Adaptive behavior planning system for sensing robot

Abstract

The invention provides an adaptive behavior planning system for a sensing robot, which relates to the technical field of sensing robots and is characterized in that multiple sensors are extracted based on the sensing robot, environment sensing is carried out, an environment sensing data set is obtained, the environment sensing data set is preprocessed, and dimension reduction is carried out on the environment sensing data set according to a preprocessing result; the method comprises the following steps: generating an environment perception dimension reduction data array, carrying out behavior analysis on a perception robot according to the environment perception dimension reduction data array, combining multiple pieces of behavior information, constructing a behavior tree, carrying out strategy training through the behavior tree in combination with the environment perception dimension reduction data array, and executing a behavior strategy to carry out adaptive behavior planning on the perception robot. The technical problems of low behavior planning efficiency and poor adaptability of a sensing robot in a complex environment in the prior art are solved. The technical effects of improving the behavior planning efficiency of the sensing robot, enhancing the complex environment adaptability and realizing intelligent behavior decision making are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of perceptual robots, and in particular to an adaptive behavior planning system for perceptual robots. Background Art

[0002] In the field of perceptual robots, behavior planning is one of the core technologies for achieving robot autonomy. The goal of behavior planning is to enable the robot to generate reasonable behavior strategies in real time based on the perceived environmental information to complete the predetermined task. However, most existing behavior planning methods are based on fixed rules or simple path planning algorithms, which have defects such as poor adaptability to environmental changes and low computational efficiency. In addition, these methods usually require a lot of computing resources and time for path planning and decision-making, which limits the real-time performance and efficiency of perceptual robots in practical applications.

[0003] The existing technology has technical problems such as low efficiency and poor adaptability of the behavior planning of perception robots in complex environments. Summary of the invention

[0004] The present application provides an adaptive behavior planning system for a perception robot, which is used to solve the technical problems in the prior art of low efficiency and poor adaptability of the behavior planning of the perception robot in a complex environment.

[0005] In view of the above problems, the present application provides an adaptive behavior planning system for a perception robot, the system comprising: an environment perception module, which is used to extract multiple sensors based on the perception robot, perform environment perception through the multiple sensors, and obtain an environment perception data set; a data dimension reduction module, which is used to preprocess the environment perception data set, reduce the dimension of the environment perception data set according to the preprocessing result, and generate an environment perception reduced dimension data array; a behavior analysis module, which is used to perform behavior analysis on the perception robot according to the environment perception reduced dimension data array, and generate multiple behavior information; a behavior tree construction module, which is used to combine the multiple behavior information to construct a behavior tree; a behavior planning module, which is used to perform strategy training through the behavior tree in combination with the environment perception reduced dimension data array, generate a behavior strategy, and execute the behavior strategy to perform adaptive behavior planning for the perception robot.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The environment perception module is used to extract multiple sensors based on the perception robot, perform environment perception through the multiple sensors, and obtain an environment perception data set. The data dimension reduction module is used to preprocess the environment perception data set, reduce the dimension of the environment perception data set according to the preprocessing result, and generate an environment perception reduced dimension data array. The behavior analysis module is used to perform behavior analysis on the perception robot according to the environment perception reduced dimension data array to generate multiple behavior information. The behavior tree construction module is used to combine the multiple behavior information to construct a behavior tree. The behavior planning module is used to perform strategy training through the behavior tree combined with the environment perception reduced dimension data array, generate a behavior strategy, and execute the behavior strategy to perform adaptive behavior planning for the perception robot. The technical effects of improving the efficiency of the behavior planning of the perception robot, enhancing the adaptability to complex environments, and realizing intelligent behavior decision-making are achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 A schematic diagram of the structure of the adaptive behavior planning system for a perception robot provided in this application;

[0010] Figure 2 A schematic diagram of a process for generating multiple behavior information in an adaptive behavior planning system for a perception robot provided in this application.

[0011] Explanation of the reference numerals: environment perception module 11, data dimension reduction module 12, behavior analysis module 13, behavior tree construction module 14, behavior planning module 15. DETAILED DESCRIPTION

[0012] This application provides an adaptive behavior planning system for a perception robot, which is used to solve the technical problems of low efficiency and poor adaptability of the behavior planning of the perception robot in a complex environment in the prior art. The system achieves the technical effects of improving the efficiency of the behavior planning of the perception robot, enhancing the adaptability to complex environments, and realizing intelligent behavior decision-making.

[0013] Below, the technical solutions in the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments described herein. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the accompanying drawings, rather than all of them.

[0014] like Figure 1 As shown, the present application provides an adaptive behavior planning system for a perception robot, the system comprising:

[0015] The environment perception module 11 is used to extract multiple sensors based on the perception robot, perform environment perception through the multiple sensors, and obtain an environment perception data set.

[0016] Specifically, the environment perception module 11 is one of the core components of the perception robot, which implements data collection and processing functions based on multiple sensors to provide comprehensive perception of the surrounding environment. The environment perception module 11 extracts multiple sensors based on the perception robot. Multiple sensors refer to a variety of sensor devices integrated by the perception robot, such as cameras, ultrasonic sensors, laser radars, infrared sensors, etc., which are used to perceive different types of information in the environment, such as vision, sound, distance, etc. The multiple sensors integrated on the perception robot can not only cover different information levels such as vision, sound, distance, etc., but also improve the perception accuracy and anti-interference ability of the perception robot by fusing various sensor data. In the specific implementation process, the environment perception module 11 performs environmental perception through each sensor, and extracts the corresponding data signals, such as images, distances or infrared radiation, to form an environment perception data set. The environment perception data set refers to a data set collected by multiple sensors, which contains multiple environmental features. The environment perception module 11 realizes efficient perception of the environment and provides a basic guarantee for the behavior planning of the perception robot.

[0017] The data dimension reduction module 12 is used to preprocess the environment perception data set, reduce the dimension of the environment perception data set according to the preprocessing result, and generate an environment perception reduced dimension data array.

[0018] Specifically, the environmental perception data set obtained by the environmental perception module 11 contains multidimensional data from multiple sensors, and there is a large amount of redundant information or noise. The data dimension reduction module 12 receives the environmental perception data set obtained from the environmental perception module 11, and preprocesses the environmental perception data set. The preprocessing includes signal filtering and noise reduction to remove random interference and useless information that may be received during the sensor acquisition process, improve data quality, and ensure the authenticity and validity of the perception data. After the data preprocessing is completed, the data dimension reduction module 12 selects a suitable dimension reduction algorithm according to the characteristics of the data, such as principal component analysis, linear discriminant analysis, etc., to reduce the dimension. The reduced data is mapped to a new low-dimensional space, and the low-dimensional space represents the core information of the original environmental perception data, and generates an environmental perception dimension reduction data array. The environmental perception dimension reduction data array is the data result after the dimension reduction of the environmental perception data set, and retains the core information of the data. The environmental perception data set is processed by the data dimension reduction module 12, the dimension of the data is reduced and the most informative features are retained, so that subsequent analysis and processing are more efficient.

[0019] Furthermore, the environmental perception data set is preprocessed, and the data dimension reduction module 12 is also used to: perform multidimensional analysis based on the environmental perception data set to determine the multidimensional environmental state signal; perform wavelet decomposition on the multidimensional environmental state signal to obtain the wavelet coefficients of the multidimensional environmental state signal; perform threshold quantization based on the wavelet coefficients of the multidimensional environmental state signal to determine the wavelet selection threshold of the multidimensional environmental state signal; intercept the wavelet coefficients of the multidimensional environmental state signal according to the wavelet selection threshold of the multidimensional environmental state signal, set the noise signal less than the wavelet selection threshold of the multidimensional environmental state signal to zero, and obtain the valid signal information greater than the wavelet selection threshold of the multidimensional environmental state signal; filter and reconstruct the valid signal information to obtain the preprocessing result.

[0020] Specifically, in the data preprocessing stage, the data dimension reduction module 12 performs multidimensional analysis on the environmental perception data set through multidimensional data analysis techniques, such as cluster analysis, association rule mining, etc. Through the multidimensional analysis results, a multidimensional environmental state signal is obtained, and the multidimensional environmental state signal includes different dimensional features of the perception robot's perception environment, such as the distance of the robot's location, the state of surrounding objects, temperature changes, etc. Then, the multidimensional environmental state signal is processed using wavelet decomposition technology. Wavelet decomposition is an effective signal processing method that can decompose complex signals into sub-signals of different frequencies. These sub-signals can reveal different scale features in the signal. Wavelet decomposition converts the multidimensional environmental state signal into wavelet coefficients to form multidimensional environmental state signal wavelet coefficients. Each wavelet coefficient in the multidimensional environmental state signal wavelet coefficients represents the change characteristics of the environmental state signal at a certain specific scale. Then, the obtained multidimensional environmental state signal wavelet coefficients are threshold quantized, and the multidimensional environmental state signal wavelet selection threshold is determined according to the distribution and importance of the wavelet coefficients. The multidimensional environmental state signal is used to distinguish between valid signals and noise. According to the set multi-dimensional environmental state signal wavelet selection threshold, the multi-dimensional environmental state signal wavelet coefficients are intercepted, and the environmental state signal wavelet coefficients that are less than the wavelet selection threshold are set to zero. The environmental state signal wavelet coefficients that are set to zero are non-important signals caused by external interference or sensor accuracy limitations. By setting these signals to zero, noise removal is achieved. The remaining environmental state signal wavelet coefficients that are greater than the multi-dimensional environmental state signal wavelet selection threshold are valid signal information. Finally, the extracted valid signal information is filtered and reconstructed. The reconstruction process smoothes the signal through a filtering method, removes the remaining noise, and restores the true characteristics of the signal to obtain the preprocessing result. Filtering methods include Gaussian filtering, mean filtering, Kalman filtering and other technologies. For example, through Gaussian filtering, the following formula is used for filtering, Among them: μ is the mean of x, σ is the standard deviation of x, which indicates the degree of data dispersion and determines the shape of the distribution. The smaller σ is, the more concentrated the distribution is and the thinner the shape is. e is the type of noise in the process. Through processing, the data dimension reduction module 12 converts the original multi-dimensional perception data into a more streamlined and reliable signal set, providing accurate data support for the robot's subsequent behavior planning and decision-making.

[0021] Furthermore, the environmental perception data set is dimensionally reduced according to the preprocessing result to generate an environmental perception reduced dimension data array, and the data dimension reduction module 12 is also used to: perform distribution calculation on the environmental perception data set based on the wavelet selection threshold of the multi-dimensional environmental state signal to obtain a data distribution probability array; construct a similarity probability distribution function, perform dimension reduction processing on the dot distribution of the data distribution probability array to generate a data dimension reduction result; and construct the environmental perception reduced dimension data array based on the data dimension reduction result.

[0022] Specifically, the multidimensional environmental state signal wavelet selection threshold generated in the preprocessing stage is used to analyze and calculate the environmental perception data set. By calculating the distribution characteristics of the environmental perception data set, a data distribution probability array is obtained, which describes the distribution characteristics of the data in the multidimensional space in the form of mathematical probability, and intuitively reflects the concentration and distribution pattern of the data. The signal screened out based on the multidimensional environmental state signal wavelet selection threshold more accurately reflects the actual effective distribution of the environmental perception data, eliminating the interference of noise data. After obtaining the data distribution probability array, based on the Euclidean distance, Manhattan distance or cosine similarity calculation method, for each pair of data points in the environmental perception data set, the similarity value between the two is calculated, all similarity values ​​are arranged, and a similarity matrix is ​​generated. The elements of the similarity matrix are normalized so that the sum of the elements in each row is 1, and the normalized similarity matrix, that is, the similarity probability distribution function, is obtained. The similarity probability distribution function reflects the similarity distribution of each data point relative to other data points. Based on the constructed similarity probability distribution function, principal component analysis, t-SNE or other nonlinear dimensionality reduction methods are used to reduce the dimensionality of the dot distribution of the data distribution probability array, map the data points in the high-dimensional space to the low-dimensional space, and generate a data dimensionality reduction result, which is a simplified low-dimensional data set that can still accurately reflect the characteristics of the original data. Finally, based on the data dimensionality reduction result, the dimensionality reduction data is further structured and organized to generate an environmental perception dimensionality reduction data array, which is a standardized representation of the dimensionality reduction data, contains the core feature information of the environment, and has low dimensionality and high information content characteristics. The data dimensionality reduction module 12 generates a concise but information-intensive environmental perception dimensionality reduction data array by performing dimensionality reduction operations on the environmental perception data set, ensuring that key features are retained while compressing the data dimension, and providing accurate and efficient data input for the subsequent behavior planning of the perception robot.

[0023] The behavior analysis module 13 is used to perform behavior analysis on the perception robot according to the environmental perception dimension reduction data array to generate multiple behavior information.

[0024] Specifically, the behavior analysis module 13 receives the reduced dimension environmental perception data array from the data dimension reduction module 12, and performs behavior analysis on the perception robot in combination with the task objectives of the perception robot to generate a plurality of behavior information, wherein the behavior information refers to the specific behavior instructions of the robot in the current environment, including the behavior type, execution conditions, priority, etc., which are used to guide the robot behavior. The behavior analysis module 13 can efficiently analyze the environmental perception data and task objectives, generate accurate and reliable behavior information, and thus improve the accuracy and reliability of the adaptive behavior planning of the perception robot.

[0025] In one embodiment, Figure 2 As shown, the perception robot is subjected to behavior analysis according to the environmental perception dimensionality reduction data array to generate multiple behavior information, and the behavior analysis module 13 is also used to: retrieve the target task, perform steering analysis according to the target task based on the environmental perception dimensionality reduction data array, and determine the target direction information; perform path analysis according to the target direction information based on the environmental perception dimensionality reduction data array to obtain a target path map; perform obstacle detection according to the target direction information, the target path map and the environmental perception dimensionality reduction data array to obtain behavior obstacle information; classify according to the target direction information, the target path map and the behavior obstacle information to determine multiple behavior patterns; and add the multiple behavior patterns to the multiple behavior information.

[0026] Specifically, the behavior analysis module 13 first retrieves the current target task of the perception robot, which refers to a specific behavior task that the perception robot needs to complete, such as reaching a specified location, path navigation, or avoiding dynamic obstacles. Based on the target task, combined with the environmental perception dimensionality reduction data array, a steering analysis is performed. The steering analysis refers to calculating the optimal moving direction based on the relative position between the robot and the target to ensure that the robot can efficiently move toward the target position. The steering analysis determines the angular deviation between the current robot position and the target position through geometric methods, and combines environmental data to avoid possible collision paths, thereby generating target direction information. The target direction information is used to indicate the direction the robot should face next, including target azimuth, distance, etc., and is an important input for path planning.

[0027] According to the target direction information, combined with the environmental perception dimension reduction data array, the perception robot is analyzed from the current starting point to the target point, and the robot's travel path in the current environment is calculated to ensure that the path is safe, accessible and efficient, and a target path map is generated. The target path map is a graph structure generated by path analysis, which describes the path information from the current starting point to the target point. Among them, the path analysis is achieved by traversing the environmental data by combining the environmental perception data and using the A* algorithm or the Dijkstra algorithm. On the basis of path generation, the behavior analysis module 13 performs obstacle detection in combination with the target direction information and the target path map. The obstacle detection uses the characteristic information about the obstacle in the environmental perception dimension reduction data array, such as position, size, and dynamics, to analyze the objects or areas on the path that may hinder the robot's movement, and mark their positions and characteristics to generate behavioral obstacle information. The obstacle information is used to describe the obstacle information on the path, including the type and position of the obstacle and the degree of influence on the path. Then, according to the target direction information, the target path map and the behavior obstacle information, the possible actions of the perception robot are classified and mapped into specific behavior patterns, which are the types of actions performed by the perception robot in the current environment, such as obstacle avoidance behavior, detour behavior or target approach behavior. Finally, multiple behavior patterns are integrated into behavior information, which is a specific description of the behavior pattern, including the action type, specific execution conditions such as trigger signals or path nodes, and the priority of each behavior, which is used to guide the actions of the perception robot. The behavior information is output in a structured form to provide a clear execution basis for the behavior planning module. For example, the behavior analysis module 13 generates a set of behavior information: avoid obstacles first, then drive in a straight line, and finally reach the target point. The behavior analysis module 13 can dynamically adapt to environmental changes and generate optimal behavior information by combining the target task with the environmental perception data. By comprehensively considering the target direction, path planning and obstacle detection results, the accuracy of behavior classification is ensured, and complete behavior analysis behavior information can be generated to enhance the flexibility of the perception robot.

[0028] Furthermore, classification is performed according to the target direction information, the target path map, and the behavioral disorder information to determine multiple behavioral patterns. The behavior analysis module 13 is also used to: extract features of the perception robot based on the target direction information, the target path map, and the behavioral disorder information to obtain multiple behavioral features; use a support vector machine to classify the multiple behavioral features to determine multiple behavior types; combine the multiple behavior types with the environmental perception dimensionality reduction data array to learn and obtain behavioral learning results; perform pattern recognition based on the behavioral learning results to determine the multiple behavior patterns.

[0029] Specifically, based on the target direction information, the target path map and the behavior obstacle information, the current environment and task state of the perception robot are feature extracted to obtain multiple behavior features, wherein the behavior features include direction features, path features and obstacle characteristic information, wherein the direction features include angle deviation of the target direction, distance of the target position, etc., the path features include path complexity, number of turning points, path length, etc., and the obstacle characteristics include obstacle position, size, dynamics, etc., wherein dynamics refers to whether the obstacle moves. Then, the extracted multiple behavior features are input into a machine learning model for classification, wherein the machine learning model is a support vector machine, and the support vector machine is used as a classifier. The support vector machine divides different behavior types by constructing a hyperplane in a high-dimensional feature space. For example, behavior types such as straight-line driving, turning obstacle avoidance, and pause waiting are classified into different labels. The support vector machine classification process includes: mapping the behavior features to a high-dimensional space, finding a hyperplane that maximizes the class interval, ensuring classification accuracy, and outputting the behavior type to which each input feature vector belongs. Next, the multiple behavior types are combined with the environmental perception dimension reduction data matrix to conduct behavior learning. Through a data-driven method, the behavior decision of the perception robot under the current environmental conditions is obtained. After the behavior learning is completed, the behavior learning results are obtained to guide the recognition of behavior patterns. Optionally, according to the classification results, the behavior priority of the robot in different situations is adjusted, and the classification standards and applicable scope of the behavior types are optimized through continuous learning and historical data feedback. Finally, based on the behavior learning results, the correlation between the behavior types is identified, and the logical relationship of the behavior pattern is constructed according to the correlation, for example, which behavior types can be combined to form a complete pattern, and then multiple behavior patterns are output for subsequent behavior planning. The behavior pattern refers to a high-level behavior set composed of multiple behavior types, which represents the action logic of the robot under specific conditions. By using feature extraction, support vector machine classification, behavior learning and pattern recognition, the environment and task data are converted into multiple specific behavior patterns, so as to dynamically guide the robot to adopt adaptive behavior strategies in different situations and achieve efficient and accurate execution of tasks.

[0030] The behavior tree construction module 14 is used to combine the multiple behavior information to construct a behavior tree.

[0031] Specifically, the behavior tree construction module 14 receives multiple behavior information, including multiple behavior types, behavior triggering conditions, and execution priorities. Then, the multiple behavior information is logically analyzed to determine the correlation between behaviors, and the relevant behavior information is combined according to task requirements and environmental characteristics. For example, obstacle avoidance behavior may be composed of two specific behavior information, namely, turning obstacle avoidance and detouring obstacle avoidance, to construct a behavior tree. The behavior tree is a hierarchical structure that can flexibly describe multiple behavior paths of the perception robot under different environments and task conditions. By generating a behavior tree, support is provided for the behavior planning of the perception robot, so that the perception robot can complete the task objectives in real time and efficiently in a complex environment.

[0032] Further, the multiple behavior information are combined to construct a behavior tree, and the behavior tree construction module 14 is also used to: perform behavior association analysis based on the multiple behavior information to obtain multiple association coefficients; construct multiple association edges in combination with the multiple behavior information according to the multiple association coefficients; map the multiple behavior information into multiple behavior tree nodes, connect the multiple behavior tree nodes according to the multiple association edges, and construct the behavior tree.

[0033] Specifically, firstly, a behavior association analysis is performed on multiple behavior information. By evaluating the dependencies and logical relationships between behaviors, multiple association coefficients are generated. The association coefficient is a numerical value used to quantify the strength of the association between two behaviors, ranging from no association 0 to complete association 1. It is obtained based on time association, spatial association and conditional association. The time association refers to whether one behavior needs to be executed after another behavior is completed, the spatial association refers to whether the behaviors involve adjacent or the same spatial area, and the conditional association refers to whether the behaviors share the same triggering conditions, such as target direction or obstacle information. Then, according to the behavior association analysis results, multiple association edges are constructed based on multiple association coefficients to connect related behavior information. The association edge is the basic structure of the behavior tree, which is used to represent the logical or conditional relationship between behavior nodes. When the association coefficient is higher than the set threshold, an association edge is added between the two behavior information. The weight of the edge is determined by the association coefficient, which represents the strength of the association between the behaviors. For example, if the condition node of the obstacle detection has a high association with the turning obstacle avoidance behavior node, an association edge is added between the two. Finally, multiple behavior information is mapped to behavior tree nodes, that is, the behavior information is converted into nodes in the behavior tree and assigned corresponding attributes. The behavior tree nodes include behavior nodes and condition nodes. The behavior nodes represent specific behavior information, such as driving in a straight line and avoiding obstacles, which are the actions that the perception robot finally performs. The condition nodes represent the specific conditions that trigger the behavior, such as detecting obstacles, which are used to determine whether to perform the corresponding behavior. Then, according to the multiple associated edges constructed, the behavior tree nodes are logically connected to form a complete behavior tree structure. The behavior tree is a hierarchical structure, with the root node representing the main task goal, the branch nodes representing different behavior paths, and the leaf nodes being specific behavior execution actions. By constructing a behavior tree, a complete logical framework is provided for the robot's adaptive behavior planning, which can guide the robot to efficiently and accurately complete task goals in a complex dynamic environment.

[0034] Further, the multiple behavior information is mapped into multiple behavior tree nodes, and the multiple behavior tree nodes are connected according to the multiple associated edges to construct the behavior tree. The behavior tree construction module 14 is also used to: analyze the multiple behavior information according to the multiple behavior types according to the environment perception dimensionality reduction data array, and set multiple behavior triggering conditions; allocate the multiple behavior information according to the multiple behavior triggering conditions, and set execution priorities; map the multiple behavior information into multiple behavior nodes according to the execution priorities, and map the multiple behavior triggering conditions into multiple condition nodes according to the execution priorities; match the multiple behavior nodes with the multiple condition nodes to determine the multiple behavior tree nodes.

[0035] Specifically, multiple behavior information is analyzed according to the environmental perception dimension reduction data array, and multiple behavior trigger conditions are set according to different behavior types. The behavior trigger condition refers to the environment or state condition that a certain behavior needs to meet. It is the activation basis of the behavior node and defines under what circumstances a certain behavior of the perception robot can be activated. For example, the target direction deviation less than a certain value can be used as a trigger condition for straight-line behavior, and the detection of obstacles can be used as a trigger condition for turning and avoiding obstacles. By analyzing the logical attributes of the behavior information and the characteristics of the environmental data, the accuracy and comprehensiveness of the trigger conditions are ensured. After the trigger conditions are set, the multiple behavior information is allocated according to the corresponding trigger conditions. Based on the importance of the behavior information and the task objectives, the execution priority is set for each behavior information. The behavior with a high priority has a higher execution order in the behavior tree. Optionally, the priority setting can be allocated based on the urgency and importance of the behavior information. The execution priority indicates the order in which the behavior information is executed in the behavior tree, which is determined by the urgency and importance of the task. According to the allocation result, the multiple behavior information is mapped to multiple behavior nodes according to the execution priority, and the corresponding behavior trigger conditions are mapped to condition nodes. The mapping of behavior nodes and condition nodes is based on the trigger conditions and execution logic described in the behavior information. For example, the conditional node corresponding to the obstacle avoidance behavior node may be that the obstacle distance is less than 0.5 meters. Finally, according to the priority and logical relationship, the behavior node is matched with the conditional node, multiple behavior tree nodes are determined, and these nodes are connected through associated edges in the execution order to build a complete behavior tree. The connection rules include: the conditional node is connected to the corresponding behavior node through an associated edge, indicating the behavior executed after the trigger condition is met, and the behavior nodes with higher priority are arranged at a higher position in the tree to ensure priority execution. Different branches represent different behavior paths for dealing with different environments or task situations. Through the analysis of behavior information, priority setting, node mapping and logical connection, a hierarchical behavior tree structure is generated to improve the flexibility of the adaptive behavior planning of the perception robot, ensuring that the perception robot can efficiently and flexibly complete adaptive tasks in a variety of scenarios.

[0036] The behavior planning module 15 is used to perform strategy training through the behavior tree in combination with the environmental perception dimensionality reduction data array, generate a behavior strategy, and execute the behavior strategy to perform adaptive behavior planning for the perception robot.

[0037] Specifically, the behavior planning module 15 is the core module for the perception robot to implement adaptive behavior planning. It evaluates the applicability and execution possibility of each behavior node in the current environment by traversing the behavior tree. During the traversal process, the strategy training is combined with the environmental perception dimensionality reduction data array to continuously adjust and optimize the nodes and paths in the behavior tree to generate a behavior strategy that adapts to the current environment. Once the behavior strategy is generated, the behavior strategy is executed to guide the perception robot to perform adaptive behavior planning. The behavior planning module 15 realizes efficient and intelligent behavior planning for the perception robot by combining the behavior tree with the environmental perception dimensionality reduction data array, so that the perception robot can make reasonable decisions and take corresponding actions in a complex and changing environment, and efficiently complete the task objectives.

[0038] Furthermore, the behavior tree is combined with the environment perception dimensionality reduction data array to perform strategy training and generate a behavior strategy. The behavior planning module 15 is also used to: map the target task to the behavior tree for node traversal to determine multiple behavior nodes; introduce a reward function to determine whether the multiple behavior nodes meet the expected behavior node interval according to the environment perception dimensionality reduction data array; if the multiple behavior nodes all meet the expected behavior node interval, calculate according to the reward function to generate positive reward data; if any one of the multiple behavior nodes does not meet the expected behavior node interval, calculate according to the reward function to generate negative reward data; feedback update the behavior tree based on the positive reward data or the negative reward data to formulate the behavior strategy.

[0039] Specifically, the current target task of the perception robot is mapped to the behavior tree, converted into a specific node in the behavior tree, and the execution path is clarified. Each node of the behavior tree is traversed, starting from the root node of the behavior tree, and gradually traversed downward according to the current environmental state and the requirements of the target task to determine multiple behavior nodes. During the traversal process, the perception robot performs corresponding operations or makes conditional judgments according to the type of behavior node. In order to optimize the behavior strategy, a reward function is introduced in the traversal process of the behavior tree node to quantitatively evaluate the state and execution result of each node. According to the input of the environmental perception dimensionality reduction data array, it is judged whether each node is within the expected behavior node interval. The expected behavior node interval is the execution range of the behavior node under ideal conditions. The reward function is used to evaluate the execution result of the behavior node. The reward guides the strategy optimization through rewards. The rewards include positive rewards and negative rewards. When the execution result of a behavior node meets the task goal or its state meets the expected conditions, a positive reward is generated. For example, when the obstacle avoidance behavior successfully bypasses the obstacle, a high positive reward is given. When the execution result of a behavior node deviates from the task goal, or its state does not meet the expected conditions, a negative reward is generated. For example, when the straight-line driving behavior causes deviation from the target direction, a negative reward is given.

[0040] According to the states of multiple behavior nodes and reward functions, reward data is calculated and classified as positive reward or negative reward. When all traversed behavior nodes meet the expected behavior node interval, positive reward data is calculated and generated. For example, the robot successfully completes all obstacle avoidance behaviors and reaches the target point, generating positive reward data. If any behavior node fails to meet the expected behavior node interval, it is calculated according to the reward function to generate negative reward data. For example, if the robot fails to complete path planning within the specified time, negative reward data will be generated. Positive reward data indicates that the execution result of the behavior node meets expectations and is used to strengthen the weight of the relevant node in the behavior tree, while negative reward data indicates that the execution result of the behavior node does not meet expectations and is used to adjust the strategy or weight. Feedback update the behavior tree based on the reward data to optimize the behavior strategy. Optionally, for the behavior node that generates positive rewards, increase its execution priority or adjust its triggering conditions to make it easier to be selected. For the behavior node that generates negative rewards, reduce its execution priority or adjust the associated conditions to avoid repeated errors. Optimize the behavior path planning by adjusting the weight of the associated edge, for example, set the path with high positive rewards as the main execution path. Finally, a specific behavior strategy is generated based on the optimized behavior tree. The generated behavior strategy is output and applied to the robot's real-time behavior planning, ensuring that the perception robot can complete the target task flexibly and efficiently in the current environment.

[0041] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0042] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations 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 equivalents, the present application intends to include these modifications and variations.

Claims

1. An adaptive behavior planning system for a perceptual robot, characterized in that The system comprises: An environmental perception module is used to extract multiple sensors based on the perception robot, perform environmental perception through the multiple sensors, and obtain an environmental perception data set; A data dimension reduction module, used for preprocessing the environment perception data set, reducing the dimension of the environment perception data set according to the preprocessing result, and generating an environment perception reduced dimension data array; A behavior analysis module, used to perform behavior analysis on the perception robot according to the environmental perception dimension reduction data array to generate a plurality of behavior information; A behavior tree construction module, used for combining the plurality of behavior information to construct a behavior tree; The behavior planning module is used to perform strategy training through the behavior tree in combination with the environmental perception dimensionality reduction data array, generate a behavior strategy, and execute the behavior strategy to perform adaptive behavior planning for the perception robot.

2. The adaptive behavior planning system for a perceptual robot according to claim 1, characterized in that: The data dimension reduction module is also used for: Performing multi-dimensional analysis based on the environmental perception data set to determine a multi-dimensional environmental state signal; Performing wavelet decomposition on the multidimensional environmental state signal to obtain wavelet coefficients of the multidimensional environmental state signal; Perform threshold quantization according to the wavelet coefficients of the multidimensional environmental state signal to determine the wavelet selection threshold of the multidimensional environmental state signal; According to the multidimensional environmental state signal wavelet selection threshold, the multidimensional environmental state signal wavelet coefficients are intercepted, the noise signal less than the multidimensional environmental state signal wavelet selection threshold is set to zero, and the valid signal information greater than the multidimensional environmental state signal wavelet selection threshold is obtained; The effective signal information is filtered and reconstructed to obtain the preprocessing result.

3. The adaptive behavior planning system for a perceptual robot according to claim 2, characterized in that: The data dimension reduction module is also used for: Perform distribution calculation on the environment perception data set based on the wavelet selection threshold of the multi-dimensional environment state signal to obtain a data distribution probability array; Constructing a similarity probability distribution function, performing dimensionality reduction processing on the dot distribution of the data distribution probability array, and generating a data dimensionality reduction result; The environment-aware dimension reduction data array is constructed based on the data dimension reduction result.

4. The adaptive behavior planning system for a perceptual robot according to claim 1, characterized in that: The behavior analysis module is also used to: Retrieving a target task, performing a steering analysis according to the target task based on the environment perception dimensionality reduction data array, and determining target direction information; Performing path analysis according to the target direction information based on the environment perception dimensionality reduction data array to obtain a target path map; Obstacle detection is performed according to the target direction information, the target path map and the environment perception dimensionality reduction data array to obtain behavior obstacle information; Classify according to the target direction information, the target path map, and the behavior obstacle information to determine a plurality of behavior patterns; The plurality of behavior patterns are added to the plurality of behavior information.

5. The adaptive behavior planning system for a perceptual robot according to claim 4, characterized in that: The behavior analysis module is also used to: Extracting features of the perception robot based on the target direction information, the target path map, and the behavior obstacle information to obtain a plurality of behavior features; Using a support vector machine to classify the multiple behavior features to determine multiple behavior types; Learning the multiple behavior types in combination with the environment perception dimensionality reduction data array to obtain a behavior learning result; Pattern recognition is performed based on the behavior learning results to determine the multiple behavior patterns.

6. The adaptive behavior planning system for a perceptual robot according to claim 5, characterized in that: The Behavior Tree Building Module is also used to: Performing behavior correlation analysis based on the multiple behavior information to obtain multiple correlation coefficients; Constructing a plurality of association edges according to the plurality of association coefficients and the plurality of behavior information; The plurality of behavior information are mapped into a plurality of behavior tree nodes, and the plurality of behavior tree nodes are connected according to the plurality of associated edges to construct the behavior tree.

7. The adaptive behavior planning system for a perceptual robot according to claim 6, characterized in that: The Behavior Tree Building Module is also used to: Analyze the plurality of behavior information according to the plurality of behavior types according to the environment perception dimension reduction data array, and set a plurality of behavior triggering conditions; Allocating the plurality of behavior information according to the plurality of behavior triggering conditions and setting execution priorities; Mapping the plurality of behavior information into a plurality of behavior nodes according to the execution priority, and mapping the plurality of behavior triggering conditions into a plurality of condition nodes according to the execution priority; The multiple behavior nodes are matched with the multiple condition nodes to determine the multiple behavior tree nodes.

8. The adaptive behavior planning system for a perceptual robot according to claim 4, characterized in that: The behavior planning module is also used to: Mapping the target task to the behavior tree to perform node traversal and determine a plurality of behavior nodes; Introducing a reward function, and judging whether the plurality of behavior nodes conform to an expected behavior node interval according to the environment perception dimensionality reduction data array; If the plurality of behavior nodes all conform to the expected behavior node interval, calculation is performed according to the reward function to generate positive reward data; If any one of the plurality of behavior nodes does not conform to the expected behavior node interval, performing calculation according to the reward function to generate negative reward data; Feedback is performed on the behavior tree based on the positive reward data or the negative reward data to formulate the behavior strategy.

Images