Robot path real-time adjustment method adaptive to environment change

Abstract

The invention discloses a robot path real-time adjustment method adaptive to environment change, and relates to the technical field of robot control, and the method comprises the steps: collecting the position information, target distribution information and dynamic change information of an environment change target, and obtaining scene extraction elements; matching with a preset multi-scene control module, and determining a matched scene module; determining a path adjustment algorithm according to the matching scene module, taking the scene extraction elements as input data, performing path environment obstacle avoidance adjustment, and determining path adjustment information; and performing node matching on the current planning path by using the path adjustment information to generate a real-time adjustment path. The technical problem that in the prior art, the operation efficiency and safety of the robot are low due to the fact that the path planning precision is affected by the complex and changeable working environment is solved, and the technical effects that the robot path is adjusted in a self-adaptive mode based on the real-time environment change, and the efficiency and safety of robot operation are improved are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of robot control, and particularly to a method for real-time adjustment of a robot path adaptable to environmental changes. Background Art

[0002] Robot path planning aims to enable a robot to find the best path in a complex environment to achieve a task or goal. With the development of technology, many intelligent path planning algorithms have emerged, such as the A* algorithm, Dijkstra algorithm, RRT algorithm, etc. These algorithms realize the basic support for robot path planning through different data structures and search strategies, as well as different processing methods for environmental information.

[0003] However, due to the complex and changeable working environment of the robot, the shapes and distributions of obstacles in the environment are often random, uneven, and even constantly changing, which affects the path planning accuracy, thereby reducing the operation efficiency and safety of the robot. Summary of the Invention

[0004] The present application provides a method for real-time adjustment of a robot path adaptable to environmental changes, which is used to solve the technical problem that the operation efficiency and safety of the robot are low due to the influence of the complex and changeable working environment on the path planning accuracy in the prior art.

[0005] In a first aspect of the present application, a method for real-time adjustment of a robot path adaptable to environmental changes is provided. The method includes: collecting environmental information, including the position information, target distribution information, and dynamic change information of an environmental change target; extracting scene elements according to the position information, target distribution information, and dynamic change information to obtain scene extraction elements; using the scene extraction elements to match with a preset multi-scene control module to determine a matching scene module; determining a path adjustment algorithm according to the matching scene module, taking the scene extraction elements as input data, and performing path environment obstacle avoidance adjustment through the path adjustment algorithm to determine path adjustment information; using the path adjustment information to perform node matching on the current planned path to generate a real-time adjustment path.

[0006] One or more technical solutions provided in the present application have at least the following technical effects or advantages: A real-time robot path adjustment method adaptable to environmental changes provided by the present application relates to the technical field of robot control. By collecting environmental information, obtaining scene extraction elements, matching them with a preset multi-scene control module, determining a matching scene module, and then determining a path adjustment algorithm, using the scene extraction elements as input data to perform path environmental obstacle avoidance adjustment, determining path adjustment information, performing node matching on the current planned path, and generating a real-time adjusted path, it solves the technical problem in the prior art that the complex and changeable working environment affects the path planning accuracy, resulting in low running efficiency and safety of the robot, and realizes the adaptive adjustment of the robot path based on real-time environmental changes, improving the technical effects of the high efficiency and safety of the robot operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following described drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0008] Figure 1 It is a schematic flowchart of a real-time robot path adjustment method adaptable to environmental changes provided by an embodiment of the present application; Figure 2 It is a schematic flowchart of constructing a preset multi-scene control module in a real-time robot path adjustment method adaptable to environmental changes provided by an embodiment of the present application; Figure 3 It is a schematic flowchart of case aggregation according to scene description information in a real-time robot path adjustment method adaptable to environmental changes provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0009] The present application provides a real-time robot path adjustment method adaptable to environmental changes, which is used to solve the technical problem in the prior art that the complex and changeable working environment affects the path planning accuracy, resulting in low running efficiency and safety of the robot.

[0010] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0011] It should be noted that the terms "first", "second", etc. in the description of this application and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0012] Embodiment 1 As Figure 1 shown, the present application provides a method for real-time adjustment of a robot path adaptable to environmental changes, and the method includes: P10: Collect environmental information, including the position information of environmental change targets, target distribution information, and dynamic change information.

[0013] Optionally, collect the current working environment information of the target robot, including usually determining the exact positions of various targets in the environment, such as obstacles, other robots, destinations, etc. through lidar, ultrasonic sensors, vision sensors, etc., to obtain the position information of environmental change targets. In addition to the position information of individual targets, the robot also needs to understand the distribution of targets in the environment, including the density of targets, distribution patterns (such as uniform distribution, clustering distribution, etc.), and the relative position relationships between targets, to obtain target distribution information, so as to better understand the structure of the environment and make reasonable path planning decisions.

[0014] It should be understood that since the environment may change over time, including the movement of targets, the emergence of new targets, and the disappearance of existing targets, etc. To cope with these dynamic changes, the robot needs to detect and track these changes in real time. By continuously collecting environmental information and comparing the data differences between different time points to obtain dynamic change information, the robot can timely adjust its path planning strategy to adapt to the environmental changes.

[0015] P20: Extract scene elements according to the position information, target distribution information, and dynamic change information to obtain scene extraction elements.

[0016] Specifically, scene element extraction is performed based on the position information, target distribution information, and dynamic change information. First, the collected position information, target distribution information, and dynamic change information are integrated, that is, the data of multiple sensors are fused to form a comprehensive environmental description. Next, key scene features are extracted from the integrated environmental data. The key scene features are closely related to the path planning and navigation tasks of the robot, such as the shape, size, and position of obstacles, as well as the relative relationship with the robot, the speed, acceleration, and direction of dynamic targets, and the overall structure of the environment, such as corridors, rooms, open areas, etc.

[0017] Furthermore, the extracted key scene features are classified and recognized so that the robot can better understand the nature and characteristics of the environment, and thus make more appropriate path planning decisions. For example, the environment can be divided into "high-risk areas" and "low-risk areas", or obstacles can be divided into two categories: "traversable" and "non-traversable". Finally, the classified scene features are integrated into a set of scene extraction elements, which serve as the basis for the robot to perform subsequent path planning and navigation tasks.

[0018] P30: Match the scene extraction elements with the preset multi-scene control module to determine the matching scene module.

[0019] Furthermore, step P30 of the embodiment of the present application further includes: P31: Traverse according to the scene extraction elements and the scene description elements of the module to determine the element matching result and the element coverage rate; P32: Determine the matching scene module according to the element coverage rate and the element matching result. The matching scene module is a module with full element matching and the element coverage rate reaching full coverage; P33: When the matching scene module is one module, use the matching scene module as the only execution module.

[0020] It should be understood that the scene extraction elements are used to match with the preset multi-scene control module to determine which preset control module the current environment most conforms to. The preset multi-scene control module is usually designed based on past experience or expert knowledge and is used to handle different types of environmental scenes.

[0021] Specifically, traverse each preset scenario control module, and compare the scenario description elements inside it with the scenario extraction elements of the current environment. Determine whether the elements in each module match the extracted elements, thereby determining the element matching result, and record the ratio of the number of matching elements to the total number of elements in the module to determine the element coverage rate. Further, based on the element matching result and the element coverage rate, determine the best matching scenario module. The definition of the matching scenario module is: the scenario extraction elements and the scenario description elements in this module are full element matches, that is, perfect matches, and the element coverage rate reaches 100%, that is, full coverage, which means that this module can perfectly describe the current environmental situation.

[0022] Further, when there is one matching scenario module, that is, the robot only finds one module that completely matches the scenario extraction elements and has a full element coverage rate, select this module as the only execution module, and perform subsequent path planning and adjustment based on this module. If no such module is found, the robot may need to adopt other strategies, such as selecting the module with the highest element matching degree as an alternative, or triggering an exception handling process to deal with the non-matching situation.

[0023] Further, as Figure 2 shown, before using the scenario extraction elements to match with the preset multi-scenario control module, the embodiment of the present application further includes step P30a, and step P30a further includes: P31a: Construct a historical case database, which includes scenario description information, path planning results, and path anomaly evaluation; P32a: Aggregate cases according to the scenario description information to construct multi-scene blocks, and the multi-scene blocks are case data sets with the same scenario description information; P33a: Based on the multi-scene blocks, respectively perform path planning relationship learning with the path planning results and path anomaly evaluation, fit the relationship between the scenario description elements and path planning, construct a path algorithm, and at the same time extract based on the relationship between the scenario description elements and path planning to determine path planning constraint information; P34a: Associate and map the scenario description elements, path planning constraint information, and path algorithms to generate control modules for each scenario, and integrate the control modules of all scenarios to construct the preset multi-scenario control module, where each path algorithm in the preset multi-scenario control module is used as the path adjustment algorithm of the module.

[0024] Optionally, before formally matching the scene extraction elements with the preset multi-scene control module, it is necessary to first construct the preset multi-scene control module. Specifically, first construct a historical case database, which contains the historical data of the target robot when performing path planning tasks in various environmental scenarios before. Each case contains scene description information, such as environmental features, obstacle distribution, etc., path planning results, such as the planned path trajectory, nodes, etc., and path anomaly evaluation, such as whether a collision occurred, whether the target was reached, etc.

[0025] Furthermore, aggregate the cases according to the scene description information in the historical case database. Classify the cases with the same or similar scene description information into the same multi-scene block. For example, the street scene has dynamic avoidance objects such as flowing crowds and vehicle traffic, and the village scene has avoidance objects such as fields that cannot be trampled. Each multi-scene block is a case data set containing the same or similar environmental features.

[0026] Furthermore, perform path planning relationship learning for the multi-scene blocks respectively with the path planning results and path anomaly evaluation, including using machine learning or statistical learning methods to fit the relationship between them based on scene description elements (such as obstacle positions, terrain features, etc.) and path planning results (such as the planned path trajectory), and construct a path algorithm. At the same time, based on the relationship between the scene description elements and path planning, extract the scene elements with the greatest influence as path planning constraint information, and the path planning constraint information reflects the rules and restrictions that need to be observed during path planning in different scenarios.

[0027] Furthermore, perform an associated mapping of the scene description elements, path planning constraint information, and path algorithm. Specifically, the robot associates the scene description elements in each multi-scene block with the corresponding path planning constraint information and path algorithm to generate a control module for this scene. These control modules contain key information such as scene recognition, path planning constraints, and path adjustment algorithms. Integrate the control modules of all scenes to construct a complete preset multi-scene control module. In the preset multi-scene control module, each path algorithm serves as the path adjustment algorithm of the corresponding module, which is used to guide the path planning and adjustment of the robot when the corresponding scene is matched.

[0028] Furthermore, as Figure 3 shown, step P32a of the embodiment of the present application further includes: P32-1a: Extract the scene description information according to a preset dimension, where the preset dimension includes environmental position information, terrain features, avoidance object recognition features, and avoidance object dynamic attributes; P32-2a: Vectorize the scene description information extracted from the preset dimensions to determine the scene feature vector. P32-3a: Aggregate the scene feature vectors based on the preset dimensions respectively, extract the scenes that meet the aggregation requirements for all preset dimensions, and construct the multi-scene block.

[0029] Optionally, cases with similar scene description information are aggregated together to form a multi-scene block. Specifically, preset key dimensions are extracted from the scene description information of each case. The preset dimensions include environmental location information (such as obstacle location, robot initial position, etc.), terrain features (such as ground material, slope, etc.), avoidance object recognition features (such as the shape and size of obstacles), and avoidance object dynamic attributes (such as moving speed, acceleration, etc.), which can comprehensively describe the key features of the environmental scene.

[0030] Furthermore, the scene description information of the extracted preset dimensions is converted into a numerical scene feature vector, so that the scene description information between different cases can be numerically compared, which is convenient for subsequent scene clustering or classification. Each dimension of the scene feature vector corresponds to a preset dimension, and its value reflects the specific performance of this dimension in the scene.

[0031] Furthermore, cluster the scene feature vectors based on each preset dimension to find scenes with similar features. To ensure the accuracy of aggregation, aggregation requirements are preset in advance, for example, requiring the similarity of all preset dimensions to reach a certain threshold. And the scenes that meet the similarity threshold are classified into a multi-scene block. The scenes within each block have high similarity in the preset dimensions, so they can be regarded as the same type of scenes. The multi-scene block can be used as the input data for subsequent path planning relationship learning to help the robot learn path planning strategies in different scenes.

[0032] Furthermore, step P33a of the embodiment of the present application further includes: P33-1a: Configure the dynamic window size according to the scene description elements of the multi-scene block, and the dynamic window size matches the safe operation space in the scene; P33-2a: Perform case screening and path annotation according to the path planning result and path anomaly evaluation. The case screening is the path planning information without abnormal paths and perform path annotation; P33-3a: Perform speed space sampling in the screened cases based on the dynamic window to generate multiple groups of sampling data, including speed, direction, and duration; P33-4a: Learn according to the multiple groups of sampling data to determine the relationship between the scene description elements and path planning, and construct the path algorithm.

[0033] Specifically, by analyzing cases in multi-scenario blocks, the relationship between scene description elements and path planning is learned, thereby constructing path algorithms for different scenarios. First, according to the scene description elements of the multi-scenario blocks, a dynamic window size is configured. The dynamic window size represents the safe operation space that the robot needs to consider when performing path planning in the current scene, and can be adjusted according to factors such as the distribution of obstacles in the scene and the speed of moving objects, so as to ensure that the robot can avoid collisions and maintain safety when planning the path.

[0034] Furthermore, according to the path planning results and path anomaly evaluation, cases in the multi-scenario blocks are screened. This includes determining whether there are anomalies in the path planning results, such as whether a collision has occurred, whether the target has not been reached within the specified time, etc. The path planning information with no abnormal paths is screened out, and these cases are path-labeled. The path labeling is to convert the path planning results into a format that can be understood by machine learning algorithms. For example, the path trajectory is converted into a series of coordinate points, and the position and speed of the avoidance objects are converted into numerical features, etc.

[0035] Furthermore, speed space sampling is performed on the screened cases. The speed space sampling refers to generating multiple sets of sampling data of speed, direction, and duration within the dynamic window of the robot according to certain rules. These sampling data represent the possible movement modes of the robot in the current scene. Through sampling, the robot can generate multiple sets of possible path planning schemes.

[0036] Furthermore, machine learning algorithms, such as supervised learning, reinforcement learning, etc., are used to analyze and learn the multiple sets of sampling data, and determine the relationship between the scene description elements and the path planning results. Through analysis, the robot can learn the path planning strategies that should be adopted in different scenarios, and how to adjust the speed and direction of the robot to avoid obstacles and reach the destination safely. Based on the learning results, path algorithms for different scenarios are constructed as part of the preset multi-scenario control module to guide the path planning of the robot in the actual environment.

[0037] Furthermore, step P33-4a of the embodiment of the present application further includes: P33-41a: According to the multiple sets of sampling data, case positioning is performed to determine the corresponding relationship between the sampling data and the position information of the avoidance objects, target distribution, and dynamic changes, and training data is constructed; P33-42a: Based on the training data, the relationship between the scene description elements and path planning is learned, and the model is optimized by using the learning result parameters and the loss values of the multiple sets of sampling data. The learning result parameters that meet the loss value convergence requirement are determined, and the relationship between the scene description elements and path planning is obtained.

[0038] It should be understood that by learning to determine the internal relationship between the scene description elements and the path planning results, an algorithm can be constructed to guide the robot to perform path planning in different scenarios. First, multiple sets of sampling data will be used for case location to determine the corresponding relationship between each set of sampling data (including speed, direction, duration) and the position information of the avoidance objects in the scene, the distribution of the targets, and the dynamic changes of the scene. Thus, the sampling data is combined with the physical elements and dynamic changes in the actual scene to construct training data.

[0039] Furthermore, the training data is used to train a machine learning model to learn the relationship between the scene description elements and the path planning results. Exemplarily, a suitable machine learning algorithm can be selected, such as a neural network, a support vector machine, etc., and the training data is input into the model for training. During the training process, the model will attempt to predict the best path planning result of the robot under a given scene description element, and a loss function is used to calculate the difference between the model prediction result and the true path planning result, that is, the loss value. The smaller the loss value, the closer the prediction result of the model is to the real situation.

[0040] Moreover, in order to optimize the model, the robot will utilize the learning result parameters, that is, the weights and bias terms in the model, and the loss values of multiple sets of sampling data to perform gradient descent or other optimization algorithms. By iteratively adjusting the learning result parameters to minimize the loss value, the prediction accuracy of the model is gradually improved. When the loss value of the model meets the convergence requirement, that is, the loss value no longer significantly decreases or reaches a preset minimum value, it is considered that the model has been trained, and the relationship between the scene description elements and the path planning is obtained. The relationship between the scene description elements and the path planning can be utilized to construct a path algorithm to generate the best path planning result of the robot in this scene according to the given scene description elements.

[0041] P40: Determine the path adjustment algorithm according to the matching scene module, use the scene extraction elements as input data, and perform path environment obstacle avoidance adjustment through the path adjustment algorithm to determine the path adjustment information.

[0042] Optionally, according to the path adjustment algorithm previously determined by the matching scene module, use the scene extraction elements as input data, and through the path adjustment algorithm, perform re-planning or local adjustment of the path. In this process, the algorithm will consider the position, shape, dynamic attributes, etc. of the obstacles, as well as the kinematic constraints of the robot, such as the maximum speed, acceleration, etc., to generate a new, collision-free path. After the path adjustment is completed, the robot will output the path adjustment information, including the new path trajectory, speed curve, direction instruction, etc., to guide the robot to move along the new path to ensure the safety and efficiency of the entire movement process.

[0043] Further, before performing path environment obstacle avoidance adjustment through the path adjustment algorithm, the embodiment of the present application further includes step P40a, and step P40a further includes: P41a: Perform element matching on the scene extraction elements based on the path planning constraint information to determine the constraint elements; P42a: Determine whether the constraint elements meet the preset requirements. When they meet, perform path environment obstacle avoidance path calculation through the path adjustment algorithm; P43a: When they do not meet, generate a constraint collection request, which includes constraint element collection requirements; P44a: Generate a collection instruction from the constraint collection request and feedback it to the robot, and perform positioning incremental collection according to the constraint element collection requirements.

[0044] Specifically, before performing path environment obstacle avoidance adjustment through the path adjustment algorithm, it is necessary to determine various constraint conditions that need to be met during path planning. First, obtain the constraint information of path planning. These constraints may include the physical limitations of the robot (such as maximum speed, maximum acceleration, steering angle, etc.), environmental limitations (such as restricted access to specific areas, restrictions during specific time periods, etc.), and task limitations (such as arrival time, energy consumption, etc.). Match the constraint information with the scene extraction elements to extract the constrained elements, that is, the constraint elements.

[0045] Further, after determining the constraint elements, determine whether the constraint elements meet the preset requirements. The preset requirements may include thresholds, ranges, etc. of the constraint elements, such as dimensional accuracy, position accuracy range, etc. If all constraint elements meet the preset requirements, continue to perform path environment obstacle avoidance path calculation, that is, call the path adjustment algorithm to perform re - planning or local adjustment of the path.

[0046] If the constraint elements do not meet the preset requirements, for example, problems such as data loss or low data collection accuracy occur, then generate a constraint collection request. The constraint collection request contains specific constraint element collection requirements, that is, the additional information that the robot needs to collect. Convert the constraint collection request into a specific collection instruction and send it to the robot. The robot will perform positioning incremental collection according to the collection instruction, that is, collect more environmental information or sensor data to meet the constraint conditions of path planning, further enhance the robot's environmental adaptability, and improve the accuracy and reliability of path planning.

[0047] P50: Perform node matching on the current planned path using the path adjustment information to generate a real - time adjustment path.

[0048] Optionally, use the path adjustment information to perform node matching on the current planned path and generate a real-time adjusted path. First, parse the path adjustment information to obtain the specific instructions and parameters contained therein, including the coordinates of new path nodes, speed requirements, steering angles, etc. Next, perform a detailed analysis of the current planned path, including identifying key nodes on the path, such as the starting point, ending point, turning points, etc., and the attributes of each node, such as position, speed, direction, etc.

[0049] Furthermore, according to the instructions in the path adjustment information, perform node matching on the current planned path, find the path nodes corresponding to the adjustment information, and determine the adjustments that need to be made to these nodes, such as changing position, speed, direction, etc. After the node matching and adjustment are completed, generate a real-time adjusted path based on the adjusted node information. The real-time adjusted path will consider all relevant constraint conditions, such as physical limitations, environmental limitations, task limitations, etc., and optimize the movement efficiency and safety of the robot as much as possible.

[0050] Furthermore, the embodiment of the present application further includes step P60, and step P60 further includes: P61: When there are multiple matching scenario modules, configure the fusion weights of each module based on the element matching results of the path planning constraint information; P62: Perform path selection fusion of multiple matching scenario modules based on the fusion weights. Among them, obtain the output paths of each matching scenario module and determine the conflicting path nodes; P63: Select the path planning information of the conflicting path nodes according to the fusion weights. When the fusion weights are the same, select the path with the largest avoidance distance, and obtain a fused path based on path smoothing fitting.

[0051] It should be understood that in a more complex scenario, there may be multiple matching scenario modules working simultaneously, and each module outputs a recommended path based on its specific scenario understanding and algorithm logic. To ensure that the robot can select the optimal and consistent path for navigation, it is first necessary to configure a fusion weight for each module. The fusion weight can be determined based on the element matching results of the path planning constraint information. The higher the similarity between the element information in the module and the element information of the path planning constraint information, the higher the corresponding weight, and vice versa.

[0052] Further, collect the output paths of each matching scenario module. These paths are generated based on their respective understanding of the environment and algorithmic logic. Analyze the output paths to identify conflicting path nodes, which may occur at positions where two or more paths cross, overlap, or interfere with each other. After determining the conflicting path nodes, select the path planning information for these nodes according to the previously configured fusion weights. The path planning information of the module with a higher weight will be given priority at the conflicting nodes.

[0053] If the fusion weights of multiple modules are the same at the conflicting nodes, select the path with the largest avoidance distance. The avoidance distance refers to the minimum distance that the robot needs to move from the current position to avoid obstacles. Selecting the path with the largest avoidance distance can ensure that the robot has a larger operating space and higher safety when encountering obstacles. Finally, based on the selected path planning information, obtain the final fused path through a path smoothing and fitting algorithm. Exemplarily, the smoothing and fitting formula can be: , where is the smoothing curve, is the basis function, are the coefficients of the control points. The control points are the path connection endpoints that do not meet the smoothness requirements. Obtain the curve values of the connection line segments of each endpoint. With the goal of the same first derivative of the curve values, perform a first derivative approximation operation using the least squares method to obtain the coefficients of the control points. Through path smoothing and fitting, unnecessary jitters and sharp corners in the path can be eliminated, enabling the robot to move more smoothly and stably along the fused path.

[0054] In summary, the embodiments of the present application at least have the following technical effects: The present application obtains the scenario extraction elements through the position information, target distribution information, and dynamic change information of the environmental change target, matches them with the preset multi-scenario control module to determine the matching scenario module, and then determines the path adjustment algorithm. Using the scenario extraction elements as input data, perform path environment obstacle avoidance adjustment to determine the path adjustment information, perform node matching on the current planned path, and generate a real-time adjustment path.

[0055] It achieves the technical effect of adaptively adjusting the robot path based on real-time environmental changes, improving the efficiency and safety of robot operation.

[0056] It should be noted that the above sequence of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of this specification. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0057] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0058] This specification and the drawings are merely exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A method for real-time adjustment of robot paths that is adaptive to environmental changes, characterized in that: The method for real-time adjustment of robot paths that adapt to environmental changes includes: Collect environmental information, including location information, target distribution information, and dynamic change information of environmental change targets; Extract scene elements according to the location information, target distribution information, and dynamic change information to obtain scene extraction elements; Using the scene extraction elements to match with the preset multi-scene control module to determine the matching scene module; Determine a path adjustment algorithm according to the matching scene module, use the scene extraction elements as input data, perform path environment obstacle avoidance adjustment through the path adjustment algorithm, and determine path adjustment information; The path adjustment information is used to perform node matching on the current planned path to generate a real-time adjusted path.

2. The method for real-time robot path adjustment in response to environmental changes as claimed in claim 1, characterized in that: The scene extraction elements are used to match the preset multi-scene control module, which includes: Build a historical case database, which includes scene description information, path planning results, and path anomaly evaluation; Aggregate cases according to the scene description information to construct multi-scene blocks, where the multi-scene blocks are case data sets with the same scene description information; Based on the multiple scene blocks, the path planning results and the path anomaly evaluation are used to respectively learn the path planning relationship, fit the relationship between the scene description elements and the path planning, and construct a path algorithm. At the same time, based on the relationship between the scene description elements and the path planning, the path planning constraint information is extracted to determine the path planning constraint information; The scene description elements, path planning constraint information, and path algorithm are associated and mapped to generate control modules for each scene, and the control modules of all scenes are integrated to construct the preset multi-scene control module, wherein each path algorithm in the preset multi-scene control module serves as the path adjustment algorithm of the module.

3. The method for real-time robot path adjustment in response to environmental changes as claimed in claim 2, characterized in that: Case aggregation is performed according to the scenario description information, including: Extracting the scene description information according to preset dimensions, wherein the preset dimensions include environmental location information, terrain features, avoidance object identification features, and avoidance object dynamic attributes; Vectorizing the scene description information extracted from the preset dimension to determine a scene feature vector; Based on the preset dimensions, scene feature vectors are aggregated respectively, scenes that meet the aggregation requirements in the preset dimensions are extracted, and the multi-scene blocks are constructed.

4. The method for real-time robot path adjustment in response to environmental changes as claimed in claim 2, characterized in that: Based on the multiple scene blocks, path planning relationship learning is performed using the path planning results and path anomaly evaluation respectively, the relationship between the scene description elements and the path planning is fitted, and a path algorithm is constructed, including: According to the scene description elements of the multiple scene blocks, a dynamic window size is configured, wherein the dynamic window size matches a safe operation space in the scene; Perform case screening and path marking according to the path planning results and path anomaly evaluation, wherein the case screening is path planning information without abnormal paths and path marking; Based on the dynamic window, speed space sampling is performed in the screening cases to generate multiple sets of sampling data, including speed, direction, and duration; Learning is performed based on the multiple groups of sampled data to determine the relationship between the scene description elements and path planning, and to construct the path algorithm.

5. The method for real-time robot path adjustment in response to environmental changes as claimed in claim 4, characterized in that: Learning according to the plurality of sets of sampled data to determine the relationship between the scene description elements and the path planning includes: Based on the multiple groups of sampling data, case positioning is performed to determine the corresponding relationship between the sampling data and the location information, target distribution, and dynamic changes of the avoidance object, and training data is constructed; Based on the training data, the relationship between the scene description elements and the path planning is learned, and the model is optimized using the learning result parameters and the loss values ​​of multiple groups of sampling data to determine the learning result parameters that meet the loss value convergence requirements, and obtain the relationship between the scene description elements and the path planning.

6. The method for real-time robot path adjustment in response to environmental changes as claimed in claim 2, characterized in that: The scene extraction elements are used as input data, and the path environment obstacle avoidance adjustment is performed through the path adjustment algorithm, which includes: Performing element matching on the scene extraction elements based on the path planning constraint information to determine constraint elements; Determine whether the constraint element meets the preset requirements, and if so, perform path environment obstacle avoidance calculation through the path adjustment algorithm; When it is not satisfied, a constraint collection request is generated, which includes constraint element collection requirements; The constraint collection request generates a collection instruction which is fed back to the robot, and positioning incremental collection is performed according to the constraint element collection requirements.

7. The method for real-time robot path adjustment in response to environmental changes as claimed in claim 2, characterized in that: Using the scene extraction elements to match the preset multi-scene control module to determine the matching scene module includes: Traversing the scene extraction elements and the scene description elements of the module to determine the element matching results and element coverage; Determine a matching scene module according to the element coverage rate and the element matching result, wherein the matching scene module is a module in which all elements are matched and the element coverage rate reaches full coverage; When the matching scene module is one module, the matching scene module is used as the only execution module.

8. The method for real-time robot path adjustment in response to environmental changes as claimed in claim 7, characterized in that: The method for real-time adjustment of robot paths that adapt to environmental changes also includes: When the matching scene module is a plurality of modules, configuring the fusion weight of each module based on the element matching result of the path planning constraint information; Based on the fusion weights, path selection fusion of multiple matching scene modules is performed, wherein the output paths of each matching scene module are obtained and conflicting path nodes are determined; The path planning information of the conflicting path nodes is selected according to the fusion weights, and when the fusion weights are the same, the path with the largest avoidance distance is selected, and the fusion path is obtained based on path smoothing fitting.

Images