Rail robot path obstacle avoidance data updating method and system

Abstract

The present application relates to the technical field of track obstacle avoidance, and discloses a track robot path obstacle avoidance data updating method and system, the method comprising: assembling a plurality of dimensional features in real-time environment data flow of the track robot into a vector group to obtain a primary feature vector, and performing abnormal mode identification on the primary feature vector and a historical normal mode library to obtain dynamic abnormal features; reconstructing a local topology of a pre-stored path knowledge graph according to the dynamic abnormal features to generate an obstacle influence domain; analyzing a field strength distribution trend of the obstacle influence domain to obtain a gradient field, and mapping a dominant direction of the gradient field into a conflict resolution strategy; correcting a predetermined path of the track robot to obtain a conflict-free path trajectory segment; performing efficiency determination on the conflict-free path trajectory segment to obtain a multi-dimensional evaluation result; and fusing the multi-dimensional evaluation result to obtain an emergent motion strategy; the present application can improve the efficiency of track robot path obstacle avoidance data updating.

Description

Technical Field

[0001] This invention relates to the field of obstacle avoidance technology, and in particular to a method and system for updating obstacle avoidance data for a track robot. Background Technology

[0002] In the field of obstacle avoidance technology for orbital robots, real-time environmental perception and abnormal feature identification are core prerequisites for ensuring the effectiveness of obstacle avoidance. However, existing technologies have significant shortcomings in this aspect. Existing technologies do not comprehensively extract multi-dimensional features from the real-time environmental data stream of orbital robots, often focusing only on a single type of feature and failing to perform effective dimensional consistency transformation and tensor synthesis for multi-dimensional features. This results in the generated feature vectors failing to accurately reflect the true state of the environment. At the same time, their abnormal pattern identification methods with historical normal pattern libraries are relatively simple, judging abnormalities only by calculating the difference between basic vectors, without taking into account the distribution characteristics and stability of feature vectors. This causes lag or misjudgment rate in dynamic abnormal feature identification, making it difficult to quickly capture abnormal changes caused by sudden obstacles in the environment, directly affecting the timeliness and accuracy of subsequent obstacle avoidance decisions.

[0003] Existing technologies also suffer from poor performance in obstacle influence domain characterization, path correction, and motion strategy generation. On the one hand, existing technologies struggle to accurately reconstruct the local topology of pre-stored path knowledge graphs based on dynamic anomaly characteristics. The generation of obstacle influence domains often relies on fixed rules or coarse spatial divisions, failing to quantify node interaction relationships and field strength distribution trends, resulting in a mismatch between the influence domain range and the actual degree of obstacle impact. On the other hand, during path correction, conflict resolution strategies are not deeply correlated with the dominant gradient field direction, leading to insufficient strategy adaptability. Furthermore, there is a lack of multi-dimensional performance evaluation for the corrected conflict-free path trajectory segments, making it impossible to generate emergent motion strategies adapted to complex environments through multi-dimensional result fusion. Ultimately, this results in obstacle avoidance paths for tracked robots exhibiting conflict risks and low operational efficiency, failing to meet the reliable operational requirements of dynamic and complex track environments. Therefore, improving the efficiency of tracked robot path obstacle avoidance data updates has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a method and system for updating path obstacle avoidance data for a track robot, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for updating path obstacle avoidance data for an orbital robot, comprising:

[0006] S1. Perform vector assembly on the multi-dimensional features in the real-time environmental data stream of the track robot to obtain the primary feature vector of the multi-dimensional features, and perform abnormal pattern identification on the primary feature vector and the historical normal pattern library to obtain the dynamic abnormal features of the real-time environmental data stream.

[0007] S2. Based on the dynamic anomaly characteristics, reconstruct the local topology of the pre-stored path knowledge graph to generate the obstacle influence domain of the orbital robot;

[0008] S3. Analyze the field strength distribution change trend of the obstacle's influence domain, obtain the gradient field of the field strength distribution change trend, and map the dominant direction of the gradient field to the conflict resolution strategy of the orbital robot.

[0009] S4. Based on the conflict resolution strategy, correct the predetermined path of the orbital robot to obtain a conflict-free path trajectory segment of the orbital robot;

[0010] S5. Based on the decision preferences in the conflict resolution strategy, the effectiveness of the conflict-free path trajectory segment is determined to obtain the multi-dimensional evaluation results of the conflict-free path trajectory segment.

[0011] S6. By integrating the multi-dimensional evaluation results, the emergent motion strategy of the orbital robot is obtained.

[0012] In a preferred embodiment, the step of assembling vectors from the multi-dimensional features in the real-time environmental data stream of the orbital robot to obtain a primary feature vector of the multi-dimensional features, and then performing anomaly pattern identification between the primary feature vector and a historical normal pattern library to obtain the dynamic anomaly features of the real-time environmental data stream, includes:

[0013] Spatial position features, motion state features, and environmental perception features are extracted from the real-time environmental data stream of the orbital robot to obtain the multi-dimensional original features of the real-time environmental data stream;

[0014] The multidimensional original features are subjected to dimension consistency transformation to obtain dimensionless features of the multidimensional original features;

[0015] Tensor synthesis is performed on the dimensionless features to obtain the primary feature vector of the dimensionless features;

[0016] Calculate the difference between the primary feature vector and the pattern vectors in the historical normal pattern library to obtain the difference sequence of the primary feature vector;

[0017] By analyzing the abnormal pattern distribution of the difference sequence, the dynamic abnormal characteristics of the real-time environmental data stream are obtained.

[0018] In a preferred embodiment, the formula for calculating the degree of difference is as follows:

[0019] ;

[0020] In the formula, The first in the historical normal pattern library A pattern vector, The primary feature vector and the first The degree of difference between pattern vectors For the primary feature vector, This is a preset numerical stability factor. The balance coefficients were obtained to analyze the distribution characteristics of pattern vectors in the historical normal pattern library. The standard deviation of the eigenvalues ​​in the primary eigenvector is given. For the first The standard deviation of the eigenvalues ​​in each pattern vector The preset distribution smoothing factor, To obtain the maximum value, Let L2 norm be the vector. To take the absolute value.

[0021] In a preferred embodiment, the step of reconstructing the local topology of the pre-stored path knowledge graph based on the dynamic anomaly characteristics to generate the obstacle influence domain of the orbital robot includes:

[0022] The dynamic anomaly features are mapped to a pre-stored path knowledge graph to construct the anomaly-node mapping relationship between the dynamic anomaly features and the path knowledge graph.

[0023] Based on the anomaly-node mapping relationship, the adjacency network of the affected nodes in the path knowledge graph is constructed to obtain the local topology of the affected nodes;

[0024] Optimize the connection strength of the edges in the local topology to obtain the reconstructed topology network of the local topology;

[0025] The interaction relationships between nodes in the reconstructed topology network are quantified to obtain the node influence magnitude distribution of the reconstructed topology network;

[0026] Spatial clustering of the influence distribution of the nodes is performed to obtain the obstacle influence domain of the orbital robot.

[0027] In a preferred embodiment, the step of analyzing the field strength distribution variation trend of the obstacle's influence domain to obtain the gradient field of the field strength distribution variation trend, and mapping the dominant direction of the gradient field to the conflict resolution strategy of the orbital robot, includes:

[0028] By monitoring the spatiotemporal evolution pattern of field strength distribution within the obstacle's influence domain, key change areas within the obstacle's influence domain can be obtained.

[0029] Analyze the spatial variation law of the field strength value in the key changing region, and construct the gradient direction set of the key changing region;

[0030] The intensity contrast between the gradient direction sets is evaluated to obtain the dominant change direction of the obstacle's influence domain;

[0031] The dominant change direction is matched with the primitive strategies in the pre-stored strategy library to obtain the matching relationship between the dominant change direction and the strategy in the pre-stored strategy library.

[0032] The strategy matching relationship is integrated into the conflict resolution strategy of the orbital robot.

[0033] In a preferred embodiment, analyzing the spatial variation law of the field strength value in the key change region and constructing the gradient direction set of the key change region includes:

[0034] By tracing the spatial variation characteristics of the field strength values ​​in the key variation region, the trajectory of the field strength variation in the key variation region can be obtained;

[0035] Establish the direction vector of the monitoring point in the field strength change trajectory, assign correlation weights according to the angle relationship between the direction vectors, and connect the monitoring points through the correlation weights to generate a spatial correlation map of the key change area;

[0036] By dividing the directional clustering intervals in the spatial correlation diagram, the dominant directional intervals of the key change regions are obtained;

[0037] Based on the statistical characteristics of the dominant direction interval, a gradient direction set for the key change region is generated.

[0038] In a preferred embodiment, the step of correcting the predetermined path of the orbital robot according to the conflict resolution strategy to obtain a conflict-free path trajectory segment of the orbital robot includes:

[0039] By deconstructing the avoidance decision elements in the conflict resolution strategy, the set of strategy elements of the conflict resolution strategy is obtained;

[0040] The strategy element set is input into the predetermined path of the orbital robot to obtain the dynamic path segment of the orbital robot;

[0041] The connection relationships of the dynamic path segments are reconstructed to obtain the topology-optimized path of the orbital robot;

[0042] The topology-optimized path is smoothed to obtain the continuous path trajectory of the orbital robot;

[0043] Verify the spatial relationship between the continuous path trajectory and the influence domain of the obstacle to obtain a conflict-free path trajectory segment of the orbital robot.

[0044] In a preferred embodiment, the step of determining the effectiveness of the conflict-free path trajectory segment based on the decision preferences in the conflict resolution strategy to obtain a multi-dimensional evaluation result of the conflict-free path trajectory segment includes:

[0045] The key parameters of decision preferences in the conflict resolution strategy are used as evaluation benchmarks;

[0046] The motion state parameters of the path points in the conflict-free path trajectory segment are collected, and the position coordinates and timestamps of the path points are recorded to construct the motion feature data of the conflict-free path trajectory segment.

[0047] The correlation analysis between the evaluation benchmark and the feature dimensions in the motion feature data is performed to obtain the coupled evaluation relationship between the evaluation benchmark and the motion feature data.

[0048] Based on the aforementioned coupling evaluation relationship, the effectiveness score of the path point is calculated, resulting in a score sequence for the path point. The calculation formula for the effectiveness score is as follows:

[0049] ;

[0050] In the formula, path point The aforementioned performance score, To evaluate the total number of benchmark parameters, For the first The weighting coefficients of each evaluation benchmark parameter, For the first The correlation coefficients between each evaluation benchmark parameter and its corresponding motion feature dimension For the first Normalized values ​​of each evaluation benchmark parameter For the first Normalized values ​​of each motion feature dimension. The preset stability factor, It is a natural exponential function;

[0051] The scoring sequence is reorganized to obtain a multi-dimensional evaluation result of the conflict-free path trajectory segment.

[0052] In a preferred embodiment, the process of fusing the multi-dimensional evaluation results to obtain the emergent motion strategy of the orbital robot includes:

[0053] By integrating the security, efficiency, and stability dimensions from the multi-dimensional evaluation results, a combination of strategy elements from the multi-dimensional evaluation results is obtained.

[0054] Based on the topological structure of the path knowledge graph, spatial constraints are applied to the combination of strategy elements to obtain the restricted strategy set of the orbital robot;

[0055] Based on the decision preferences in the conflict resolution strategy, the set of restricted strategies is prioritized to obtain the strategy behavior sequence of the orbital robot;

[0056] The emergent motion strategy of the orbital robot is obtained by spatiotemporally aligning the strategy behavior sequence with the conflict-free path trajectory segment.

[0057] To address the above problems, the present invention also provides a path obstacle avoidance data update system for an orbital robot, the system comprising:

[0058] The feature vectorization and anomaly detection module is used to assemble multi-dimensional features in the real-time environmental data stream of the track robot into vectors to obtain the primary feature vectors of the multi-dimensional features, and to identify the anomaly patterns by comparing the primary feature vectors with the historical normal pattern library to obtain the dynamic anomaly features of the real-time environmental data stream.

[0059] The topology reconstruction and influence domain generation module is used to reconstruct the local topology of the pre-stored path knowledge graph based on the dynamic anomaly characteristics, so as to generate the obstacle influence domain of the orbital robot.

[0060] The gradient analysis and strategy mapping module is used to analyze the field strength distribution change trend of the obstacle influence domain, obtain the gradient field of the field strength distribution change trend, and map the dominant direction of the gradient field to the conflict resolution strategy of the orbital robot.

[0061] The path correction and trajectory generation module is used to correct the predetermined path of the orbital robot according to the conflict resolution strategy, so as to obtain a conflict-free path trajectory segment of the orbital robot.

[0062] The performance evaluation and decision optimization module is used to determine the performance of the conflict-free path trajectory segment based on the decision preferences in the conflict resolution strategy, and obtain a multi-dimensional evaluation result of the conflict-free path trajectory segment.

[0063] The strategy fusion and emergent control module is used to fuse the multi-dimensional evaluation results to obtain the emergent motion strategy of the orbital robot.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] 1. This invention assembles multi-dimensional features of the real-time environmental data stream of a tracked robot into vectors, and combines this with a historical normal pattern library to accurately identify abnormal patterns, effectively acquiring dynamic abnormal features. Then, based on these dynamic abnormal features, it reconstructs the local topology of a pre-stored path knowledge graph, generating an obstacle influence domain, and analyzes its field strength distribution trend to obtain the gradient field, mapping the dominant direction to a conflict resolution strategy. This process improves the accuracy and timeliness of dynamic abnormal feature identification, achieves precise characterization of the obstacle influence domain, provides a reliable basis for subsequent path optimization, and ensures the scientific nature of obstacle avoidance decisions.

[0066] 2. This invention corrects the predetermined path of a tracked robot through a conflict resolution strategy to obtain conflict-free path trajectory segments. Then, it combines the decision preferences in the strategy to perform multi-dimensional performance evaluation of the trajectory segments, and merges the evaluation results to generate an emergent motion strategy. This ensures the conflict-free nature of the path trajectory, improves the adaptability of the motion strategy to dynamic environments, significantly improves the efficiency of path obstacle avoidance data updates for the tracked robot, and enhances the robot's operational stability and reliability, meeting the operational requirements in complex track environments. Attached Figure Description

[0067] Figure 1 This is a flowchart illustrating a method for updating path obstacle avoidance data for an orbital robot, provided in an embodiment of the present invention.

[0068] Figure 2 This is a functional block diagram of a path obstacle avoidance data update system for an orbital robot provided in an embodiment of the present invention;

[0069] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0070] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0071] This application provides a method for updating path obstacle avoidance data for a tracked robot. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for updating path obstacle avoidance data for a tracked robot can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0072] Reference Figure 1 The diagram shown is a flowchart illustrating a method for updating path obstacle avoidance data for a tracked robot according to an embodiment of the present invention. In this embodiment, the method includes:

[0073] S1. Perform vector assembly on the multi-dimensional features in the real-time environmental data stream of the track robot to obtain the primary feature vector of the multi-dimensional features, and perform abnormal pattern identification on the primary feature vector and the historical normal pattern library to obtain the dynamic abnormal features of the real-time environmental data stream.

[0074] In this embodiment of the invention, the step of assembling vectors from the multi-dimensional features in the real-time environmental data stream of the orbital robot to obtain a primary feature vector of the multi-dimensional features, and then performing abnormal pattern identification on the primary feature vector with a historical normal pattern library to obtain the dynamic abnormal features of the real-time environmental data stream, includes:

[0075] Spatial position features, motion state features, and environmental perception features are extracted from the real-time environmental data stream of the orbital robot to obtain the multi-dimensional original features of the real-time environmental data stream;

[0076] The multidimensional original features are subjected to dimension consistency transformation to obtain dimensionless features of the multidimensional original features;

[0077] Tensor synthesis is performed on the dimensionless features to obtain the primary feature vector of the dimensionless features;

[0078] Calculate the difference between the primary feature vector and the pattern vectors in the historical normal pattern library to obtain the difference sequence of the primary feature vector;

[0079] By analyzing the abnormal pattern distribution of the difference sequence, the dynamic abnormal characteristics of the real-time environmental data stream are obtained.

[0080] The formula for calculating the degree of difference is as follows:

[0081] ;

[0082] In the formula, The first in the historical normal pattern library A pattern vector, The primary feature vector and the first The degree of difference between pattern vectors The primary feature vector, This is a preset numerical stability factor. The balance coefficients were obtained to analyze the distribution characteristics of pattern vectors in the historical normal pattern library. The standard deviation of the eigenvalues ​​in the primary eigenvector is given. For the first The standard deviation of the eigenvalues ​​in each pattern vector The preset distribution smoothing factor, To obtain the maximum value, Let L2 norm be the vector. To take the absolute value.

[0083] Spatial position features, motion state features, and environmental perception features are extracted from the real-time environmental data stream of the track robot to obtain multi-dimensional original features of the real-time environmental data stream. The GPS positioning module on the track robot obtains the robot's coordinate information in three-dimensional space, and the encoder obtains the robot's attitude angles. These constitute spatial position features. The linear velocity and angular velocity of the robot are obtained by the velocity sensor, and the linear acceleration and angular acceleration of the robot are obtained by the accelerometer. These constitute motion state features. The track and surrounding environment are scanned by the lidar to obtain the distance, direction, and size information of obstacles. The track images are captured by the camera and the track wear degree, track joint position, etc. are identified. These constitute environmental perception features. All the above features are sorted into preset categories to form multi-dimensional original features containing three types of information: spatial position, motion state, and environmental perception.

[0084] The multi-dimensional original features are subjected to dimension consistency transformation to obtain dimensionless features. For the coordinate information in the spatial location features, the maximum and minimum values ​​of the feature in historical data are calculated. Each coordinate value is subtracted from the minimum value and divided by the difference between the maximum and minimum values ​​to obtain the normalized coordinate features. For the velocity information in the motion state features, the same normalization method is used to transform it based on the maximum and minimum values ​​of historical velocity data. For the obstacle distance, track wear degree, etc. in the environmental perception features, the features with units are normalized, and the original values ​​of the dimensionless features are kept. Through the above operations, all multi-dimensional original features are transformed into dimensionless features with values ​​in the range [0,1].

[0085] Tensor synthesis is performed on the dimensionless features to obtain the primary feature vectors of the dimensionless features. The dimensionless features corresponding to the spatial position features are arranged in the order of "X-axis coordinate - Y-axis coordinate - Z-axis coordinate - roll angle - pitch angle - yaw angle" to form a spatial feature sub-vector. The dimensionless features corresponding to the motion state features are arranged in the order of "linear velocity - angular velocity - linear acceleration - angular acceleration" to form a motion feature sub-vector. The dimensionless features corresponding to the environmental perception features are arranged in the order of "obstacle distance - obstacle direction - obstacle size - track wear degree - track joint position" to form an environmental feature sub-vector. The spatial feature sub-vectors, motion feature sub-vectors, and environmental feature sub-vectors are concatenated in sequence to form a one-dimensional numerical sequence, which is the primary feature vector of the dimensionless features.

[0086] The first in the historical normal pattern library The pattern vectors are derived from a set of primary feature vectors formed by continuously collecting environmental data streams and extracting multi-dimensional features during normal operation of the orbital robot. These vectors are then transformed using dimensional consistency conversion and tensor synthesis. This set constitutes the historical normal pattern library, where the pattern vectors are arranged in order as follows: The element is the th element. A pattern vector.

[0087] The numerical stability factor is a fixed value that is pre-set manually based on the application scenario and data calculation requirements of the track robot. Its purpose is to prevent the denominator in the formula from approaching zero and to ensure the stability of the calculation process.

[0088] The balance coefficient is a fixed value obtained by statistically analyzing the distribution characteristics of all pattern vectors in the historical normal pattern library. Specifically, it involves statistically analyzing the distribution information such as the feature value range and dispersion of all pattern vectors, and then determining a suitable value based on this information to weigh the two parts of the calculation results in the balance formula.

[0089] The standard deviation of the eigenvalues ​​in the primary eigenvector is obtained by calculating the standard deviation of all eigenvalues ​​in the primary eigenvector. The calculation process is as follows: first, calculate the average of all eigenvalues ​​in the primary eigenvector; then, calculate the difference between each eigenvalue and the average; square all the differences and sum them; divide the sum by the total number of eigenvalues; finally, take the square root of the result, which is the standard deviation of the eigenvalues ​​in the primary eigenvector.

[0090] No. The calculation process for the standard deviation of the eigenvalues ​​in the pattern vector is the same as that for the standard deviation of the eigenvalues ​​in the primary eigenvector, that is, first calculate the standard deviation of the eigenvalues ​​in the pattern vector. The average of all eigenvalues ​​in the pattern vector is calculated, then the difference between each eigenvalue and this average is calculated, squared, and summed. The sum is divided by the total number of eigenvalues ​​in the pattern vector, and finally the square root of the result is taken to obtain the nth pattern vector. The standard deviation of the eigenvalues ​​in each pattern vector.

[0091] The distribution smoothing factor is a fixed value that is manually set in advance based on the standard deviation distribution of the eigenvalues ​​of all pattern vectors in the historical normal pattern library. It is used to avoid the denominator in the formula being too small and to prevent abnormal fluctuations in the calculation results.

[0092] The difference between the primary feature vector and the pattern vectors in the historical normal pattern library is calculated to obtain the difference sequence of the primary feature vector. The historical normal pattern library stores multiple primary feature vectors collected by the orbital robot under normal operating conditions, and each pattern vector has a collection timestamp. For the currently acquired primary feature vector, each pattern vector in the historical normal pattern library is extracted one by one. The difference is obtained by combining the results of the two parts of the calculation. The first part of the calculation is done by calculating the difference between the primary feature vector and the pattern vector in the historical normal pattern library. The L2 norm of the difference between the pattern vectors and the L2 norm of their sum are multiplied together and then divided by the first L2 norm. The sum of the squares of the L2 norms of the first pattern vectors and the numerical stability factor mainly reflects the difference between the two vectors at the magnitude level. The second part calculates the difference by first finding the standard deviation of the eigenvalues ​​in the primary eigenvectors and then... The absolute value of the standard deviation of the eigenvalues ​​in the nth pattern vector is then divided by the nth pattern vector. The standard deviation of the eigenvalues ​​in the pattern vectors and the maximum value of the distribution smoothing factor are multiplied by the balance coefficient. This part mainly reflects the difference in the dispersion of the eigenvalue distributions between the two vectors. The sum of the two calculation results is the value of the primary eigenvector and the secondary eigenvector. The difference degree of each pattern vector is calculated, which takes into account the differences in vector magnitude and the dispersion of feature value distribution to ensure the comprehensiveness and accuracy of the difference degree calculation. The smallest value among all single vector difference degrees is selected as the final difference degree of the current primary feature vector. The final difference degree of each time step is arranged in the order of the acquisition time of the primary feature vector to form a difference degree sequence.

[0093] Analyzing the abnormal pattern distribution of the difference degree sequence yields the dynamic abnormal characteristics of the real-time environmental data stream. A difference degree threshold is set, which is determined by statistically analyzing the difference degree corresponding to all pattern vectors in the historical normal pattern library and taking the 95th percentile as the standard. The difference degree sequence is iterated, and when a difference degree value exceeds the set threshold, that moment is marked as an abnormal moment. The distribution of abnormal moments on the time axis is statistically analyzed. If three or more abnormal moments occur consecutively, it is determined as an abnormal time period. The difference degree value of each abnormal moment is calculated from the threshold. The larger the difference value, the higher the degree of abnormality. The trend of the degree of abnormality over time is recorded. At the same time, the differences between the primary feature vector of the abnormal moment and each component of the most similar pattern vector in the historical normal pattern library are compared. The feature category corresponding to the component with the largest difference is identified, and the dominant abnormal feature is determined. The abnormal time period, the trend of the degree of abnormality, and the dominant abnormal feature are integrated to form the dynamic abnormal characteristics of the real-time environmental data stream.

[0094] The beneficial effects are as follows: by accurately extracting multi-dimensional original features and performing dimensional consistency transformation, the comparability and validity of feature data are ensured; by forming primary feature vectors through tensor synthesis, the orderly integration of multi-source features is achieved; and by taking into account the difference in vector amplitude and feature value distribution dispersion calculation, as well as anomaly pattern distribution analysis, dynamic anomaly features in the real-time environment can be quickly and accurately identified, providing reliable anomaly information support for the subsequent path avoidance decision-making of the track robot, effectively improving the robot's adaptability to complex environments and the timeliness of obstacle avoidance response.

[0095] S2. Based on the dynamic anomaly characteristics, reconstruct the local topology of the pre-stored path knowledge graph to generate the obstacle influence domain of the orbital robot;

[0096] In this embodiment of the invention, the step of reconstructing the local topology of the pre-stored path knowledge graph based on the dynamic anomaly characteristics to generate the obstacle influence domain of the orbital robot includes:

[0097] The dynamic anomaly features are mapped to a pre-stored path knowledge graph to construct the anomaly-node mapping relationship between the dynamic anomaly features and the path knowledge graph.

[0098] Based on the anomaly-node mapping relationship, the adjacency network of the affected nodes in the path knowledge graph is constructed to obtain the local topology of the affected nodes;

[0099] Optimize the connection strength of the edges in the local topology to obtain the reconstructed topology network of the local topology;

[0100] The interaction relationships between nodes in the reconstructed topology network are quantified to obtain the node influence magnitude distribution of the reconstructed topology network;

[0101] Spatial clustering of the influence distribution of the nodes is performed to obtain the obstacle influence domain of the orbital robot.

[0102] The path knowledge graph pre-stores various nodes related to the track and the connections between them. Nodes include track segment nodes, spatial location nodes, and environmental status nodes. Dynamic anomaly features include the anomaly time period, the trend of anomaly severity, and the dominant anomaly feature. The corresponding node type is determined based on the dominant anomaly feature. If the dominant anomaly feature is a spatial location feature, it matches a spatial location node in the path knowledge graph; if it is an environmental perception feature, it matches an environmental status node. The range of graph nodes corresponding to the robot within that time period is then determined by combining the anomaly time period with the specific information of each dynamic anomaly feature. This binds the specific information of each dynamic anomaly feature to its corresponding node, clarifying which anomaly feature is associated with which node(s), thus forming an anomaly-node mapping relationship.

[0103] Extract all associated nodes from the anomaly-node mapping relationship as affected nodes. Query the direct neighboring nodes of each affected node in the path knowledge graph, i.e., nodes that have a direct connection with the affected node. Collect all affected nodes and their direct neighboring nodes, retain all connections between these nodes in the original path knowledge graph, and remove nodes and connections outside this range to form a network structure that only contains affected nodes, their neighboring nodes, and their interconnections. This structure is the local topology of the affected nodes.

[0104] The initial connection strength of each edge in the local topology is set based on the original node association tightness in the path knowledge graph; the tighter the association, the greater the connection strength. Combining the anomaly severity trend in the dynamic anomaly features, regions with higher anomaly severity show a greater reduction in the connection strength of edges between affected nodes and their neighbors; regions with lower anomaly severity show a slight reduction or no change in edge connection strength. This method adjusts the connection strength of all edges in the local topology to accurately reflect the impact of anomalies on node associations. The adjusted local topology is the reconstructed topology network.

[0105] For each node in the reconstructed topology, the number of nodes directly connected to it is counted, i.e., the node degree. Simultaneously, the optimized connection strength of the edges between the node and each directly connected node is recorded. The average of the sum of the node's degree and the strength of all directly connected edges is taken as the node's base influence weight. Then, based on the node's position in the network, if the node is in the network center, a fixed proportion of weight is added to the base influence weight; if it is in the edge region, the base influence weight remains unchanged. Each node is assigned a unique influence weight value, and the distribution of node influence weights is formed according to the correspondence between nodes and influence weights.

[0106] Obtain the actual spatial coordinates of each node in the reconstructed topology network. These coordinates correspond one-to-one with the physical location of the track and are pre-stored. Divide all nodes into several spatial units according to their spatial coordinates. Each spatial unit contains a certain number of adjacent nodes. Calculate the average influence strength of all nodes in each spatial unit. Set an average influence strength threshold and filter out spatial units with an average value higher than the threshold. Merge adjacent qualified spatial units and remove isolated unqualified spatial units. The resulting continuous and complete spatial range is the obstacle influence domain of the track robot.

[0107] The beneficial effects are that by accurately mapping dynamic anomaly features to the path knowledge graph, the range of affected nodes is clearly defined. Then, through adjacency network construction, edge connection strength optimization, node interaction quantification, and spatial clustering, the influence domain of obstacles is accurately characterized. This ensures that the influence domain can truly reflect the range and degree of impact of abnormal obstacles on the operation of the track robot, providing accurate spatial basis for subsequent conflict resolution strategy formulation and path correction, and effectively improving the scientificity and reliability of obstacle avoidance decision-making for track robots.

[0108] S3. Analyze the field strength distribution change trend of the obstacle's influence domain, obtain the gradient field of the field strength distribution change trend, and map the dominant direction of the gradient field to the conflict resolution strategy of the orbital robot.

[0109] In this embodiment of the invention, the step of analyzing the field strength distribution variation trend of the obstacle's influence domain to obtain the gradient field of the field strength distribution variation trend, and mapping the dominant direction of the gradient field to the conflict resolution strategy of the orbital robot, includes:

[0110] By monitoring the spatiotemporal evolution pattern of field strength distribution within the obstacle's influence domain, key change areas within the obstacle's influence domain can be obtained.

[0111] Analyze the spatial variation law of the field strength value in the key changing region, and construct the gradient direction set of the key changing region;

[0112] The intensity contrast between the gradient direction sets is evaluated to obtain the dominant change direction of the obstacle's influence domain;

[0113] The dominant change direction is matched with the primitive strategies in the pre-stored strategy library to obtain the matching relationship between the dominant change direction and the strategy in the pre-stored strategy library.

[0114] The strategy matching relationship is integrated into the conflict resolution strategy of the orbital robot.

[0115] The analysis of the spatial variation law of the field strength value in the key changing region and the construction of the gradient direction set of the key changing region include:

[0116] By tracing the spatial variation characteristics of the field strength values ​​in the key variation region, the trajectory of the field strength variation in the key variation region can be obtained;

[0117] Establish the direction vector of the monitoring point in the field strength change trajectory, assign correlation weights according to the angle relationship between the direction vectors, and connect the monitoring points through the correlation weights to generate a spatial correlation map of the key change area;

[0118] By dividing the directional clustering intervals in the spatial correlation diagram, the dominant directional intervals of the key change regions are obtained;

[0119] Based on the statistical characteristics of the dominant direction interval, a gradient direction set for the key change region is generated.

[0120] A full-domain scan of the obstacle's influence area is performed at fixed time intervals. Field strength values ​​for all spatial points within the domain are collected at each scan time, and the field strength data for each spatial point at different time points is recorded, forming a time-series data of the field strength distribution. By comparing the field strength distributions of two adjacent time points, the change in field strength at each spatial point is calculated. The average value of the change in field strength within the domain is statistically analyzed. Spatial points with changes greater than twice the average value are marked as active points, and the area covered by continuously distributed active points is designated as a critical change region.

[0121] Several fixed monitoring points are set at uniform intervals within the key change area, each corresponding to a unique spatial coordinate. The electric field strength (EVS) value of each monitoring point is continuously collected in chronological order, recording the change in ENS value from the initial moment to the current moment, forming an ENS time series for each monitoring point. The ENS time series of all monitoring points are then organized according to their spatial coordinate distribution order to clarify the increasing and decreasing trend and magnitude of ENS over time at each spatial location. This information is then integrated to form the ENS change trajectory of the key change area.

[0122] For each monitoring point, the field strength values ​​at two adjacent time points are selected. The change in field strength value is used as the magnitude of the vector, and the spatial direction of the increase in field strength value is used as the direction of the vector, thus establishing a direction vector for each monitoring point. The angle between the direction vectors of any two monitoring points is calculated. When the angle is 0 degrees, the correlation weight is set to the maximum value of 1. The weight decreases linearly for each degree increase in the angle, and the weight is set to 0 when the angle is 90 degrees or higher. For any two monitoring points, if their correlation weight is greater than 0, a connecting edge is established between the two monitoring points, and the weight of the edge is the correlation weight value. All monitoring points and the connecting edges between them together constitute a spatial correlation diagram of the key change area.

[0123] Extract the direction vectors of all monitoring points in the spatial correlation map, and convert the direction angle of each direction vector into a numerical range of 0-360 degrees. Divide the 0-360 degree range into 24 direction intervals, each in 15-degree intervals, and count the number of direction vectors contained in each interval. Identify the interval with the highest number of direction vectors; this interval is the dominant direction interval of the key change area.

[0124] Calculate the average angle of all direction vectors within the dominant direction interval, and use this average angle as the core gradient direction. Calculate the deviation between the direction vectors within the dominant direction interval and the average angle, and select the angles corresponding to direction vectors with a deviation of less than 5 degrees as supplementary gradient directions. Arrange the core gradient direction and all supplementary gradient directions in ascending order of angle to form the gradient direction set for the key change region.

[0125] For each gradient direction in the gradient direction set, the sum of the field strength changes at all monitoring points along that direction is calculated, and this sum is taken as the intensity value of that gradient direction. By comparing the intensity values ​​of all gradient directions, the gradient direction with the largest intensity value is identified, and this direction is the dominant change direction of the obstacle's influence domain.

[0126] The pre-stored strategy library contains multiple primitive strategies, each corresponding to a specific avoidance direction. Each primitive strategy's avoidance direction is pre-converted into an angle range of 0-360 degrees. The angle of the dominant change direction is compared with the angle range corresponding to each primitive strategy in the pre-stored strategy library. If the angle of the dominant change direction falls within the angle range of a certain primitive strategy, then the dominant change direction and that primitive strategy are considered to have a matching relationship. This matching relationship is recorded as the strategy matching relationship between the dominant change direction and the pre-stored strategy library.

[0127] If the strategy matching relationship corresponds to only one primitive strategy, then that primitive strategy is directly determined as the conflict resolution strategy for the orbital robot. If the strategy matching relationship corresponds to multiple primitive strategies, then the core avoidance logic of these primitive strategies is extracted. Combining the spatial range of the obstacle's influence domain and the intensity of field strength changes, a core avoidance direction with the highest priority is determined. Based on the primitive strategy corresponding to this core avoidance direction, auxiliary avoidance measures from other matching primitive strategies are integrated to form a unified and complete conflict resolution strategy for the orbital robot.

[0128] The beneficial effects are that by monitoring the spatiotemporal evolution of the field strength distribution in the obstacle's influence domain, tracking key change areas, and generating gradient direction sets, the dominant change direction of the obstacle's influence is accurately captured. Then, through primitive strategy matching and integration, a conflict resolution strategy is formed, ensuring that the strategy can accurately adapt to the dynamic change characteristics of the obstacle. This provides a scientific and effective basis for obstacle avoidance decision-making for the orbital robot, significantly improving the pertinence and effectiveness of robot conflict resolution and ensuring the safety of robot operation.

[0129] S4. Based on the conflict resolution strategy, correct the predetermined path of the orbital robot to obtain a conflict-free path trajectory segment of the orbital robot;

[0130] In this embodiment of the invention, the step of correcting the predetermined path of the orbital robot according to the conflict resolution strategy to obtain a conflict-free path trajectory segment of the orbital robot includes:

[0131] By deconstructing the avoidance decision elements in the conflict resolution strategy, the set of strategy elements of the conflict resolution strategy is obtained;

[0132] The strategy element set is input into the predetermined path of the orbital robot to obtain the dynamic path segment of the orbital robot;

[0133] The connection relationships of the dynamic path segments are reconstructed to obtain the topology-optimized path of the orbital robot;

[0134] The topology-optimized path is smoothed to obtain the continuous path trajectory of the orbital robot;

[0135] Verify the spatial relationship between the continuous path trajectory and the influence domain of the obstacle to obtain a conflict-free path trajectory segment of the orbital robot.

[0136] The conflict resolution strategy contains a complete decision-making logic to achieve the obstacle avoidance goal. It clearly breaks down the core avoidance instructions involved in the strategy, extracts key contents such as the specific direction of avoidance, the speed adjustment range during the avoidance process, the standard for maintaining a safe distance from the obstacle, the robot's posture adjustment angle requirements, and the time node for path switching. These independent and key decision information are organized into categories of "direction-speed-distance-posture-time" to form a set containing all the core elements of avoidance decision-making. This set is the strategy element set.

[0137] The predetermined path is a complete path from the starting point to the ending point pre-planned by the orbital robot. It contains multiple path nodes arranged in chronological order, each labeled with specific spatial coordinates and motion parameters. The various elements in the strategy element set are assigned to the corresponding nodes on the predetermined path according to the time nodes. The spatial coordinates of the nodes are adjusted according to the avoidance direction, the movement speed of the nodes is corrected according to the speed adjustment range, and the movement attitude of the nodes is optimized by adjusting the angle according to the safety distance standard and attitude. This allows the predetermined path to exhibit dynamic adjustments within the corresponding intervals to adapt to obstacle avoidance requirements. The adjusted path is then divided into several segments according to the element's effective range; each segment is a dynamic path fragment of the orbital robot.

[0138] The spatial coordinates, velocity, and direction of motion of the starting and ending nodes of each dynamic path segment are analyzed one by one. The spatial distance and motion parameter difference between the ending and starting nodes of adjacent segments are calculated. If the difference exceeds a preset connection threshold, the coordinates of the starting node of the next segment are adjusted to shorten the distance between it and the ending node of the previous segment to within the threshold range. At the same time, the velocity and direction of the starting node of the next segment are corrected to keep the motion parameters consistent with those of the ending node of the previous segment. The core path information of all dynamic path segments is retained, and only the connecting nodes between segments are optimized to form a path structure with smooth node connections and continuous motion parameters. This structure is the topology-optimized path of the orbital robot.

[0139] The topology-optimized path consists of multiple path segments, which may contain areas with excessively steep turns. Within these steep transition areas, transition nodes are added at uniform intervals. The coordinates of these transition nodes are determined by linear interpolation of the coordinates of adjacent original nodes, and their speeds gradually change according to the speed gradient of the original nodes. All original nodes and transition nodes are connected sequentially, and a smooth curve is used to fit the paths between nodes, ensuring that the tangent direction of the curve is continuous at the nodes without abrupt angles. This results in a continuous and smooth path, which is the continuous path trajectory of the orbital robot.

[0140] Obtain the spatial coordinates of all nodes on the continuous path trajectory, and simultaneously define the boundary coordinates of the spatial range of the obstacle's influence domain. Determine whether the coordinates of each node are within the boundary coordinate range of the obstacle's influence domain. If the node's coordinates are within the range, mark it as a conflict node; otherwise, mark it as a non-conflict node. Organize all non-conflict nodes chronologically, extract the path segments corresponding to consecutive non-conflict nodes, and remove path segments containing conflict nodes. The resulting continuous and conflict-free path segment is the conflict-free path trajectory segment of the orbital robot.

[0141] The beneficial effects are that by deconstructing the conflict resolution strategy, a precise set of strategy elements is obtained. Through path input, topology reconstruction, smoothing processing and spatial relationship verification, conflict-free path trajectory segments are gradually generated, ensuring that the path not only meets the obstacle avoidance requirements but also has continuous and smooth motion characteristics, completely avoiding the influence domain of obstacles, providing a safe and reliable driving path for the track robot, effectively reducing the risk of path conflict during obstacle avoidance, and improving the stability and smoothness of robot motion.

[0142] S5. Based on the decision preferences in the conflict resolution strategy, the effectiveness of the conflict-free path trajectory segment is determined to obtain the multi-dimensional evaluation results of the conflict-free path trajectory segment.

[0143] In this embodiment of the invention, the step of determining the effectiveness of the conflict-free path trajectory segment based on the decision preferences in the conflict resolution strategy to obtain a multi-dimensional evaluation result of the conflict-free path trajectory segment includes:

[0144] The key parameters of decision preferences in the conflict resolution strategy are used as evaluation benchmarks;

[0145] The motion state parameters of path points in the conflict-free path trajectory segment are collected, and the position coordinates and timestamps of the path points are recorded to construct the motion feature data of the conflict-free path trajectory segment.

[0146] The correlation analysis between the evaluation benchmark and the feature dimensions in the motion feature data is performed to obtain the coupled evaluation relationship between the evaluation benchmark and the motion feature data.

[0147] Based on the aforementioned coupling evaluation relationship, the effectiveness score of the path point is calculated, resulting in a score sequence for the path point. The calculation formula for the effectiveness score is as follows:

[0148] ;

[0149] In the formula, path point The aforementioned performance score, To evaluate the total number of benchmark parameters, For the first The weighting coefficients of each evaluation benchmark parameter, For the first The correlation coefficients between each evaluation benchmark parameter and its corresponding motion feature dimension For the first Normalized values ​​of each evaluation benchmark parameter For the first Normalized values ​​of each motion feature dimension. The preset stability factor, It is a natural exponential function;

[0150] The scoring sequence is reorganized to obtain a multi-dimensional evaluation result of the conflict-free path trajectory segment.

[0151] The key parameters of the decision preferences in the conflict resolution strategy are used as evaluation benchmarks. The decision preferences of the conflict resolution strategy clarify the core orientation of robot obstacle avoidance. Key parameters such as safety priority setting, speed efficiency threshold, posture stability requirements, and path smoothness standards are extracted from this strategy. These parameters are fixed requirements determined during strategy formulation and do not require additional adjustment. These parameters are directly integrated to form evaluation benchmarks to measure the performance of conflict-free path trajectory segments.

[0152] The motion state parameters of path points in the conflict-free path trajectory segment are collected, and the position coordinates and timestamps of the path points are recorded to construct the motion feature data of the conflict-free path trajectory segment. The conflict-free path trajectory segment consists of continuous path points, selected at equal time intervals. The motion state parameters of each path point are collected by sensors on the robot, including linear velocity, angular velocity, linear acceleration, angular acceleration, and attitude angle. At the same time, the three-dimensional spatial coordinates of each path point are obtained by the positioning module, and the acquisition timestamp of each path point is recorded by the timing module. The position coordinates and timestamps of each path point are mapped one-to-one with the corresponding motion state parameters, and the data is organized in chronological order to form the complete motion feature data of the conflict-free path trajectory segment.

[0153] The correlation analysis between the evaluation benchmark and the feature dimensions in the motion feature dataset is performed to obtain the coupled evaluation relationship between the evaluation benchmark and the motion feature data. First, the attribute categories of each parameter in the evaluation benchmark are clarified. For example, safety priority corresponds to the obstacle distance-related dimension in the motion feature data, speed efficiency threshold corresponds to the speed parameter dimension in the motion feature data, posture stability requirement corresponds to the posture angle and acceleration dimensions in the motion feature data, and path smoothness standard corresponds to the speed change rate dimension in the motion feature data. The correspondence between each evaluation benchmark parameter and each feature dimension in the motion feature data is determined one by one. If a certain evaluation benchmark parameter is directly related to a certain feature dimension and affects the performance judgment result, it is marked as a strong correlation; if it is indirectly related and has a small impact, it is marked as a weak correlation. All the correlation types and correspondences between evaluation benchmark parameters and feature dimensions are compiled and summarized to form the coupled evaluation relationship between the evaluation benchmark and the motion feature data.

[0154] The total number of evaluation benchmark parameters is the actual number of key parameters of decision preferences extracted from conflict resolution strategies. This number is determined by the core evaluation dimensions contained in the decision preferences and is a fixed statistical value.

[0155] No. The weighting coefficient of the first evaluation benchmark parameter is determined based on the coupling evaluation relationship between the evaluation benchmark and the motion characteristic data. If the first... Each evaluation benchmark parameter is strongly correlated with its corresponding motion feature dimension and is assigned a higher fixed weight value; if the correlation is weak, it is assigned a lower fixed weight value. The sum of the weight coefficients of all evaluation benchmark parameters is a fixed value to ensure the rationality of weight allocation.

[0156] No. The correlation coefficient between each evaluation benchmark parameter and its corresponding motion characteristic dimension was obtained by analyzing the changing patterns of both in historical data. This involved statistically analyzing the correlation coefficients within the same time period. The degree of fit between the changing trend of each evaluation benchmark parameter and the changing trend of the corresponding motion feature dimension is determined by the correlation coefficient. When the changing trends are completely consistent, the correlation coefficient is taken at its maximum value; when the changing trends are completely opposite, the correlation coefficient is taken at its minimum value; and when the degree of fit is between the two, the corresponding values ​​are allocated proportionally.

[0157] No. The normalized value of the first evaluation benchmark parameter is obtained by applying the first... The evaluation benchmark parameter is obtained by numerical standardization. First, the maximum and minimum values ​​of this benchmark parameter in historical applications are statistically analyzed. Then, the current benchmark parameter is used... The normalized value is obtained by subtracting the minimum value from the value of each evaluation benchmark parameter, and then dividing by the difference between the maximum and minimum values, within a fixed range.

[0158] No. The method for obtaining the normalized value of the first motion feature dimension is the same as that of the second. The normalized values ​​of each evaluation benchmark parameter are consistent. First, determine the maximum and minimum values ​​of the motion feature dimension in the historical data. Then, subtract the minimum value from the current value of the motion feature dimension and divide by the difference between the maximum and minimum values ​​to achieve the standardization transformation of the values.

[0159] The preset stability factor is a fixed value set manually based on the actual needs of the evaluation calculation. Its value is determined according to the smallest possible value of the denominator in historical calculations, ensuring that the calculation result will not fluctuate abnormally due to the denominator being too small.

[0160] The natural exponential function is used to quantify the first... The normalized value of the first evaluation benchmark parameter and the first The function calculates the degree of difference between the normalized values ​​of each motion feature dimension. When the two values ​​are exactly the same, the function result is the maximum value of 1; the greater the difference between the two values, the closer the function result is to 0. This variation pattern accurately reflects the degree of matching between the two.

[0161] The effectiveness score of each path point is calculated based on the aforementioned coupling evaluation relationship, resulting in a score sequence for each path point. During the calculation, for each evaluation benchmark parameter, its weight coefficient, correlation coefficient with the corresponding motion feature dimension, and the result of the natural exponential function reflecting the degree of matching between their normalized values ​​are multiplied together. Then, the products of all evaluation benchmark parameters are summed to obtain the numerator. Simultaneously, the weight coefficients of all evaluation benchmark parameters are added together to obtain the denominator. Dividing the numerator by the denominator yields the effectiveness score of that path point. This score comprehensively considers the importance of each evaluation benchmark parameter, the closeness of its correlation with the motion feature dimension, and the degree of numerical matching, thus comprehensively and objectively reflecting the overall performance of the path point. The effectiveness scores of each path point are arranged sequentially according to their temporal order within the conflict-free path trajectory segment, forming a score sequence for each path point.

[0162] The scoring sequence is reorganized to obtain a multi-dimensional evaluation result of the conflict-free path trajectory segment. Based on the core categories of the evaluation benchmark, the scoring sequence is divided into sub-sequences corresponding to three feature dimensions: safety, efficiency, and stability. The safety sub-sequence includes scores for all path points related to safety priority; the efficiency sub-sequence includes scores for all path points related to speed efficiency thresholds; and the stability sub-sequence includes scores for all path points related to attitude stability requirements and path smoothness standards. The average score of each sub-sequence is calculated as the evaluation score for the corresponding dimension. Simultaneously, the percentage of path points in each sub-sequence with scores above a set pass threshold is calculated as the pass rate for the corresponding dimension. The evaluation scores and pass rates for the three dimensions are integrated and presented in the order of "safety-efficiency-stability" to form the multi-dimensional evaluation result of the conflict-free path trajectory segment.

[0163] The beneficial effects are as follows: by using the key parameters of the decision preference of the conflict resolution strategy as the evaluation benchmark, the motion characteristic data of the path points are comprehensively collected and correlation analysis is performed. The path point effectiveness score is obtained by combining the coupled evaluation relationship and the accurate calculation method that comprehensively considers various factors. After feature recombination, multi-dimensional evaluation results are obtained, ensuring that the evaluation results meet the obstacle avoidance decision requirements and fully reflect the path effectiveness. This provides an accurate and reliable decision basis for the subsequent generation of emergent motion strategies, and helps to improve the overall effectiveness of obstacle avoidance operation of the track robot.

[0164] S6. By integrating the multi-dimensional evaluation results, the emergent motion strategy of the orbital robot is obtained.

[0165] In this embodiment of the invention, the step of fusing the multi-dimensional evaluation results to obtain the emergent motion strategy of the orbital robot includes:

[0166] By integrating the security, efficiency, and stability dimensions from the multi-dimensional evaluation results, a combination of strategy elements from the multi-dimensional evaluation results is obtained.

[0167] Based on the topological structure of the path knowledge graph, spatial constraints are applied to the combination of strategy elements to obtain the restricted strategy set of the orbital robot;

[0168] Based on the decision preferences in the conflict resolution strategy, the set of restricted strategies is prioritized to obtain the strategy behavior sequence of the orbital robot;

[0169] The emergent motion strategy of the orbital robot is obtained by spatiotemporally aligning the strategy behavior sequence with the conflict-free path trajectory segment.

[0170] By integrating the safety, efficiency, and stability dimensions from the multi-dimensional evaluation results, a strategy element combination based on the multi-dimensional evaluation results is obtained. The multi-dimensional evaluation results include the evaluation score and pass rate for each dimension. Core requirements such as safe distance compliance standards and obstacle avoidance success rate are extracted from the safety dimension. Key indicators such as speed maintenance range and path time upper limit are extracted from the efficiency dimension. Core parameters such as attitude fluctuation allowable value and speed change rate threshold are extracted from the stability dimension. These core requirements, key indicators, and core parameters from the three dimensions are categorized and summarized, duplicate information is removed, and the core constraints and guiding information of each dimension for strategy formulation are retained, forming a strategy element combination covering the core requirements of safety, efficiency, and stability.

[0171] Based on the topology of the path knowledge graph, spatial constraints are applied to the strategy element combinations to obtain the restricted strategy set for the orbital robot. The topology of the path knowledge graph includes the spatial coordinates of track-related nodes, the connection relationships between nodes, the physical boundary range of the track, and the reachable path rules between nodes. Referring to this topology, each element in the strategy element combination is verified one by one. Path adjustment elements exceeding the physical boundary range of the track are eliminated, strategy options that disrupt the inherent connection relationships between nodes are excluded, and action commands that do not conform to the reachable path rules between nodes are discarded. Only strategy elements that fully adapt to the spatial coordinates, connection relationships, boundary range, and reachability rules in the topology are retained. These constrained elements are then sorted according to their original categories to form the restricted strategy set for the orbital robot.

[0172] Based on the decision preferences in the conflict resolution strategy, the set of restricted strategies is prioritized to obtain the strategy behavior sequence of the orbital robot. The decision preferences of the conflict resolution strategy clearly define the core orientation that safety priority is higher than efficiency priority, and efficiency priority is higher than stability priority. For each strategy in the set of restricted strategies, the degree to which it meets the core requirements of the safety dimension is first evaluated, with higher safety compliance rate resulting in higher priority; if the safety compliance rate is the same, the level of compliance with the key indicators of the efficiency dimension is evaluated, with better efficiency performance resulting in higher priority; when safety and efficiency performance are consistent, the fit of the stability dimension is considered, with better stability resulting in higher priority. Following this logic, the priority of all strategies is completed, and the ranked strategies are arranged in descending order to form an ordered strategy behavior sequence of the orbital robot.

[0173] The emergent motion strategy of the orbital robot is obtained by spatiotemporally aligning the strategy behavior sequence with the conflict-free path trajectory segment. The conflict-free path trajectory segment contains path points arranged by timestamps and their corresponding spatial coordinates and motion parameters. Each strategy in the strategy behavior sequence has a clearly defined execution time interval and core action command. Based on the timestamp, the execution time interval of each strategy in the strategy behavior sequence is matched with the corresponding time range in the trajectory segment, so that each strategy accurately corresponds to a path point within a specific time period in the trajectory segment. Spatially, the action commands in the strategy are combined with the spatial coordinates of the corresponding path point to clarify the specific strategy action to be executed at each path point. By integrating all spatiotemporally aligned strategy actions and path trajectory information, a continuous, coherent, and real-time obstacle avoidance-adaptive emergent motion strategy for the orbital robot is formed.

[0174] The beneficial effects are that by integrating the core elements of multi-dimensional evaluation results, combining the spatial constraints of the path knowledge graph with the decision preference ranking of conflict resolution strategies, and then achieving precise fusion of strategies and trajectories through spatiotemporal alignment, the resulting emergent motion strategies not only comprehensively consider safety, efficiency and stability, but also conform to the actual operational constraints of the orbital space. They can accurately adapt to dynamic obstacle avoidance scenarios, effectively improve the scientific nature of the orbital robot's motion decision-making and the reliability of its execution, and ensure that the robot can complete its tasks stably, efficiently and safely in complex environments.

[0175] like Figure 2 The diagram shown is a functional block diagram of a path obstacle avoidance data update system for a track robot provided in an embodiment of the present invention.

[0176] The track robot path obstacle avoidance data update system 100 of the present invention can be installed in an electronic device. Depending on the functions implemented, the track robot path obstacle avoidance data update system 100 may include a feature vectorization and anomaly detection module 101, a topology reconstruction and influence domain generation module 102, a gradient analysis and policy mapping module 103, a path correction and trajectory generation module 104, an efficiency evaluation and decision optimization module 105, and a policy fusion and emergent control module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0177] In this embodiment, the functions of each module / unit are as follows:

[0178] The feature vectorization and anomaly detection module 101 is used to perform vector assembly on multi-dimensional features in the real-time environmental data stream of the track robot to obtain the primary feature vector of the multi-dimensional features, and to perform anomaly pattern identification on the primary feature vector and the historical normal pattern library to obtain the dynamic anomaly features of the real-time environmental data stream.

[0179] The topology reconstruction and influence domain generation module 102 is used to reconstruct the local topology of the pre-stored path knowledge graph based on the dynamic anomaly characteristics, so as to generate the obstacle influence domain of the orbital robot.

[0180] The gradient analysis and strategy mapping module 103 is used to analyze the field strength distribution change trend of the obstacle influence domain, obtain the gradient field of the field strength distribution change trend, and map the dominant direction of the gradient field to the conflict resolution strategy of the orbital robot.

[0181] The path correction and trajectory generation module 104 is used to correct the predetermined path of the orbital robot according to the conflict resolution strategy, so as to obtain a conflict-free path trajectory segment of the orbital robot.

[0182] The performance evaluation and decision optimization module 105 is used to determine the performance of the conflict-free path trajectory segment based on the decision preferences in the conflict resolution strategy, and obtain a multi-dimensional evaluation result of the conflict-free path trajectory segment.

[0183] The strategy fusion and emergence control module 106 is used to fuse the multi-dimensional evaluation results to obtain the emergence motion strategy of the orbital robot.

[0184] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0185] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0186] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0187] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0188] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0189] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for updating obstacle avoidance data for a tracked robot, characterized in that, The method includes: S1. Perform vector assembly on the multi-dimensional features in the real-time environmental data stream of the track robot to obtain the primary feature vector of the multi-dimensional features, and perform abnormal pattern identification on the primary feature vector and the historical normal pattern library to obtain the dynamic abnormal features of the real-time environmental data stream. S2. Based on the dynamic anomaly characteristics, reconstruct the local topology of the pre-stored path knowledge graph to generate the obstacle influence domain of the orbital robot; S3. Analyze the field strength distribution change trend of the obstacle's influence domain, obtain the gradient field of the field strength distribution change trend, and map the dominant direction of the gradient field to the conflict resolution strategy of the orbital robot. S4. Based on the conflict resolution strategy, correct the predetermined path of the orbital robot to obtain a conflict-free path trajectory segment of the orbital robot; S5. Based on the decision preferences in the conflict resolution strategy, the effectiveness of the conflict-free path trajectory segment is determined to obtain the multi-dimensional evaluation results of the conflict-free path trajectory segment. S6. By integrating the multi-dimensional evaluation results, the emergent motion strategy of the orbital robot is obtained.

2. The method for updating path obstacle avoidance data for a tracked robot as described in claim 1, characterized in that, The process involves assembling vectors from multi-dimensional features in the real-time environmental data stream of the orbital robot to obtain primary feature vectors, and then comparing these primary feature vectors with a historical normal pattern library to identify abnormal patterns, thereby obtaining dynamic abnormal features of the real-time environmental data stream, including: Spatial position features, motion state features, and environmental perception features are extracted from the real-time environmental data stream of the orbital robot to obtain the multi-dimensional original features of the real-time environmental data stream; The multidimensional original features are subjected to dimension consistency transformation to obtain dimensionless features of the multidimensional original features; Tensor synthesis is performed on the dimensionless features to obtain the primary feature vector of the dimensionless features; Calculate the difference between the primary feature vector and the pattern vectors in the historical normal pattern library to obtain the difference sequence of the primary feature vector; By analyzing the abnormal pattern distribution of the difference sequence, the dynamic abnormal characteristics of the real-time environmental data stream are obtained.

3. The method for updating path obstacle avoidance data for a tracked robot as described in claim 2, characterized in that, The formula for calculating the degree of difference is as follows: ； In the formula, The first in the historical normal pattern library A pattern vector, The primary feature vector and the first The degree of difference between pattern vectors For the primary feature vector, This is a preset numerical stability factor. The balance coefficients were obtained to analyze the distribution characteristics of pattern vectors in the historical normal pattern library. The standard deviation of the eigenvalues ​​in the primary eigenvector is given. For the first The standard deviation of the eigenvalues ​​in each pattern vector The preset distribution smoothing factor, To obtain the maximum value, Let L2 norm be the vector. To take the absolute value.

4. The method for updating path obstacle avoidance data for a tracked robot as described in claim 1, characterized in that, The step of reconstructing the local topology of the pre-stored path knowledge graph based on the dynamic anomaly characteristics to generate the obstacle influence domain of the orbital robot includes: The dynamic anomaly features are mapped to a pre-stored path knowledge graph to construct the anomaly-node mapping relationship between the dynamic anomaly features and the path knowledge graph. Based on the anomaly-node mapping relationship, the adjacency network of the affected nodes in the path knowledge graph is constructed to obtain the local topology of the affected nodes; Optimize the connection strength of the edges in the local topology to obtain the reconstructed topology network of the local topology; The interaction relationships between nodes in the reconstructed topology network are quantified to obtain the node influence magnitude distribution of the reconstructed topology network; Spatial clustering of the influence distribution of the nodes is performed to obtain the obstacle influence domain of the orbital robot.

5. The method for updating path obstacle avoidance data for a tracked robot as described in claim 1, characterized in that, The process of analyzing the field strength distribution variation trend in the obstacle's influence domain to obtain the gradient field of the field strength distribution variation trend, and mapping the dominant direction of the gradient field to the conflict resolution strategy of the orbital robot, includes: By monitoring the spatiotemporal evolution pattern of field strength distribution within the obstacle's influence domain, key change areas within the obstacle's influence domain can be obtained. Analyze the spatial variation law of the field strength value in the key changing region, and construct the gradient direction set of the key changing region; The intensity contrast between the gradient direction sets is evaluated to obtain the dominant change direction of the obstacle's influence domain; The dominant change direction is matched with the primitive strategies in the pre-stored strategy library to obtain the matching relationship between the dominant change direction and the strategy in the pre-stored strategy library. The strategy matching relationship is integrated into the conflict resolution strategy of the orbital robot.

6. The method for updating path obstacle avoidance data for a tracked robot as described in claim 5, characterized in that, The analysis of the spatial variation law of the field strength value in the key changing region and the construction of the gradient direction set of the key changing region include: By tracing the spatial variation characteristics of the field strength values ​​in the key variation region, the trajectory of the field strength variation in the key variation region can be obtained; Establish the direction vector of the monitoring point in the field strength change trajectory, assign correlation weights according to the angle relationship between the direction vectors, and connect the monitoring points through the correlation weights to generate a spatial correlation map of the key change area; By dividing the directional clustering intervals in the spatial correlation diagram, the dominant directional intervals of the key change regions are obtained; Based on the statistical characteristics of the dominant direction interval, a gradient direction set for the key change region is generated.

7. The method for updating path obstacle avoidance data for a tracked robot as described in claim 1, characterized in that, The step of correcting the predetermined path of the orbital robot according to the conflict resolution strategy to obtain a conflict-free path trajectory segment of the orbital robot includes: By deconstructing the avoidance decision elements in the conflict resolution strategy, the set of strategy elements of the conflict resolution strategy is obtained; The strategy element set is input into the predetermined path of the orbital robot to obtain the dynamic path segment of the orbital robot; The connection relationships of the dynamic path segments are reconstructed to obtain the topology-optimized path of the orbital robot; The topology-optimized path is smoothed to obtain the continuous path trajectory of the orbital robot; Verify the spatial relationship between the continuous path trajectory and the influence domain of the obstacle to obtain a conflict-free path trajectory segment of the orbital robot.

8. The method for updating path obstacle avoidance data for a tracked robot as described in claim 1, characterized in that, The step of evaluating the effectiveness of the conflict-free path segment based on the decision preferences in the conflict resolution strategy, and obtaining a multi-dimensional evaluation result of the conflict-free path segment, includes: The key parameters of decision preferences in the conflict resolution strategy are used as evaluation benchmarks; The motion state parameters of path points in the conflict-free path trajectory segment are collected, and the position coordinates and timestamps of the path points are recorded to construct the motion feature data of the conflict-free path trajectory segment. The correlation analysis between the evaluation benchmark and the feature dimensions in the motion feature data is performed to obtain the coupled evaluation relationship between the evaluation benchmark and the motion feature data. Based on the aforementioned coupling evaluation relationship, the effectiveness score of the path point is calculated, resulting in a score sequence for the path point. The calculation formula for the effectiveness score is as follows: ； In the formula, path point The aforementioned performance score, To evaluate the total number of benchmark parameters, For the first The weighting coefficients of each evaluation benchmark parameter, For the first The correlation coefficients between each evaluation benchmark parameter and its corresponding motion feature dimension For the first Normalized values ​​of each evaluation benchmark parameter For the first Normalized values ​​of each motion feature dimension. The preset stability factor, It is a natural exponential function; The scoring sequence is reorganized to obtain a multi-dimensional evaluation result of the conflict-free path trajectory segment.

9. The method for updating path obstacle avoidance data for a tracked robot as described in claim 1, characterized in that, The emergenceary motion strategy of the orbital robot, obtained by fusing the multi-dimensional evaluation results, includes: By integrating the security, efficiency, and stability dimensions from the multi-dimensional evaluation results, a combination of strategy elements from the multi-dimensional evaluation results is obtained. Based on the topological structure of the path knowledge graph, spatial constraints are applied to the combination of strategy elements to obtain the restricted strategy set of the orbital robot; Based on the decision preferences in the conflict resolution strategy, the set of restricted strategies is prioritized to obtain the strategy behavior sequence of the orbital robot; The emergent motion strategy of the orbital robot is obtained by spatiotemporally aligning the strategy behavior sequence with the conflict-free path trajectory segment.

10. A path obstacle avoidance data update system for a tracked robot, used to implement the path obstacle avoidance data update method for a tracked robot as described in claim 1, the system comprising: The feature vectorization and anomaly detection module is used to assemble multi-dimensional features in the real-time environmental data stream of the track robot into vectors to obtain the primary feature vectors of the multi-dimensional features, and to identify the anomaly patterns by comparing the primary feature vectors with the historical normal pattern library to obtain the dynamic anomaly features of the real-time environmental data stream. The topology reconstruction and influence domain generation module is used to reconstruct the local topology of the pre-stored path knowledge graph based on the dynamic anomaly characteristics, so as to generate the obstacle influence domain of the orbital robot. The gradient analysis and strategy mapping module is used to analyze the field strength distribution change trend of the obstacle influence domain, obtain the gradient field of the field strength distribution change trend, and map the dominant direction of the gradient field to the conflict resolution strategy of the orbital robot. The path correction and trajectory generation module is used to correct the predetermined path of the orbital robot according to the conflict resolution strategy, so as to obtain a conflict-free path trajectory segment of the orbital robot. The performance evaluation and decision optimization module is used to determine the performance of the conflict-free path trajectory segment based on the decision preferences in the conflict resolution strategy, and obtain a multi-dimensional evaluation result of the conflict-free path trajectory segment. The strategy fusion and emergent control module is used to fuse the multi-dimensional evaluation results to obtain the emergent motion strategy of the orbital robot.

Images