy, ensures the safety, smoothness and dynamic performance of the robotic arm trajectory, providing a reliable technical guarantee for substation oil sample collection operations.

[0012] Furthermore, S3 includes the following sub-steps: S31. Send the optimal strategy to the physical robot control system to control the physical robot to perform the task; S32. Continuously synchronize the status of the physical robot to the digital twin, monitor the task progress and key indicators in real time, and set up anomaly diagnosis triggers. S33. Determine whether the abnormal diagnosis trigger is activated. If yes, send a pause or safety stop command to the physical robot control system and proceed to S34. If no, control the physical robot to continue executing the task until it is completed. S34. In the digital twin system of the oil sample collection robot, taking the current actual state of the physical robot as a new starting point, and combining the current environmental information, execute S2 to obtain the robot's current optimal strategy, and return to S31.

[0013] Furthermore, S4 includes the following sub-steps: S41. After the physical robot finishes performing the task, generate a comparative analysis report of the task's pre-performance and execution. S42. Extract the triplet data of scenario features, strategy used, and execution effect for the task execution, and update the knowledge accordingly. Specifically: If any strategy scores above the excellence threshold in a specific scenario, the weight of the algorithm combination of that strategy in the knowledge base is increased; if any strategy activates the anomaly diagnosis trigger, the scenario in which the strategy failed is recorded, and the weights of various performance indicators are adjusted to optimize the selection of the optimal strategy in the future.

[0014] The beneficial effects of the above-mentioned further solutions are as follows: The present invention utilizes the abnormal feedback of the physical robot to trigger the re-decision mechanism of the digital twin. That is, after the algorithm is switched again in the virtual domain, the parameters are adjusted and verified to be successful, the feasible path is sent to the physical robot again, realizing an intelligent closed loop from "execution failure" to "strategy reconstruction", effectively ensuring the safety and reliability of the task.

[0015] The beneficial effects of this invention are as follows: (1) Reduced debugging risks and maintenance costs: This invention constructs a digital twin system for an oil sample collection robot. The digital twin system performs competitive pre-playing of the task, selecting the optimal strategy before execution. This eliminates risks such as collisions, equipment damage, and oil leaks caused by algorithm incompatibility, achieving "zero-risk" debugging and solving the problem of insufficient feasibility decisions before task execution in existing technologies. When the physical execution encounters a sudden anomaly, the system can automatically trigger a "closed-loop reconstruction decision," reverting to the virtual domain to quickly generate a new strategy and resume the task.

[0016] (2) Improve trajectory planning accuracy: This invention constructs a virtual-real real-time data mapping, integrates A*, DWA, improved RRT and fifth-order polynomial interpolation algorithms, comprehensively evaluates time consumption, safety and smoothness, decides the optimal algorithm and parameter configuration and generates a feasible path. This feasible path takes into account path optimality, obstacle avoidance safety and motion smoothness, effectively solves the deviation problem between traditional simulation and on-site execution, and reduces the positioning error of the robotic arm end effector by more than 30%.

[0017] (3) Enhancing Robot Adaptability: Existing robot algorithms and strategies are fixed and cannot cope with the complex working conditions at substation sites. This invention, based on online scene feature recognition and an algorithm strategy knowledge base, monitors the status of the physical robot in real time during task execution. When the physical robot experiences execution failure, prolonged stagnation, or planning deadlock, the system automatically activates a path planning failure trigger, reconstructs the feasible path of the physical robot, and quickly switches to a backup algorithm or adjusts constraint parameters for a second rehearsal in the digital twin system. After verification, the new strategy is reissued to the physical robot, realizing an online adaptive closed loop of "perception-decision-execution," solving the problem of existing technologies lacking closed-loop reconstruction decisions based on virtual and real feedback during task execution. Whether it is a change in static obstacle distribution or the sudden appearance of dynamic obstacles, the system can autonomously switch to the optimal algorithm combination. This intelligent adaptive capability enables the robot to effectively cope with uncertainties such as ground slippage, joint back gaps, and temporary obstacles, transforming task failures under a single strategy into opportunities for autonomous optimization of the system, and improving the success rate and continuity of the task by more than 80%. Attached Figure Description

[0018] Figure 1 This is a flowchart of the intelligent pre-simulation decision-making method for substation oil sample collection robots based on digital twins, as described in this invention.

[0019] Figure 2 This is a schematic diagram of data communication transmission according to the present invention.

[0020] Figure 3 This is a framework diagram of the digital twin system for oil sample collection robots of the present invention. Detailed Implementation The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0021] like Figure 1 As shown, in one embodiment of the present invention, the intelligent pre-simulation decision-making method for substation oil sample collection robots based on digital twins includes the following steps: S1. Construct a digital twin system for an oil sample collection robot; S2. Based on the digital twin system of the oil sample collection robot, the task execution is rehearsed to obtain the optimal strategy of the robot; S3. The physical robot executes the task according to the optimal strategy and performs closed-loop reconfiguration of the strategy; S4. Conduct an evaluation after the physical robot has completed its task.

[0022] S1 includes the following steps: S11. Based on the measured dimensions in the station, establish a high-fidelity digital twin including the scene, sampling points, obstacles and robots, and complete the scale calibration and kinematic link configuration; S12. Configure a communication module on the digital twin to enable two-way data interaction between the digital twin and the physical robot; In this embodiment, the bidirectional data interaction between the digital twin and the physical robot is as follows: Figure 2 As shown, this communication link is not only used for read-only acquisition of the robot's real-time status (pose, joint angle, sensor data), but more importantly, it is used to issue control strategies (such as algorithm configuration and trajectory parameter packages) after decision-making.

[0023] S13. Preset a safety rule base (collision, joint limit, singular configuration, etc.), initialize the knowledge base, and store the mapping relationship between different scene features (such as static / dynamic obstacle density, channel width) and the optimal algorithm combination.

[0024] In this embodiment, the constructed oil sample collection robot digital twin system is as follows: Figure 3 As shown.

[0025] In S11, the specific method for creating a high-fidelity digital twin is as follows: Using 3ds Max and SolidWorks 3D modeling software, 3D geometric modeling of the scene, sampling points, obstacles, and robot was performed. The 3D model was exported in .stl format. 3ds Max was used to convert the 3D model format, unnecessary parts were removed for lightweight processing, and abnormal coordinate axes were corrected to obtain a 3D model in .fbx format. This model was then imported into the Unity 3D digital twin platform to complete the creation of a high-fidelity digital twin.

[0026] In this embodiment, to achieve a one-to-one reproduction of the geometric dimensions and shape of the physical entity and the work object, 3ds Max and SolidWorks 3D modeling software are used to perform 3D geometric modeling of the physical entity, work object, and scene. The robot scene entity set SolidWorks model mainly includes the physical entity geometric modeling of the robotic arm, mobile chassis, and work object, achieving a 1:1 accurate reproduction from reality to simulation. Its export format is .stl. Since the Unity 3D digital twin platform does not support the .stl model format exported from SolidWorks, and the exported model contains a large number of unnecessary parts, it consumes too much memory during real-time visualization rendering in Unity 3D, reducing rendering efficiency. This problem will further increase latency and reduce the overall system performance in the rendering-on-demand mode of the digital twin system. Therefore, this embodiment uses 3ds Max to convert the format and remove some unnecessary parts for lightweight processing. At the same time, it observes whether the coordinate axes of individual models are abnormal, corrects the abnormal coordinate axes, and finally exports the .fbx format and imports it into the Unity 3D platform.

[0027] S2 includes the following steps: S21. Based on the constructed digital twin system of the oil sample collection robot, in the digital twin environment, the scene feature vector of the current task is identified according to the data of the digital twin and virtual sensors, and a preset number of algorithm combinations are selected from the knowledge base according to the scene feature vector. S22. Use a combination of algorithms in parallel or serial mode to perform task rehearsals. Each rehearsal drives the digital twin to execute the task from the starting point to the sampling point and records the performance indicators of the rehearsal. The performance indicators include the total task time, trajectory smoothness, minimum safe distance, peak joint torque, and whether safety rules are triggered. S23. Based on the preset performance index weights, perform weighted scoring on each performance index, and take the algorithm combination corresponding to the highest score of the pre-performance as the robot's optimal strategy.

[0028] In this embodiment, based on the determined optimal strategy, the pre-simulation results of the strategy are subjected to a final security check. If the check passes, an access instruction, an executable trajectory / parameter package, and a review report are generated.

[0029] In S22, the method for rehearsing the task includes the following steps: The specific method for task pre-simulation is as follows: a hierarchical planning strategy is adopted to plan feasible trajectories for the chassis motion and robotic arm control of the oil sample robot, and competitive pre-simulation and evaluation are carried out based on scene characteristics to obtain the optimal feasible trajectory with no collision. The competitive pre-simulation method is as follows: the selected algorithm combination is added to the competition pool, and the algorithm combination is switched according to scene characteristics to perform task pre-simulation, generating feasible trajectories and recording the performance indicators of each pre-simulation.

[0030] In this embodiment, the present invention employs a hierarchical planning strategy to plan the motion trajectory of the oil sample collection robot. The digital twin system of the oil sample collection robot automatically analyzes the preset scene features (static / dynamic obstacle density, channel width, ground flatness, etc.) based on the virtual scene, and reads the selected algorithm combinations for competitive pre-playing and evaluation.

[0031] Regarding the motion control of the oil sample robot chassis, in high-density static obstacle scenarios, the digital twin system will perform parallel pre-playing of algorithm combinations such as A* + DWA and RRT* + DWA; in dynamic obstacle scenarios, it will perform pre-playing of combinations such as RRT* + DWA; and in open scenarios, it will use A* and pure tracking algorithms. The following uses a single pre-playing process of the "A+DWA" combination in a high-density static obstacle scenario as an example to illustrate its specific implementation and data acquisition method.

[0032] In a digital twin virtual environment, the A* algorithm is used to generate a global static path from the starting point to the target sampling point. Based on this, the Dynamic Window Method (DWA) is employed for local trajectory planning. By sampling and simulating short-term trajectories within the velocity space, dynamic obstacle avoidance in the virtual environment is achieved. Subsequently, the system uses Unity's physics engine to perform discrete collision detection on the integrated global and local paths, completing the pre-verification of trajectory feasibility. During this process, the system simultaneously records key evaluation metrics for this pre-simulation, such as total task time, trajectory smoothness, and minimum safe distance, providing data support for subsequent decisions. If a collision risk is detected, the A* algorithm is immediately invoked in the virtual environment, starting from the current position, for local replanning, and the new path is smoothed and optimized until a completely collision-free feasible trajectory is generated. This fully verified final trajectory is authorized for execution on the physical chassis, thus ensuring operational safety and stability from the source. After algorithm scheduling is completed, the system enters the trajectory generation stage. Taking the A*+DWA combination used in a high-density static obstacle scene as an example, the specific implementation is as follows: The 3D virtual scene of the substation is reduced in dimension and projected into a 2D raster map, with different access costs assigned to areas such as equipment areas and safety passages. The A* algorithm is used, and the evaluation function is evaluated... Perform an iterative search to generate a globally optimal static path from the starting point to the target area, which serves as the navigation baseline.

[0033] The main advantage of the A* algorithm is its ability to plan the globally optimal path from the starting point to the target point, continuously searching for the nearest node until the target point is reached. This reduces the complexity of practical problems and narrows the search range. Considering the practical application of robots, the A* algorithm is the shortest and most efficient path planning method in static environment models, enabling relatively fast global path planning. The evaluation function of the A* algorithm is as follows: in, This indicates that the robot generates a path from the starting point to reach the node. n The cost, Indicates the current node of the robot. n The estimated cost to reach the target point; This represents the minimum total cost for the robot to move from its initial state to the target position.

[0034] Guided by a global path, dynamic windowing (DWA) is used for local real-time motion planning. This process includes: (1) Velocity sampling: In the velocity space (v, ω), the robot's velocity is limited by its own performance, including motor performance, safe speed, and its own motion, as well as numerous external factors. Velocity sampling cannot be endless. Therefore, appropriate velocity sampling constraints are needed to ensure that the obtained robot velocities are applicable in practice. These constraints can be mainly categorized as follows: Limit speed constraint: Due to the inherent hardware limitations of the robot, there will inevitably be a maximum speed. Therefore, a boundary is set for it, which is the limit speed constraint. The specific expression is: In the formula, Linear velocity, Angular velocity, This represents the minimum linear velocity. This represents the maximum linear velocity. This is the minimum angular velocity. This represents the maximum angular velocity.

[0035] Robot motor performance constraints: In practical robot applications, due to motor torque, there are upper and lower limits to motor acceleration and deceleration. These constraints define the performance constraints of robot motors. The specific expression is: In the formula, This is the robot's maximum linear velocity. and These represent the maximum deceleration and maximum acceleration of the linear velocity that the robot's motors can achieve, respectively. This indicates the robot's maximum angular velocity; and These represent the maximum deceleration and maximum acceleration of the angular velocity achievable by the motor carried by the robot, respectively, and T is the velocity sampling time.

[0036] Robot safety constraints: When the robot detects an obstacle using its sensors, it must have sufficient reaction time to ensure it can avoid the obstacle. Therefore, a safe speed is set. The set safe speed range... The expression is: In the formula, This is the obstacle distance evaluation function; (2) Trajectory prediction: Using the kinematic model of a differential-driven mobile robot, trajectory prediction is calculated as follows: In the formula, For robots at all times k x-axis coordinates For robots at all times k -1 is the x-axis coordinate. For robots at all times k The vertical axis coordinate, For robots at all times k -1 is the ordinate on the vertical axis. For robots at all times k The orientation angle, For robots at all times k -1 orientation angle, Indicates the sampling interval.

[0037] (3) Trajectory evaluation: In the entire velocity space, each set of velocity (v, ω) data can predict a trajectory. The optimal trajectory needs to be selected from numerous trajectories for path planning. An evaluation function is introduced to select the best trajectory to control the robot and achieve path planning. The trajectory evaluation function in the dynamic window method... The specific expression is: In the formula, This represents the azimuth evaluation function. For the velocity evaluation function, As azimuth weight, For safety distance weighting, For speed weighting, This is a smoothing factor or normalization parameter. The robot travels at a simulated speed, and this parameter evaluates the angle difference between the direction and the target point when the robot reaches the end of the trajectory at the current speed; that is, the difference between the predicted azimuth and the actual azimuth. The purpose of this function is to find the optimal trajectory with the smallest angle between the predicted and actual azimuth. The algorithm generates simulated trajectories by sampling in the velocity space and evaluates their safety and efficiency. The algorithm eliminates trajectories that are too close to obstacles until it finds an optimal, collision-free path, or traverses all possible trajectories, ensuring that the robot plans a path safely and efficiently.

[0038] Finally, the Unity physics engine is used to perform discrete collision detection on the integrated global and local paths, completing the pre-verification of trajectory feasibility. If a collision risk is detected, an adaptive mechanism is immediately triggered. If a slight deviation occurs, the A* algorithm is invoked in the virtual environment, starting from the current position, for local replanning, and the new path is smoothed using B-spline curves. If scene feature indicators exceed preset thresholds (such as a sudden increase in the speed of dynamic obstacles exceeding the threshold), the system automatically reverts to scene perception, rematches the algorithm combination (such as switching to RRT*+DWA combination), loads the corresponding parameter template, and restarts the planning process in the current state. This detection and optimization process is repeated until a completely collision-free feasible trajectory is generated. The final trajectory, after thorough verification, is authorized for execution on the physical chassis, thus ensuring operational safety and stability from the source.

[0039] Regarding the control of the robotic arm for oil sampling, when the accessibility of the sampling port is assessed as "difficult" (e.g., severe occlusion, high pose singularity), the digital twin system will test different algorithm combinations, such as standard RRT and bidirectional RRT, in parallel, and enable redundant joint optimization mode. The sampling range and exploration weight in the parameter template will be increased to ensure success rate under complex conditions. The system will record the planning time, success rate, and trajectory quality of different algorithms under the same difficult scenario for comparison by the decision module. Finally, the optimized trajectory is validated in the Unity environment. Dense discrete sampling collision detection is performed on the complete trajectory to verify the satisfaction of kinematic and dynamic constraints. Only trajectories that pass all safety checks are authorized for execution, ensuring the accuracy, stability, and safety of the robotic arm's movement during oil sample collection. The following uses a rapidly expanding random tree as an example to illustrate its specific implementation and data acquisition method.

[0040] First, based on the spatial pose of the transformer sampling port identified by the virtual vision system, the target joint angle state of the robotic arm is obtained through inverse kinematics, establishing an accurate mapping relationship from the task space to the joint space. Second, an improved fast expanding random tree algorithm is used for collision-free path search in the joint space. A search tree is constructed with the current joint state as the root node, and sampling points are generated using a target bias sampling strategy. New nodes are expanded from the nearest node to the sampling points with a fixed step size. During node expansion, Unity's collision detection function is invoked in real time to verify the safety of the path segment with the transformer body, insulating bushings, and other equipment. A reconnection mechanism is used to evaluate the cost of neighboring nodes, continuously optimizing the path quality to ensure a complete collision-free feasible path is obtained. Next, motion smoothness optimization is performed on the obtained collision-free path. A fifth-order polynomial interpolation method is used to construct the joint trajectory function. By setting boundary constraints for the position, velocity, and acceleration of the starting and ending points, a complete system of equations is established to solve for the polynomial coefficients. This method ensures that the generated trajectory remains continuous in terms of position, velocity, and acceleration, effectively eliminating mechanical shock and improving motion stability.

[0041] Then, the smoothed trajectory is optimized for time parameters. An optimization function is established to minimize joint torque variations and motion time, comprehensively considering the velocity limits, acceleration constraints, and torque constraints of each joint of the robotic arm. Through optimization algorithms, the time-optimal trajectory parameters are obtained while satisfying all dynamic constraints, ensuring the smoothness and efficiency of the robotic arm's motion.

[0042] S3 includes the following steps: S31. Send the optimal strategy to the physical robot control system to control the physical robot to perform the task; S32. Continuously synchronize the status of the physical robot to the digital twin, monitor the task progress and key indicators in real time, and set up anomaly diagnosis triggers, including path planning failure triggers, progress stagnation triggers and safety threshold triggers. In this embodiment, when a feasible path cannot be generated within a preset time, a path planning failure trigger is activated; when the robot stays at a certain position for a period of time exceeding a time threshold, a progress stagnation trigger is activated; and when the virtual-real deviation or joint limit proximity indicators exceed the limits, a safety threshold trigger is activated. S33. Determine whether the abnormal diagnosis trigger is activated. If yes, send a pause or safety stop command to the physical robot control system and proceed to S34. If no, control the physical robot to continue executing the task until it is completed. S34. In the digital twin system of the oil sample collection robot, taking the current actual state of the physical robot as a new starting point, and combining the current environmental information, execute S2 to obtain the robot's current optimal strategy, and return to S31.

[0043] In this embodiment, based on the digital twin system of the oil sample collection robot, the robot's current actual state is used as a new starting point. Combined with the latest environmental information, the decision-making mechanism in step S2 is triggered to obtain the robot's optimal strategy. At this time, the digital twin system will exclude failed algorithms, re-select candidate algorithms from the knowledge base (such as switching A+DWA to RRT+DWA), or adjust key parameters (such as increasing the sampling step size and relaxing smoothness constraints) to perform rapid secondary pre-playing and decision-making, generating a new and feasible "reconstruction strategy". The "reconstruction strategy" is then sent to the physical robot, instructing it to resume task execution from the breakpoint. Thus, a complete intelligent closed loop of "anomaly perception - rollback twin - reconstruction decision - resumption of execution" is completed.

[0044] S4 includes the following steps: S41. After the physical robot finishes performing the task, generate a comparative analysis report of the task's pre-performance and execution. S42. Extract the triplet data of scenario features, strategy used, and execution effect for the task execution, and update the knowledge accordingly. Specifically: If any strategy scores above the excellence threshold in a specific scenario, the weight of the algorithm combination of that strategy in the knowledge base is increased; if any strategy activates the anomaly diagnosis trigger, the scenario in which the strategy failed is recorded, and the weights of various performance indicators are adjusted to optimize the selection of the optimal strategy in the future.

[0045] In this embodiment, by extracting the triplet data of scenario features, strategy used, and execution effect of this task, if a certain strategy performs well in a specific scenario, its weight in the knowledge base is strengthened; if a certain strategy leads to reconstruction, its failure scenario is recorded to optimize the initial screening logic of future strategies, so as to realize the continuous evolution of the system's decision-making ability.

[0046] In the description of this invention, it should be understood that the above content is merely a preferred embodiment of the invention and is not intended to limit the scope of protection of the invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A digital twin-based intelligent pre-simulation decision-making method for substation oil sample collection robots, characterized in that, Includes the following steps: S1. Construct a digital twin system for an oil sample collection robot; S2. Based on the digital twin system of the oil sample collection robot, the task execution is rehearsed to obtain the optimal strategy of the robot; S3. The physical robot executes the task according to the optimal strategy and performs closed-loop reconfiguration of the strategy; S4. Conduct an evaluation after the physical robot has completed its task.

2. The intelligent pre-simulation decision-making method for substation oil sample collection robots based on digital twins according to claim 1, characterized in that, S1 includes the following steps: S11. Based on the measured dimensions in the station, establish a high-fidelity digital twin including the scene, sampling points, obstacles and robots, and complete the scale calibration and kinematic link configuration; S12. Configure a communication module on the digital twin to enable two-way data interaction between the digital twin and the physical robot; S13. Preset a security rule base and initialize a knowledge base. The knowledge base stores the mapping relationship between different scenario features and the optimal algorithm combination.

3. The intelligent pre-simulation decision-making method for substation oil sample collection robot based on digital twin as described in claim 2, characterized in that, In S11, the specific method for creating a high-fidelity digital twin is as follows: Using 3ds Max and SolidWorks 3D modeling software, 3D geometric modeling of the scene, sampling points, obstacles, and robot was performed. The 3D model was exported in .stl format. 3ds Max was used to convert the 3D model format, unnecessary parts were removed for lightweight processing, and abnormal coordinate axes were corrected to obtain a 3D model in .fbx format. This model was then imported into the Unity 3D digital twin platform to complete the creation of a high-fidelity digital twin.

4. The intelligent pre-simulation decision-making method for substation oil sample collection robot based on digital twin as described in claim 1, characterized in that, S2 includes the following steps: S21. Based on the constructed digital twin system of the oil sample collection robot, in the digital twin environment, the scene feature vector of the current task is identified according to the data of the digital twin and virtual sensors, and a preset number of algorithm combinations are selected from the knowledge base according to the scene feature vector. S22. Use a combination of algorithms in parallel or serial mode to perform task rehearsals. Each rehearsal drives the digital twin to execute the task from the starting point to the sampling point and records the performance indicators of the rehearsal. The performance indicators include the total task time, trajectory smoothness, minimum safe distance, peak joint torque, and whether safety rules are triggered. S23. Based on the preset performance index weights, perform weighted scoring on each performance index, and take the algorithm combination corresponding to the highest score of the pre-performance as the robot's optimal strategy.

5. The intelligent pre-simulation decision-making method for substation oil sample collection robot based on digital twin as described in claim 4, characterized in that, In S22, the method for rehearsing the task includes the following steps: The specific method for task pre-simulation is as follows: a hierarchical planning strategy is adopted to plan feasible trajectories for the chassis motion and robotic arm control of the oil sample robot, and competitive pre-simulation and evaluation are carried out based on scene characteristics to obtain the optimal feasible trajectory with no collision. The competitive pre-simulation method is as follows: the selected algorithm combination is added to the competition pool, and the algorithm combination is switched according to scene characteristics to perform task pre-simulation, generating feasible trajectories and recording the performance indicators of each pre-simulation.

6. The intelligent pre-simulation decision-making method for substation oil sample collection robot based on digital twin as described in claim 1, characterized in that, S3 includes the following steps: S31. Send the optimal strategy to the physical robot control system to control the physical robot to perform the task; S32. Continuously synchronize the status of the physical robot to the digital twin, monitor the task progress and key indicators in real time, and set up anomaly diagnosis triggers. S33. Determine whether the abnormal diagnosis trigger is activated. If yes, send a pause or safety stop command to the physical robot control system and proceed to S34. If no, control the physical robot to continue executing the task until it is completed. S34. In the digital twin system of the oil sample collection robot, taking the current actual state of the physical robot as a new starting point, and combining the current environmental information, execute S2 to obtain the robot's current optimal strategy, and return to S31.

7. The intelligent pre-simulation decision-making method for substation oil sample collection robot based on digital twin as described in claim 1, characterized in that, S4 includes the following steps: S41. After the physical robot finishes performing the task, generate a comparative analysis report of the task's pre-performance and execution. S42. Extract the triplet data of scenario features, strategy used, and execution effect for the task execution, and update the knowledge accordingly. Specifically: If any strategy scores above the excellence threshold in a specific scenario, the weight of the algorithm combination of that strategy in the knowledge base is increased; if any strategy activates the anomaly diagnosis trigger, the scenario in which the strategy failed is recorded, and the weights of various performance indicators are adjusted to optimize the selection of the optimal strategy in the future.

Images