Industrial robot path optimization method and system based on machine learning

Abstract

The invention relates to the technical field of industrial robots, in particular to an industrial robot path optimization method and system based on machine learning, which is characterized in that dynamic data in an operation environment of an industrial robot are collected in real time, historical path data are combined, a path planning model is constructed, and a path is optimized by utilizing a machine learning algorithm, especially reinforcement learning. And the path candidate set is evaluated and optimized, so that adaptive adjustment of the robot path is realized, and the method specifically comprises the steps of construction of a path planning model, multi-dimensional path evaluation, path optimization and dynamic adjustment. In the path execution process, if the environment changes or the robot deviates from the preset path, the system can automatically correct the path, it is ensured that the robot executes tasks along the optimal path, and collision with obstacles is avoided. According to the invention, the operation efficiency, safety and reliability of the industrial robot in a complex and dynamic environment can be improved, and efficient and flexible production requirements are met.

Description

Technical Field

[0001] This invention relates to the field of industrial robot technology, and in particular to a method and system for optimizing industrial robot paths based on machine learning. Background Technology

[0002] With the rapid development of industrial automation and intelligence, industrial robots are being used more and more widely in manufacturing, playing an irreplaceable role, especially in complex production tasks. Path planning, as one of the core issues in robotics technology, directly affects the robot's working efficiency, accuracy, and safety.

[0003] While traditional path planning methods can achieve a certain degree of path optimization in static environments, existing technologies suffer from several problems in dynamic and complex working environments. First, traditional methods ignore the dynamic changes in the working environment and cannot respond to unexpected situations in real time, such as moving obstacles, resulting in unsatisfactory path planning performance. Second, existing path planning algorithms typically cannot comprehensively evaluate multiple optimization objectives (such as path length, energy consumption, time consumption, and safety), leading to overly simplistic path selection that fails to comprehensively improve the robot's working efficiency and safety in complex environments. Summary of the Invention

[0004] This invention provides a method and system for optimizing the path of industrial robots based on machine learning.

[0005] The machine learning-based industrial robot path optimization method includes the following steps: S1, Data Acquisition and Environment Modeling: Real-time acquisition of dynamic data in the industrial robot's working environment, including the robot's current working status, workpiece position, obstacle distribution, production task requirements, and environmental parameters, and modeling of the robot's working environment to build a dynamic environment model. At the same time, historical path data is obtained from historical path records. S2, Path planning model construction: Based on machine learning algorithms, combined with historical path data and dynamic environment models, a path planning model is constructed to capture the non-linear relationship between the path and the environment and generate multiple path candidate sets; S3, Path Evaluation and Optimization: The candidate path set is evaluated according to the preset evaluation criteria, including path length, time consumption, energy consumption and path safety. The reinforcement learning algorithm is used to optimize the path evaluation results by evaluating and iterating each path multiple times. The path planning model is then optimized based on the evaluation results to select the optimal path scheme. S4, Path Execution and Dynamic Adjustment: When the robot is executing the optimal path plan, the dynamic data in the robot's working environment is monitored in real time. If the robot deviates from the predetermined path or the environment changes (such as the appearance of new obstacles), the path adjustment mechanism is automatically triggered to correct the path and ensure that the robot performs the task along the optimal path and avoids collisions with obstacles.

[0006] Optionally, S1 includes: S11, Real-time acquisition of dynamic data in the industrial robot's operating environment, including the robot's current working status, workpiece position, distribution of surrounding obstacles, production task requirements, and environmental parameters; The robot's current working status includes its position, speed, and acceleration. The position is measured in real time by the positioning system on the industrial robot, and the speed and acceleration are obtained through the industrial robot's inertial measurement unit. The position of the workpiece in the working environment is acquired in real time using a vision sensor; The distribution of surrounding obstacles is detected in real time using lidar and infrared sensors; Obtain production task requirements through a task management system or production scheduling system, including the target location, work path, and processing task; Environmental parameters, including temperature and humidity, are collected in real time using temperature and humidity sensors. S12, retrieve historical path data from historical path records; S13, preprocess the collected dynamic data and historical path data, including noise reduction and outlier detection.

[0007] Optionally, the step of modeling the robot's operating environment and constructing a dynamic environment model includes: S14. Based on the collected data on the robot's current position, workpiece position, and obstacle distribution, construct a dynamic environment model in a spatial coordinate system. S15, based on the collected temperature, humidity and production task requirements, updates the relevant variables in the dynamic environment model to form a real-time representation of the environment's state.

[0008] Optionally, S2 includes: S21, preprocess the collected historical path data and the constructed dynamic environment model; S22, uses deep neural networks to extract features from historical path data and dynamic environment models; S23, a path planning model is constructed based on a deep neural network algorithm.

[0009] S24. After the model training is completed, the path planning model is used to generate multiple path candidate sets. Each path candidate takes into account the obstacle positions, workpiece positions and production task requirements in the dynamic environment model.

[0010] Optionally, S3 includes: S31 sets evaluation criteria for each path candidate set, including path length, time consumption, energy consumption, and path safety; S32, based on the set evaluation criteria, performs a preliminary evaluation of each path candidate set; S33. During the evaluation process, the objectives of path optimization include minimizing path length, time consumption, and energy consumption, while maximizing path safety.

[0011] Optionally, the path optimization includes: S34 uses the Q-learning algorithm as a reinforcement learning algorithm to optimize the path evaluation results.

[0012] S35, through the iterative training process of reinforcement learning, evaluates and optimizes the path candidate set multiple times; S36. In each iteration of path optimization, the path planning model is optimized based on the optimal path evaluation result obtained by the reinforcement learning algorithm, and the weight values ​​and parameters in the model are adjusted.

[0013] Optionally, the path optimization specifically includes: S37. In the path evaluation and optimization process, after obtaining the path evaluation results through the reinforcement learning algorithm, the path candidate set is screened and the path with the highest evaluation score is retained. S38. After selecting the optimal path, the robot executes the path and monitors the execution.

[0014] Optionally, S4 includes: S41 collects and monitors dynamic data of the robot's working environment in real time while the robot is executing the optimal path plan.

[0015] S42, If the robot deviates from the predetermined path or the environment changes (such as the appearance of new obstacles or changes in environmental conditions), the path adjustment mechanism is automatically triggered. S43, After the path adjustment mechanism is activated, the A* algorithm is used to correct the path; S44, the robot continues to perform its task according to the corrected path, monitoring the path execution status in real time. If a path deviation or new obstacle occurs again, the readjustment mechanism is triggered.

[0016] Optionally, during the path execution and dynamic adjustment process, the robot's task execution status and environmental changes are continuously monitored, and the path planning model is updated and retrained using the new execution data.

[0017] The machine learning-based industrial robot path optimization system, used to implement the aforementioned machine learning-based industrial robot path optimization method, includes the following modules: Data acquisition and environment modeling module: Real-time acquisition of dynamic data in the industrial robot's operating environment, including the robot's current working status, workpiece position, obstacle distribution, production task requirements, and environmental parameters, and construction of a dynamic environment model. At the same time, historical path data is obtained from historical path records. Path planning model construction module: Based on machine learning algorithms, combined with historical path data and dynamic environment models, a path planning model is constructed, and multiple path candidate sets are generated through the path planning model; Path evaluation and optimization module: Evaluates each candidate path set according to preset evaluation criteria, optimizes the path evaluation results using reinforcement learning algorithms, and optimizes the path planning model based on the evaluation results to select the optimal path scheme; Path execution and dynamic adjustment module: When the robot is executing the optimal path plan, it monitors the dynamic data in the robot's working environment in real time. If the robot deviates from the predetermined path or the environment changes (such as the appearance of new obstacles), the path adjustment mechanism is automatically triggered to correct the path.

[0018] The beneficial effects of this invention are: This invention, by comprehensively utilizing historical path data and a real-time dynamic environment model, enables the method to capture complex nonlinear relationships within the robot's operating environment. The construction of the path planning model effectively integrates various environmental factors (such as obstacle positions, workpiece changes, and environmental parameters), resulting in more accurate path planning and the ability to adapt to environmental changes promptly. Particularly in dynamic environments, the robot can adjust its path in real time, ensuring optimal path performance during execution and preventing reduced operational efficiency due to environmental variations.

[0019] This invention utilizes reinforcement learning algorithms for path evaluation and optimization. The method intelligently evaluates multiple candidate paths based on preset evaluation criteria (such as path length, time consumption, energy consumption, and path safety). Path optimization not only improves the execution efficiency of robot tasks but also reduces wasted time and energy while ensuring safety. Multiple evaluations and iterative optimizations through reinforcement learning ensure that the path planning results accurately reflect the actual needs of complex working environments, significantly improving the efficiency and reliability of robot operations.

[0020] This invention monitors dynamic data in the robot's operating environment in real time while the robot is executing the optimal path. If the environment changes or the robot deviates from the predetermined path, the method automatically triggers a path adjustment mechanism, replanning the path in real time using the A* algorithm path correction method. This mechanism effectively avoids collisions between the robot and newly encountered obstacles, while ensuring that the robot successfully completes the task along the optimal path. Through dynamic adjustment, path planning not only has real-time performance but also continuously optimizes execution results, improving operational safety and efficiency. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the system flow according to an embodiment of the present invention. Detailed Implementation

[0023] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0024] It should be noted that the use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when a specific feature, structure, or characteristic is described in connection with an embodiment, implementing such a feature, structure, or characteristic in conjunction with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the art.

[0025] Generally, terms can be understood at least partly from their use in context. For example, depending at least partly on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood not necessarily to convey an exclusive set of factors, but rather, alternatively, depending at least partly on the context, to allow for the presence of other factors that are not necessarily explicitly described.

[0026] like Figure 1 As shown, the machine learning-based industrial robot path optimization method includes the following steps: S1, Data Acquisition and Environment Modeling: Real-time acquisition of dynamic data in the industrial robot's working environment, including the robot's current working status, workpiece position, obstacle distribution, production task requirements, and environmental parameters, and modeling of the robot's working environment to build a dynamic environment model. At the same time, historical path data is obtained from historical path records. S2, Path planning model construction: Based on machine learning algorithms, combined with historical path data and dynamic environment models, a path planning model is constructed to capture the non-linear relationship between the path and the environment and generate multiple path candidate sets; S3, Path Evaluation and Optimization: The candidate path set is evaluated according to the preset evaluation criteria, including path length, time consumption, energy consumption and path safety. The reinforcement learning algorithm is used to optimize the path evaluation results by evaluating and iterating each path multiple times. The path planning model is then optimized based on the evaluation results to select the optimal path scheme. S4, Path Execution and Dynamic Adjustment: When the robot is executing the optimal path plan, the dynamic data in the robot's working environment is monitored in real time. If the robot deviates from the predetermined path or the environment changes (such as the appearance of new obstacles), the path adjustment mechanism is automatically triggered to correct the path and ensure that the robot performs the task along the optimal path and avoids collisions with obstacles.

[0027] S1 includes: S11 collects dynamic data in the industrial robot's operating environment in real time. The dynamic data includes the robot's current working status, workpiece position, distribution of surrounding obstacles, production task requirements, and environmental parameters. The robot's current working status includes its position, speed, and acceleration. The position is measured in real time by the positioning system on the industrial robot (such as lidar, encoder, etc.), and the speed and acceleration are obtained through the industrial robot's inertial measurement unit. The position of the workpiece in the working environment is acquired in real time using visual sensors (such as cameras, laser scanners, etc.); The distribution of surrounding obstacles is detected in real time using lidar and infrared sensors; Obtain production task requirements through a task management system or production scheduling system, including the target location, work path, and processing task; Environmental parameters, including temperature and humidity, are collected in real time using temperature and humidity sensors. S12, retrieve historical path data from historical path records; S13, preprocess the collected dynamic data and historical path data, including noise reduction and outlier detection; By collecting dynamic data from the industrial robot's operating environment in real time, the comprehensiveness, timeliness, and accuracy of various environmental information during robot operations are ensured. This provides reliable data support for subsequent path planning and environmental modeling, helping to improve the effectiveness of path optimization and the safety of robot operations.

[0028] Modeling the robot's operating environment and constructing a dynamic environment model includes: S14. Based on the collected data on the robot's current position, workpiece position, and obstacle distribution, construct a dynamic environment model in a spatial coordinate system. The environment model includes the position and shape information of static obstacles (such as workbench, walls, etc.) and dynamic obstacles (such as moving objects or people, etc.). S15. Based on the collected temperature, humidity and production task requirements, update the relevant variables in the dynamic environment model to form a real-time state representation of the environment. For example, temperature and humidity can affect the robot's operating accuracy, and production task requirements can affect the robot's working path and execution strategy.

[0029] S2 includes: S21, preprocess the collected historical path data and the constructed dynamic environment model. The historical path data includes information such as the robot's path trajectory, operation time, and path length when performing past tasks. The dynamic environment model includes information such as obstacle distribution, workpiece position, and environmental parameters in the working environment. Preprocessing includes: Historical path data is preprocessed through data standardization and noise reduction (such as using the sliding window method) to eliminate outliers and ensure data quality; Dynamic environment models ensure their timeliness and accuracy through spatial coordinate mapping and dynamic updates (combined with real-time sensor data); S22 utilizes deep neural networks to extract features from historical path data and dynamic environment models. These features include: Geometric features in the robot's execution path (such as path curvature, turning angle, etc.); The dynamic characteristics of obstacles, such as type, shape, location, and speed; The influence characteristics of the requirements of the workpiece and the task on the path (such as path length, task accuracy requirements, etc.).

[0030] The weighted average method is used for data fusion to combine features from historical path data and dynamic environment models into a unified input feature set. The specific steps are as follows: Different weights are assigned to various features in the historical path data, and the weight values ​​are adjusted according to the relevance and importance of the features (for example, more critical features such as path length and time consumption have larger weights). Features in the dynamic environment model (such as obstacle positions, workpiece positions, etc.) are weighted, and the weight allocation is adjusted according to the degree of influence of environmental changes; The weighted average method is used to fuse all features to form a unified feature set as the input of the path planning model. The weighted average method calculation formula is expressed as follows: ; in, It is the fused feature set. It is the i-th feature. It is the weight of the i-th feature. It is the total number of features; S23, Construct a path planning model based on a deep neural network algorithm. The input of the path planning model is the feature set extracted in step S42, and the output is a path candidate set. The construction process includes: Input layer: The feature set is used as input, and the number of nodes in the input layer is the same as the feature dimension; Hidden layers: By using multiple hidden layers (such as two or three fully connected layers) and the ReLU activation function, the non-linear relationship between the path and the environment is learned layer by layer; Output layer: The output layer consists of multiple path candidates. The softmax function is used to calculate the path probability, and the paths are finally sorted according to their probability.

[0031] Model training employs backpropagation and gradient descent optimization, adjusting the neural network weights by minimizing a loss function (such as mean squared error or cross-entropy loss). Training and validation are performed using training and validation sets to progressively optimize model performance.

[0032] S24, After the model training is completed, the path planning model is used to generate multiple path candidate sets. Each path candidate considers the obstacle positions, workpiece positions, and production task requirements in the dynamic environment model. The process of generating path candidate sets includes: The current dynamic environment model is input into the trained path planning model, and multiple path candidates are output. Constraint optimization is performed on the path candidate set to ensure that the path avoids obstacles and meets the task requirements.

[0033] S3 includes: S31 sets evaluation criteria for each candidate path, including path length, time consumption, energy consumption, and path safety, as detailed below: Path length: the total distance of the path, taking into account the distance the robot travels from the starting point to the end point; Time consumption: The total time for the robot to perform the task along the path, taking into account the effects of factors such as speed changes and path tortuosity; Energy consumption: The energy consumed by the robot in performing tasks along the path, taking into account factors such as the robot's acceleration, friction, and load during movement; Path safety: Assess the distance between the path and obstacles in the environment to ensure that the robot will not collide while executing the path, thus guaranteeing path safety; These evaluation criteria can reflect different dimensions of path quality, providing a comprehensive evaluation basis for subsequent path optimization; S32, based on the established evaluation criteria, performs a preliminary evaluation of each path candidate set, specifically including: Each path is assigned an initial evaluation score by calculating indicators such as path length, time consumption, energy consumption, and safety. Using a weighted approach, the various evaluation indicators are weighted according to different path attributes to obtain a comprehensive score, expressed as follows: Path Score Length Time Energy Safety; in, The weights for the path evaluation metrics are Length, Time, Energy, and Safety, which are the scores for path length, time consumption, energy consumption, and safety, respectively. S33, During the evaluation process, the objectives of path optimization include minimizing path length, time consumption, and energy consumption, while maximizing path safety. The objective function is set as follows: Objective Function Length Time Energy Safety; in, The weights for each objective are given, and the minimization or maximization of the objective function will be adjusted according to the actual application requirements.

[0034] Path optimization includes: S34 employs the Q-learning algorithm as a reinforcement learning algorithm to optimize path evaluation results. Specific methods include: Reward function design: Based on the path evaluation criteria, design a suitable reward function so that the path optimization objective can be reflected during the algorithm training process.

[0035] Application of Q-learning algorithm: When using Q-learning algorithm for training, the optimization process of the path candidate set is to continuously update the Q value to select the optimal path. The Q value is defined as the expected reward under each path state, and the calculation formula is: ; in, For state Take action below Q value, This is the reward value for the current path. As a discount factor, For the next state, For the next action; Training process: Through repeated trials and feedback, the Q-learning algorithm gradually adjusts the path selection strategy and learns to optimize path evaluation metrics.

[0036] S35, through an iterative training process of reinforcement learning, evaluates and optimizes the path candidate set multiple times, including: After each evaluation, the path selection strategy is adjusted by updating the Q value based on the feedback of the reward function to evaluate the quality of the path, and gradually converges to the optimal path. After each iteration, the algorithm comprehensively evaluates the path length, time consumption, energy consumption, and safety, and optimizes the objective function. S36, In each iteration of path optimization, the path planning model is optimized based on the optimal path evaluation result obtained by the reinforcement learning algorithm, adjusting the weight values ​​and parameters in the model, including: By learning the optimal path selection strategy obtained from Q-learning, the path planning model is optimized to adapt to different environments and task requirements. Fine-tune the parameters in the path planning model to enable it to better predict the optimal path in future tasks.

[0037] Path optimization specifically includes: S37. In the path evaluation and optimization process, after obtaining the path evaluation results through the reinforcement learning algorithm, the path candidate set is screened and the path with the highest evaluation score is retained. Each optimized path is selected based on minimizing length, time consumption, energy consumption, and maximizing safety to ensure that the selected path has the highest performance while meeting the requirements of production tasks. S38. After selecting the optimal path, the robot executes the path and monitors the execution.

[0038] S4 includes: S41, while the robot is executing the optimal path plan, real-time dynamic data of the robot's working environment is collected and monitored. The dynamic data includes the robot's current position, speed, acceleration, workpiece position, and the distribution of surrounding obstacles. The specific steps are as follows: The robot's position, speed, and acceleration are monitored in real time using a positioning system (such as lidar, encoder, etc.) installed on the robot.

[0039] The robot collects the workpiece position in real time using visual sensors (such as cameras and laser scanners) to ensure that it performs tasks along a predetermined path.

[0040] By using lidar and infrared sensors to obtain the location of surrounding obstacles in real time, the robot can be ensured to perceive changes in the surrounding environment in real time during operation.

[0041] S42, if the robot deviates from the predetermined path or the environment changes (such as the appearance of new obstacles or changes in environmental conditions), the path adjustment mechanism is automatically triggered. The specific steps are as follows: Based on real-time monitoring data, the deviation between the robot's current position and the optimal path is calculated. If the deviation exceeds a preset threshold, a path adjustment mechanism is triggered.

[0042] At the same time, if new obstacles (such as moving objects or people) appear in the environment, the system will automatically detect and identify the new obstacles and trigger path replanning. S43, After the path adjustment mechanism is activated, the A* algorithm is used for path correction, including: S431, Path Correction Calculation: Based on the robot's current position, obstacle distribution, and environmental parameters, a new feasible path is planned. The specific process is as follows: The robot obtains its current position (including position, velocity, acceleration, etc.) and obstacle distribution information (such as the position and shape of obstacles detected by LiDAR and visual sensors) through sensors.

[0043] The A* algorithm is used for path planning. It calculates the cost of each candidate path node (including the actual cost from the starting point to the node and the estimated cost from the node to the destination), and selects the path with the minimum cost as the optimal path. The calculation formula for the A* algorithm is as follows: ; in, This represents the total cost from the starting point to the current node. It is the actual cost (path length, time, etc.) from the starting point to the current node. It is the heuristic cost estimate from the current node to the target point (i.e., the estimated distance from the current node to the target point); In practical applications, the A* algorithm evaluates all possible path nodes and selects the path with the lowest cost, ensuring that the robot can find an effective path in a complex environment. S432, Optimize and Correct the Path: The corrected path will be further optimized based on preset evaluation criteria, including path length, safety, time consumption, and energy consumption. The specific steps are as follows: After the corrected path is generated, it is optimized according to evaluation criteria (such as path safety, whether it avoids obstacles, whether it needs to pass through high-risk areas, etc.). Optimization algorithms adjust paths by calculating the total cost (including time, energy, etc.) to better suit task requirements and environmental constraints. The optimization process may include reducing path length, avoiding obstacles or dangerous areas, etc. S44, the robot continues its task based on the corrected path, monitoring the path execution status in real time. If a path deviation or new obstacle occurs again, a readjustment mechanism is triggered, with the specific steps as follows: With the path corrected, the robot begins to perform tasks along the new path and continuously monitors its current position through the positioning system; During execution, sensors continue to monitor the surrounding environment to ensure that the robot completes the task along the corrected path; Once a deviation or new obstacle is detected in the path execution, the path adjustment mechanism is immediately triggered to replan the path and ensure that every segment of the path executed by the robot meets the optimal conditions.

[0044] During path execution and dynamic adjustment, the robot's task execution and environmental changes are continuously monitored, and the path planning model is updated and retrained using the new execution data. The specific steps are as follows: During the robot's execution path, execution feedback data is collected, including the robot's actual movement path, time consumption, energy consumption, and task completion rate.

[0045] Based on this execution data, the path planning model is retrained using regression analysis. Regression analysis updates the parameters of the path planning model by fitting the relationship between path execution data and environmental variables, enabling the model to more accurately predict the robot's performance in future tasks. The regression analysis model can capture the nonlinear relationship between path length, time, energy consumption, and environmental factors (such as obstacle positions and workpiece movements).

[0046] The updated path planning model will be optimized based on new training data, improving the adaptive capability of path planning and ensuring that the robot can adjust its path execution in real time in dynamic environments, thereby improving the efficiency and success rate of tasks.

[0047] like Figure 2 As shown, the machine learning-based industrial robot path optimization system, used to implement the aforementioned machine learning-based industrial robot path optimization method, includes the following modules: Data acquisition and environment modeling module: Real-time acquisition of dynamic data in the industrial robot's operating environment, including the robot's current working status, workpiece position, obstacle distribution, production task requirements, and environmental parameters, and construction of a dynamic environment model. At the same time, historical path data is obtained from historical path records. Path planning model construction module: Based on machine learning algorithms, combined with historical path data and dynamic environment models, a path planning model is constructed, and multiple path candidate sets are generated through the path planning model; Path evaluation and optimization module: Evaluates each candidate path set according to preset evaluation criteria, optimizes the path evaluation results using reinforcement learning algorithms, and optimizes the path planning model based on the evaluation results to select the optimal path scheme; Path execution and dynamic adjustment module: When the robot is executing the optimal path plan, it monitors the dynamic data in the robot's working environment in real time. If the robot deviates from the predetermined path or the environment changes (such as the appearance of new obstacles), the path adjustment mechanism is automatically triggered to correct the path.

[0048] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0049] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A machine learning-based method for optimizing the path of an industrial robot, characterized in that, Includes the following steps: S1, Data Acquisition and Environment Modeling: Real-time acquisition of dynamic data in the industrial robot's working environment, including the robot's current working status, workpiece position, obstacle distribution, production task requirements, and environmental parameters, and modeling of the robot's working environment to build a dynamic environment model. At the same time, historical path data is obtained from historical path records. S2, Path planning model construction: Based on machine learning algorithms, combined with historical path data and dynamic environment models, a path planning model is constructed to capture the non-linear relationship between the path and the environment and generate multiple path candidate sets; S3, Path Evaluation and Optimization: The candidate path set is evaluated according to the preset evaluation criteria, including path length, time consumption, energy consumption and path safety. The reinforcement learning algorithm is used to optimize the path evaluation results by evaluating and iterating each path multiple times. The path planning model is then optimized based on the evaluation results to select the optimal path scheme. S4, Path Execution and Dynamic Adjustment: When the robot is executing the optimal path plan, the dynamic data in the robot's working environment is monitored in real time. If the robot deviates from the predetermined path or the environment changes (such as the appearance of new obstacles), the path adjustment mechanism is automatically triggered to correct the path.

2. The machine learning-based industrial robot path optimization method according to claim 1, characterized in that, S1 includes: S11, Real-time acquisition of dynamic data in the industrial robot's operating environment, including the robot's current working status, workpiece position, distribution of surrounding obstacles, production task requirements, and environmental parameters; The robot's current working status includes its position, speed, and acceleration. The position is measured in real time by the positioning system on the industrial robot, and the speed and acceleration are obtained through the industrial robot's inertial measurement unit. The position of the workpiece in the working environment is acquired in real time using a vision sensor; The distribution of surrounding obstacles is detected in real time using lidar and infrared sensors; Obtain production task requirements through a task management system or production scheduling system, including the target location, work path, and processing task; Environmental parameters, including temperature and humidity, are collected in real time using temperature and humidity sensors. S12, retrieve historical path data from historical path records; S13, preprocess the collected dynamic data and historical path data, including noise reduction and outlier detection.

3. The machine learning-based industrial robot path optimization method according to claim 2, characterized in that, The process of modeling the robot's operating environment and constructing a dynamic environment model includes: S14. Based on the collected data on the robot's current position, workpiece position, and obstacle distribution, construct a dynamic environment model in a spatial coordinate system. S15, based on the collected temperature, humidity and production task requirements, updates the relevant variables in the dynamic environment model to form a real-time representation of the environment's state.

4. The machine learning-based industrial robot path optimization method according to claim 3, characterized in that, S2 includes: S21, preprocess the collected historical path data and the constructed dynamic environment model; S22, uses deep neural networks to extract features from historical path data and dynamic environment models; S23, a path planning model is constructed based on a deep neural network algorithm; S24. After the model training is completed, the path planning model is used to generate multiple path candidate sets. Each path candidate takes into account the obstacle positions, workpiece positions and production task requirements in the dynamic environment model.

5. The machine learning-based industrial robot path optimization method according to claim 4, characterized in that, S3 includes: S31 sets evaluation criteria for each path candidate set, including path length, time consumption, energy consumption, and path safety; S32, based on the set evaluation criteria, performs a preliminary evaluation of each path candidate set; S33. During the evaluation process, the objectives of path optimization include minimizing path length, time consumption, and energy consumption, while maximizing path safety.

6. The machine learning-based industrial robot path optimization method according to claim 5, characterized in that, The path optimization includes: S34 uses the Q-learning algorithm as a reinforcement learning algorithm to optimize the path evaluation results; S35, through the iterative training process of reinforcement learning, evaluates and optimizes the path candidate set multiple times; S36. In each iteration of path optimization, the path planning model is optimized based on the optimal path evaluation result obtained by the reinforcement learning algorithm, and the weight values ​​and parameters in the model are adjusted.

7. The machine learning-based industrial robot path optimization method according to claim 6, characterized in that, The path optimization specifically includes: S37. In the path evaluation and optimization process, after obtaining the path evaluation results through the reinforcement learning algorithm, the path candidate set is screened and the path with the highest evaluation score is retained. S38. After selecting the optimal path, the robot executes the path and monitors the execution.

8. The machine learning-based industrial robot path optimization method according to claim 7, characterized in that, S4 includes: S41: When the robot is executing the optimal path plan, collect and monitor the dynamic data of the robot's working environment in real time; S42, If the robot deviates from the predetermined path or the environment changes, the path adjustment mechanism is automatically triggered; S43, After the path adjustment mechanism is activated, the A* algorithm is used to correct the path; S44, the robot continues to perform the task according to the corrected path, and monitors the path execution in real time. If the path deviation or new obstacles occur again, the readjustment mechanism is triggered.

9. The machine learning-based industrial robot path optimization method according to claim 8, characterized in that, During the path execution and dynamic adjustment process, the robot's task execution and environmental changes are continuously monitored, and the path planning model is updated and retrained using the new execution data.

10. A machine learning-based industrial robot path optimization method and system, used to implement the machine learning-based industrial robot path optimization method as described in any one of claims 1-9, characterized in that, Includes the following modules: Data acquisition and environment modeling module: Real-time acquisition of dynamic data in the industrial robot's operating environment, including the robot's current working status, workpiece position, obstacle distribution, production task requirements, and environmental parameters, and construction of a dynamic environment model. At the same time, historical path data is obtained from historical path records. Path planning model construction module: Based on machine learning algorithms, combined with historical path data and dynamic environment models, a path planning model is constructed, and multiple path candidate sets are generated through the path planning model; Path evaluation and optimization module: Evaluates each candidate path set according to preset evaluation criteria, optimizes the path evaluation results using reinforcement learning algorithms, and optimizes the path planning model based on the evaluation results to select the optimal path scheme; Path execution and dynamic adjustment module: When the robot is executing the optimal path plan, it monitors the dynamic data in the robot's working environment in real time. If the robot deviates from the predetermined path or the environment changes (such as the appearance of new obstacles), the path adjustment mechanism is automatically triggered to correct the path.

Images