Event-driven control method for rescue robot

Abstract

The invention relates to the technical field of event-driven control, in particular to an event-driven control method for a rescue robot, which comprises the following steps of: extracting event triggering conditions, triggering time sequence data and state parameters based on identification and serial number management of various events in a rescue environment; event type classification, spatial region partitioning and triggering probability modeling are carried out in combination with an SPFA algorithm, the corresponding relation between an event and a response state can be accurately mapped in the space-time dimension, the structural expression and state scheduling capability of a complex environment is enhanced, path optimization is carried out in combination with weights, and the path optimization efficiency is improved. Quantitative allocation of response resources and quantitative evaluation of response execution efficiency can be realized, a response path cost model is constructed by adopting a Dijkstra algorithm, and a mapping relation between path suitability and a triggering probability is compared, so that dynamic screening of an optimal path can be realized under an environment constraint condition; and the execution accuracy of the response path and the task completion collaboration are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of event-driven control, and in particular to an event-driven control method for a rescue robot. Background Art

[0002] Event-driven control technology belongs to the field of control engineering, especially technologies related to automation and robotic control systems. It starts and adjusts the behavior and state of the system through the triggering of external and internal events. It does not rely on timers and predetermined time intervals, but instead starts corresponding control actions when a specific event occurs, improving response speed and efficiency, adapting to environmental changes and uncertainties, and being able to adjust system strategies and behaviors in real time when executing tasks in a dynamic environment.

[0003] A rescue robot event-driven control method aims to optimize the rescue robot's response speed and task execution efficiency. When the rescue robot performs tasks in a complex and dynamically changing environment, an event-driven mechanism is introduced to trigger the control system to perform corresponding operations based on real-time detected environmental changes. The purpose is to ensure that the robot can respond quickly to external events while improving the accuracy and safety of task execution.

[0004] Existing technologies rely on event triggering conditions to schedule control actions. However, when dealing with the intersection of complex events and the evolution of spatial dynamic environments, there are problems such as unclear control behavior response boundaries, lack of cost evaluation of path selection, and unclear task priority processing. In addition, they emphasize immediate response to event triggering. In the absence of state mapping and path cost modeling, they are prone to problems such as redundant response paths and simultaneous execution of conflicting events, affecting the overall task execution continuity of the system. At the same time, existing technologies do not incorporate cost parameters such as energy consumption and response delay into the evaluation system. Path execution lacks quantitative standards, which is not conducive to path optimization in resource-limited scenarios. A mapping model between paths and environmental states has not been established, and the task path cannot be dynamically corrected according to the environmental state, resulting in a lack of adaptability of the task execution plan in the face of environmental changes, reducing the stability and robustness of task completion. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an event-driven control method for a rescue robot.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a rescue robot event-driven control method, comprising the following steps:

[0007] S1: Identify various events in the rescue environment, obtain the triggering conditions and time series data of each event, use the SPFA algorithm to classify event types, spatially partition the environment, record the event triggering probability corresponding to each partition, combine the event response strategy and weight to perform state mapping, and generate an event state mapping model;

[0008] S2: Based on the event state mapping model, calculate the response delay, task completion rate and energy consumption parameters of the event state, evaluate each response path according to the weight, optimize the path selection, and generate an event response optimization model;

[0009] S3: Based on the event response optimization model, the Dijkstra algorithm is used to compare the trigger probability of the event state with the response path cost to perform path selection and judgment. The adaptive response path is selected by comparing the path cost flow, and the event response path solution is generated by combining the state mapping information.

[0010] S4: Based on the event response path plan, determine whether the current event conflicts with an event that has already occurred. If a conflict occurs, sort the conflicting events by priority, roll back the response path with a lower priority, adjust and optimize the response sequence, and generate an event conflict resolution path.

[0011] S5: Based on the event conflict resolution path, obtain environmental change information and event triggering timing, adjust the control strategy in combination with real-time feedback, recalculate the adaptive task path, and generate a dynamic task path planning solution through path adjustment and task replanning.

[0012] As a further solution of the present invention, the specific steps of generating the event state mapping model are:

[0013] Identify various events in the rescue environment, match event names and numbers, mark trigger locations, read trigger time nodes, and extract status parameters. After data formatting, field column cleaning, numerical anomalies removal, and linear sorting of time series, an event trigger data set is generated.

[0014] Based on the event triggering dataset, the SPFA algorithm is used to extract the event type field, perform statistical coding on the tag value, obtain the corresponding spatial coordinates, set the boundaries of the spatial partitions, mark the partition numbers, construct a distribution matrix and the correspondence between event type and spatial location, and generate a spatial area event distribution map;

[0015] Based on the spatial area event distribution map, the number of event records in each partition is counted, the average trigger frequency is extracted, and a comparison is made according to the historical response action list. The weight value is verified, and a mapping relationship between the response status value and the weight is established. At the same time, the partition number is bound to the mapping relationship to generate an event status mapping model.

[0016] As a further solution of the present invention, the SPFA algorithm, according to the formula

[0017]

[0018] in: Represents the spatial region from the source event node to the target node The shortest path distance, Indicates the path from the source event node to the current node The shortest path distance, Representation node With node The event label weight value between represents the event tag semantic similarity adjustment coefficient, Representation node and The timestamp difference of the corresponding events, The weight coefficient representing the influence of time difference, Representation node With node The cosine value of the spatial angle between The weight coefficient representing the consistency of spatial propagation direction, Representation node The event density coefficient of the spatial region, represents the event density suppression coefficient.

[0019] As a further solution of the present invention, the SPFA algorithm first takes any starting node in the graph to be processed as the source point, initializes the shortest path from the source point to itself to 0, sets the path values ​​of the remaining nodes to positive infinity, and initializes a queue to store the nodes to be relaxed. After the source point is added to the queue, the current node is taken out of the queue in turn, and the target nodes pointed to by all its outgoing edges are traversed. The path is judged according to the path relaxation principle. If it exists, the path value of the target node is updated, and on the premise that the target node is not in the queue, the queue is added to continue processing, and the number of times each node enters the queue is recorded. If the number of times any node enters the queue exceeds the total number of nodes, it means that there is a negative cycle in the graph, and the algorithm is terminated. Under the condition of ensuring that there is no negative weight cycle in the graph, the calculation of the path of each node is gradually completed.

[0020] As a further solution of the present invention, the specific steps of generating the event response optimization model are:

[0021] Based on the event state mapping model, the response delay, task completion rate, and energy consumption parameters corresponding to each event state are calculated. By weighted evaluation of the parameters in each response path, an adaptive path is selected, and its path cost is compared with the response performance to generate an event response parameter evaluation value.

[0022] Based on the event response parameter evaluation value, different response paths are prioritized, and based on the cost and weight of each path, the cost flow of path selection is analyzed, the path selection strategy is optimized, and combined with the path evaluation, an event response optimization model is generated.

[0023] As a further solution of the present invention, the specific steps of generating the event response path solution are:

[0024] Based on the event response optimization model, the trigger probability and response path cost of each event state are obtained, and the Dijkstra algorithm is used to calculate the adaptability of the path. By comparing the responsiveness and cost flow of the path, the execution effect of each path is evaluated, and whether the cost flow of each path meets the task requirements is determined, and a path selection evaluation result is generated;

[0025] Based on the path selection evaluation results, the adaptive response path is screened out by comparing the cost flow of the path with the trigger probability, analyzing whether the path conforms to the current environmental changes and task objectives, adjusting the optimized path, and generating the adaptive response path;

[0026] Based on the adaptive response path, each path is compared with the environmental state mapping information, the execution feasibility of the path is evaluated, the environmental changes and task requirements that may be encountered during the execution of the path are checked, and an event response path plan is generated.

[0027] As a further solution of the present invention, the Dijkstra algorithm, according to the formula

[0028]

[0029] in: Indicates the starting node to the node The minimum path evaluation cost value, Indicates the starting node to the node The known minimum path evaluation cost value, A set of event nodes representing completed path cost updates, Representation node To Node The basic path cost value, Represents a path segment arrive The risk correction factor, Representation node To Node The probability of successful event triggering, Indicates the adjustment coefficient of the trigger probability, Represents a path segment arrive responsiveness score, Represents the weight adjustment coefficient of responsiveness, represents the structural accessibility coefficient.

[0030] As a further solution of the present invention, the Dijkstra algorithm first selects the source node in the graph, initializes the path distance to 0, initializes the path distances of all other nodes to positive infinity, and constructs a set of unvisited nodes, records the nodes to be processed, selects a node from the set as the current processing node according to the distance value in each round, and marks it as visited, then traverses all adjacent nodes. If the distance from the node to the adjacent node is less than the currently recorded distance of the adjacent node, the distance value of the adjacent node is updated, and the predecessor node of the adapted path is recorded until all nodes are visited and the adapted path of the target node is confirmed.

[0031] As a further solution of the present invention, the specific steps of generating the event conflict resolution path are:

[0032] Based on the event response optimization model, the trigger probability of each path and the cost data of the response path are obtained, the cost value and trigger conditions between the paths are extracted, the relationship between the cost and the trigger probability is compared, the adaptation path that meets the trigger conditions is screened out, and a comparison result of the path cost and the trigger probability is generated;

[0033] Based on the comparison results of the path cost and trigger probability, the response time and execution order information of each path are extracted, the adaptability of each path is evaluated, the path selection is optimized, the path priority is adjusted in combination with the state mapping information, the path order is optimized and adjusted, and the event response path plan is generated.

[0034] As a further solution of the present invention, the specific steps of generating the dynamic task path planning solution are:

[0035] Based on the event conflict resolution path, environmental change information is obtained and the timing of event triggering is recorded. By monitoring real-time environmental data, the state changes of each event and environmental feedback are analyzed, control parameters are adjusted, and environmental and event feedback data are generated;

[0036] Based on the environment and event feedback data, the task path corresponding to each event is recalculated, and by adjusting the node order and task priority in the path, the path is replanned and the execution order is optimized to generate a dynamic task path planning solution.

[0037] Compared with the prior art, the advantages and positive effects of the present invention are:

[0038] In this invention, by identifying and numbering various events in the rescue environment, extracting event trigger conditions, trigger time series data and state parameters, and combining the SPFA algorithm to classify event types, partition spatial regions and model trigger probability, it is possible to accurately map the correspondence between events and response states in the spatiotemporal dimension, thereby enhancing the structured expression and state scheduling capabilities of complex environments.

[0039] In this invention, by introducing response delay, task completion rate and energy consumption parameters as comprehensive evaluation indicators of response paths, and combining weights for path optimization, it helps to achieve quantitative allocation of response resources and quantitative evaluation of response execution efficiency. The response path cost model is constructed using the Dijkstra algorithm, and the mapping relationship between path adaptability and trigger probability is compared. It can realize dynamic screening of the optimal path under environmental constraints, and improve the execution accuracy of the response path and the synergy of task completion.

[0040] In the present invention, by identifying the mutual interference relationship between events and combining response priorities for sorting and response path rollback, response consistency is guaranteed in the case of high-frequency dynamic event triggering. By combining real-time environmental changes with event triggering timing, feedback is provided to correct control strategies and task paths, forming a closed-loop control feedback mechanism in rescue missions, and realizing re-planning and response execution of multi-path tasks in a dynamic environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the main steps of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0043] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0044] Example 1

[0045] See also Figure 1The present invention provides a technical solution: an event-driven control method for a rescue robot, comprising the following steps:

[0046] S1: Identify various events in the rescue environment, obtain the triggering conditions and time series data of each event, use the SPFA algorithm to classify event types, spatially partition the environment, record the event triggering probability corresponding to each partition, combine the event response strategy and weight to perform state mapping, and generate an event state mapping model;

[0047] S2: Based on the event state mapping model, the response delay, task completion rate, and energy consumption parameters of the event state are calculated, each response path is evaluated according to the weight, the path selection is optimized, and the event response optimization model is generated;

[0048] S3: Based on the event response optimization model, the Dijkstra algorithm is used to compare the trigger probability of the event state with the response path cost to perform path selection and judgment. By comparing the path cost flow, an adaptive response path is selected. Combined with the state mapping information, an event response path solution is generated.

[0049] S4: Based on the event response path plan, determine whether the current event conflicts with an event that has already occurred. If a conflict occurs, sort the conflicting events by priority, roll back the low-priority response path, adjust and optimize the response sequence, and generate an event conflict resolution path.

[0050] S5: Based on the event conflict resolution path, obtain environmental change information and event triggering timing, adjust the control strategy based on real-time feedback, recalculate the adaptive task path, and generate a dynamic task path planning solution through path adjustment and task replanning.

[0051] The specific steps to generate the event state mapping model are:

[0052] Identify various events in the rescue environment, match event names and numbers, mark trigger locations, read trigger time nodes, and extract status parameters. After data formatting, field column cleaning, numerical anomalies removal, and linear sorting of time series, an event trigger data set is generated.

[0053] Based on the event-triggered dataset, the SPFA algorithm is used to extract the event type field, perform statistical coding on the tag values, obtain the corresponding spatial coordinates, set the boundaries of the spatial partitions, and mark the partition numbers. The distribution matrix and the correspondence between event types and spatial locations are constructed to generate a spatial regional event distribution map.

[0054] Based on the spatial area event distribution map, the number of event records in each partition is counted, the average trigger frequency is extracted, and the response action list is matched against the historical response action list. The weight value is verified, and a mapping relationship between the response status value and the weight is established. At the same time, the partition number is bound to the mapping relationship to generate an event status mapping model.

[0055] Based on the event record data collected in the rescue environment, the KMP algorithm is used to match the event name and event number. The main string input parameter is the complete text in the event record data, and the pattern string input parameter is the standard event name and number list. The pattern string prefix array is constructed and initialized with all values ​​0. The matching step is set to 1, and the scanning direction is set to left to right. During the matching process, if the characters at the corresponding positions of the main string and the pattern string are different, the pattern string position is retracted according to the prefix array and the matching is continued. After a successful match, the event name and event number are recorded. The regular expression processing method is used to read the trigger position. The string segmentation processing method is used for state parameter extraction. The input parameter is the state parameter field text. The delimiter is set to a colon ":". Each field is processed in a field column manner. The delimiter parameter is set to a comma ",". When performing data format normalization, the numerical type conversion is used and all fields are converted to floating point numbers. When removing numerical anomalies, the anomaly threshold parameter is set to 99999, and data records with values ​​greater than the anomaly threshold are deleted. The time node field is used as the sorting basis. The ascending sorting rule is set during the sorting process. All data records are processed to generate an event trigger data set.

[0056] Based on the event-triggered data set, the SPFA algorithm is used for path search, and the node information is stored in the adjacency table. The initial distance of all nodes is set to positive infinity, and the initial distance of the source node is set to 0. When the queue is initialized, the source node is added to the queue. The loop condition is that the queue is not empty. After the current node is dequeued, all adjacent nodes are traversed. If the distance between the adjacent nodes is greater than the distance between the current node and the edge weight, the distance between the adjacent nodes is updated to the distance between the current node and the edge weight, and the updated adjacent nodes are added to the queue. When the label value is statistically encoded, the initial value of the label number is set to 0, and the event type field is read one by one. The event type is not repeated and the label number is assigned incrementally. When reading the spatial coordinates, the longitude and latitude fields are directly read. When setting the spatial partition boundary, Set the minimum and maximum longitude values ​​to the minimum and maximum longitudes in the data, respectively, and the minimum and maximum latitude values ​​to the minimum and maximum latitudes in the data, respectively. Set the longitude partition step to 0.1 and the latitude partition step to 0.1. When generating the partition number, concatenate the longitude area code and the latitude area code as strings to generate a unique partition number. When constructing the distribution matrix, set the matrix rows and columns to the total number of longitude area codes and the total number of latitude area codes, respectively. Initialize the matrix elements to 0. After reading each event record, determine the corresponding partition based on the longitude and latitude, and add 1 to the corresponding position value in the matrix. When establishing the correspondence between event type and spatial position, set the association structure to a dictionary structure, with the key being the event type label number and the value being the corresponding partition number set, to generate a spatial area event distribution map.

[0057] Based on the spatial area event distribution map, when counting the number of event records in each partition, read the value of each position in the distribution matrix, extract the number of events in each partition, and read the cumulative number of events in each partition and the total number of event records when calculating the average trigger frequency. Calculate the average frequency value. When performing comparison and matching according to the historical response action list, set the matching field to the event type field, read the event type field of the historical response action list and match it one-to-one with the event type field of the current spatial area event distribution map. When verifying the weight value, read the historical response action weight field value and the current partition statistical weight value. Set the allowable error to 0.001, and the match is considered within the allowable error range. When establishing the mapping relationship between the response status value and the weight, set the mapping structure to a dictionary structure, the key is the response status value, and the value is the statistical weight value. When binding the partition number and the mapping relationship, set the storage structure to a nested dictionary, the outer key is the partition number, the inner key is the response status value, and the inner value is the corresponding weight value to generate an event status mapping model.

[0058] SPFA algorithm, according to the formula

[0059]

[0060] in: Represents the spatial region from the source event node to the target node The shortest path distance, Indicates the path from the source event node to the current node The shortest path distance, Representation node With node The event label weight value between represents the event tag semantic similarity adjustment coefficient, Representation node and The timestamp difference of the corresponding events, The weight coefficient representing the influence of time difference, Representation node With node The cosine value of the spatial angle between The weight coefficient representing the consistency of spatial propagation direction, Representation node The event density coefficient of the spatial region, represents the event density suppression coefficient;

[0061] Execution process: First, take any starting node in the graph to be processed as the source point, initialize the shortest path from the source point to itself to 0, set the path values ​​of the remaining nodes to positive infinity, and initialize a queue to store the nodes to be relaxed. After adding the source point to the queue, take the current node from the queue in turn, traverse all the target nodes pointed to by its outgoing edges, and judge the path according to the path relaxation principle. If the path exists, update the path value of the target node, and add the target node to the queue to continue processing if the target node is not in the queue. Record the number of times each node has been queued. If the number of times any node has been queued exceeds the total number of nodes, it means that there is a negative cycle in the graph, and terminate the algorithm. Under the condition that there is no negative weight cycle in the graph, the calculation of the path of each node is gradually completed;

[0062] Execution process: First, collect the label information and spatial coordinate data of various events in the rescue mission, perform frequency statistics on the event labels, build the correlation strength between event label pairs and calculate the semantic similarity, and determine the label adjustment coefficient. , then extract the timestamp of each event and calculate the adjacent event nodes and Time difference By statistically analyzing the standard deviation and distribution characteristics of all time differences, the time weight coefficient is calculated using the ratio of the quintile range to the mean. , then construct the direction vector based on the node space coordinates and calculate the cosine value of the angle between adjacent nodes , calculate the skewness value of the historical path direction data distribution and normalize it to generate the direction consistency weight , then divide the task area into fixed grids, count the number of events per unit area in each node area, and calculate the node event density The suppression function is constructed by the ratio of the global average density and the density suppression coefficient is generated by Logistic function mapping. Finally, it is introduced into the SPFA path optimization algorithm to gradually calculate the minimum path cost value of each event node in the rescue mission to reach each spatial area, and construct the robot task response path graph to drive the robot to achieve event-oriented optimal control and rapid response path planning in complex dynamic scenes.

[0063] The specific steps to generate an event response optimization model are:

[0064] Based on the event state mapping model, the response delay, task completion rate, and energy consumption parameters corresponding to each event state are calculated. By weighted evaluation of the parameters in each response path, the adaptive path is selected and its path cost is compared with the response performance to generate the event response parameter evaluation value.

[0065] Based on the evaluation values ​​of event response parameters, different response paths are prioritized. Based on the cost and weight of each path, the cost flow of path selection is analyzed to optimize the path selection strategy. Combined with the path evaluation, an event response optimization model is generated.

[0066] Based on the event state mapping model, the AHP hierarchical analysis method is used to perform parameter quantitative evaluation of the event state. The first-level evaluation indicators are set as response delay, task completion rate, and energy consumption parameters, and the second-level indicators are set as the corresponding values ​​in each response path. When constructing the judgment matrix, the matrix dimension is set to 3×3, and the numerical scale 1 to 9 is used to represent the relative importance of the indicators. The input parameters are the delay and completion rate comparison score of 5, the completion rate and energy consumption comparison score of 3, and the delay and energy consumption comparison score of 7. After normalizing the columns of the judgment matrix using the normalization method, the mean value of each row is calculated as the indicator weight. When performing weighted evaluation, Set the path parameter structure to a list structure. Each item in the list contains three fields: delay, completion rate, and energy consumption. For each item, use the weight vector to multiply the corresponding parameter value, and calculate the weighted sum as the path score. When selecting the adaptive path, traverse all path score values ​​and use the min function to select the path with the minimum score. Define the path cost as the sum of the delay and energy consumption in the path, and the response performance as the task completion rate. Use a parallel comparison method to compare the path cost and response performance. Construct a comparison structure as a dictionary structure, with the key being the path number and the value being a substructure containing the cost and performance fields, to generate the event response parameter evaluation value.

[0067] Based on the event response parameter evaluation value, the TOPSIS ranking method is used to prioritize the score values ​​of all response paths. First, a decision matrix is ​​constructed. The matrix rows correspond to each path, and the matrix columns correspond to the three scores of delay, completion rate, and energy consumption. The size of the decision matrix is ​​set to m×3. The normalized matrix processing adopts the vector normalization method. The square root of the sum of squares of each column is taken as the denominator. Each element is divided by the denominator to obtain the normalized value. The weight vector is set to the three weight values ​​obtained from the AHP results, corresponding to delay, completion rate, and energy consumption respectively. After weighting the normalized matrix, a weighted decision matrix is ​​obtained. The positive ideal solution is set as the column extreme value set corresponding to the maximum completion rate, minimum delay, and minimum energy consumption. The negative ideal solution is set as the column extreme value set corresponding to the minimum completion rate, maximum delay, and maximum energy consumption. The extreme value set calculates the Euclidean distance between each path and the positive and negative ideal solutions, expressed as D_plus and D_minus, respectively. The distance calculation formula is the weighted decision value minus the square of the ideal solution value, the sum of the results, and the square root. The ranking index is D_minus divided by D_plus plus D_minus. When analyzing the path selection cost flow, a partition structure is used to group the path costs, and the grouping interval is set to every 10 units. The path number and cost value within each group are statistically analyzed. When optimizing the path selection strategy, the screening condition is set to the path set with a ranking index greater than 0.75 and a path cost less than the cost mean. Combined with the path evaluation results, a path selection strategy structure is constructed. The fields include path number, ranking index, and cost value, and the event response optimization model is generated.

[0068] The specific steps to generate an incident response path plan are:

[0069] Based on the event response optimization model, the trigger probability and response path cost of each event state are obtained. The Dijkstra algorithm is used to calculate the adaptability of the path. By comparing the responsiveness and cost flow of the path, the execution effect of each path is evaluated. The cost flow of each path is determined to determine whether it meets the task requirements, and the path selection evaluation results are generated.

[0070] Based on the path selection evaluation results, the adaptive response path is screened by comparing the path cost flow with the trigger probability, analyzing whether the path conforms to the current environmental changes and task objectives, adjusting the optimized path, and generating the adaptive response path;

[0071] Based on the adaptive response path, each path is compared with the environmental state mapping information to evaluate the feasibility of the path, check the environmental changes and task requirements that may be encountered during the execution of the path, and generate an event response path plan;

[0072] Based on the event response optimization model, when obtaining the trigger probability of each event state, the cumulative frequency method is used to count the historical trigger records, the total number of events is set to N, the number of times a single event state is triggered is n, the trigger probability is calculated by dividing n by N, and the Dijkstra algorithm is used to calculate the adaptability of each response path. The path graph structure is initialized as a weighted directed graph, and the nodes in the graph are set as the response task node set. The edge weight is the path cost. The priority queue method is used to manage unprocessed nodes. The initial node cost is set to 0, and the costs of all other nodes are infinite. When constructing the priority queue, each element is set to a tuple containing the node number and the current cumulative cost. The minimum heap method is used to manage the queue priority. The loop condition is When the queue is not empty, after dequeuing the current minimum cost node, traverse all its adjacent edges. If the current cost of the adjacent node is greater than the current node cost plus the edge weight, update the adjacent node cost and add the adjacent node to the queue. After completion, a list of minimum path costs for each node is formed. When comparing the responsiveness of each path with the cost flow, set the responsiveness as the task completion rate field, and the path cost as the cumulative value of the costs of each node on the path. Construct a path evaluation structure with fields including path number, responsiveness value, and path cost value. When evaluating the execution effect of the path, a logical judgment method is used. The task requirement constraint is set to the maximum cost not exceeding a certain threshold. Judge whether the path cost is less than the threshold and record the judgment result to generate the path selection evaluation result.

[0073] Based on the path selection evaluation results, when comparing the path cost flow with the trigger probability, the path number is used as the matching field for structure joint query. When screening the adaptive response path, the screening condition is set as the path trigger probability is greater than the given probability threshold and the path cost is less than the cost threshold. The probability threshold is set to 0.6, and the cost threshold is the average of the historical path costs. The path evaluation structure is screened by the filtering function, and the path number set that meets the conditions is output. When analyzing whether the path conforms to the current environment changes and task goals, the environment status structure is set to a structure containing the current resource status and task priority fields. The resource status field of the node on which the path depends is compared with the current resource status, and a field-by-field matching method is adopted. If the path dependency field contains a resource field that does not exist in the current environment, the path is marked as unsuitable, otherwise it is marked as suitable. When adjusting the optimized path, a reordering mechanism is used to sort the adaptive path set in ascending order by the cost field, retain the top k paths as the adjustment results, and generate the adaptive response path.

[0074] Based on the adaptive response path, when comparing each path with the environmental state mapping information, the environmental state mapping information is set as a key-value structure containing the node number and the current state value. The path node list is read, and the node number is compared with the state value in the state mapping one by one. If the state value conflicts with the task requirement field, it is marked as a conflict path. When evaluating the feasibility of path execution, the feasibility judgment condition is set to that no node in the path is in a resource conflict state and the path length is less than the maximum task step threshold. When checking the environmental changes that may be encountered during the execution of the path, a change prediction structure is established. The fields include future time points and corresponding node state change information. The step simulation time interval is set to 5 seconds. The future state change values ​​of each step node in the path are compared one by one. If the state becomes unavailable within the prediction window, the path is marked as unstable. After eliminating all unstable paths, a structure is constructed. The fields include path number, feasibility judgment result, and conflict marking status to generate an event response path plan.

[0075] Dijkstra's algorithm, according to the formula

[0076]

[0077] in: Indicates the starting node to the node The minimum path evaluation cost value, Indicates the starting node to the node The known minimum path evaluation cost value, A set of event nodes representing completed path cost updates, Representation node To Node The basic path cost value, Represents a path segment arrive The risk correction factor, Representation node To Node The probability of successful event triggering, Indicates the adjustment coefficient of the trigger probability, Represents a path segment arrive responsiveness score, Represents the weight adjustment coefficient of responsiveness, represents the structural accessibility coefficient;

[0078] Dijkstra's algorithm first selects the source node in the graph, initializes the path distance to 0, and initializes the path distances of all other nodes to positive infinity. It then constructs a set of unvisited nodes and records the nodes to be processed. In each round, a node is selected from the set as the current processing node based on the distance value and marked as visited. It then traverses all adjacent nodes. If the distance from a node to an adjacent node is less than the currently recorded distance of the adjacent node, the distance value of the adjacent node is updated, and the predecessor node of the adapted path is recorded until all nodes are visited and the adapted path to the target node is confirmed.

[0079] Execution process: First, construct the event state graph structure of the rescue scene and extract all reachable path node pairs. and , and calculate the basic cost value of each path segment based on the map data , combined with the robot's moving distance, energy consumption and terrain traversal cost, and then collect risk factors of the area passed by the path, including building damage rate, visibility interference and personnel density level, to calculate the path risk correction factor , and then count the events from the node The probability of triggering success , obtained based on the normalization of the state transition frequency in the historical task execution record, and the inverse normalization of the variance of the probability in all paths to obtain the probability adjustment coefficient , and then evaluate the responsiveness of the path Based on the comprehensive scores of task response time, path accessibility and resource utilization, the response adjustment coefficient is further determined by mapping the task response coverage rate and the average response delay ratio. , then analyze the connectivity of the graph structure, identify whether the path is a cut point channel, redundant alternative and bridge path, and calculate the structural accessibility coefficient , and finally substitute all the parameters into the formula, from the set Select the node with the minimum path cost , update adjacent nodes Path evaluation value , until all path evaluations are completed, the optimal control path that meets the task response requirements is generated to guide the rescue robot to perform event-driven response and path scheduling control in complex scenarios.

[0080] The specific steps for generating an event conflict resolution path are as follows:

[0081] Based on the event response optimization model, the trigger probability of each path and the cost data of the response path are obtained, the cost value and trigger conditions between the paths are extracted, the relationship between the cost and the trigger probability is compared, the adaptation path that meets the trigger conditions is screened out, and the comparison result of the path cost and trigger probability is generated;

[0082] Based on the comparison results of path cost and trigger probability, the response time and execution order information of each path are extracted, the adaptability of each path is evaluated, the path selection is optimized, the path priority is adjusted based on the state mapping information, the path order is optimized and adjusted, and the event response path plan is generated;

[0083] Based on the event response optimization model, when obtaining the trigger probability and response path cost data of each path, the probability cumulative counting method is used to build a path probability mapping table, set each path number as the key, the number of triggers as the value, set the total number of trigger samples as N, and calculate the probability value by dividing the number of triggers by N. The response path cost data is read by traversing the path cost field, extracting the field value and storing it in the cost vector structure, and using the Pearson correlation coefficient calculation method to extract the relationship between the cost value and the trigger condition between the paths, constructing two vectors X and Y, where X is the path cost value vector and Y is the path trigger probability vector. The vector dot product operation is used to calculate the covariance, and the calculation formula is each item multiplied by the mean difference. The products are summed and divided by the vector dimension. The standard deviation is calculated by taking the square root of the sum of the squared differences between each item and the mean. The Pearson correlation coefficient is calculated as the covariance divided by the product of the standard deviations of the two vectors. When comparing the relationship between path cost and trigger probability, a threshold of 0.5 is set. The path set with a correlation coefficient greater than the threshold is selected as the initial adaptation path set. The trigger condition field is extracted as the trigger type and environmental parameter matching field set of each path. The set comparison method is used to determine whether the path contains the target trigger field and environmental parameter value. The set of path numbers that meet the conditions is selected and a matching structure is output. The fields include path number, trigger probability, and path cost. The comparison result of path cost and trigger probability is generated.

[0084] Based on the comparison results of path cost and trigger probability, the AHP hierarchical analysis method is used to extract the response time and execution sequence information of each path. The first-level indicators are set as the path response time and execution sequence value. The response time of each path is set as the cumulative result of the timestamp difference of the task nodes in the path. The path execution sequence information is read as the sequence number of the original path sequence field. The judgment matrix dimension is 2×2. The scoring scale 1 to 9 is used to represent the importance of response time to sequence. The response time score to sequence is set to 5, and the sequence score to response time is set to 1 / 5. After normalization, the average value is calculated by row to obtain the weights of the two indicators. The weighted method is used to evaluate the adaptability of each path. And scoring method, the structure fields include path number, response time value, sequence number, and adaptability score value. When optimizing path selection, the screening condition is set to the adaptability score value being greater than the average score of all paths, and the set of path numbers with score values ​​greater than the average is extracted as the optimized target path set. When adjusting the path priority in combination with the state mapping information, the mapping structure is set as the key to the path number and the value to the set of all node state identifiers on the path. The current environment state set is read, and a Boolean judgment is performed on whether the required state of the path exists in the environment state set. The path numbers that fully match are given priority, and the scores are sorted from high to low. The sequence fields are adjusted and renumbered to generate an event response path plan.

[0085] The specific steps to generate a dynamic task path planning solution are:

[0086] Based on the event conflict resolution path, it obtains environmental change information and records the timing of event triggering. By monitoring real-time environmental data, it analyzes the state changes of each event and environmental feedback, adjusts control parameters, and generates environmental and event feedback data.

[0087] Based on the environment and event feedback data, the task path corresponding to each event is recalculated. By adjusting the node order and task priority in the path, the path is replanned and the execution order is optimized to generate a dynamic task path planning solution.

[0088] Based on the event conflict resolution path, the sliding window state sampling method is used to obtain environmental change information and record the timing of event triggering. The time window size is set to 10 seconds, the sampling frequency is set to once per second, and the sampling variables include ambient temperature, humidity, wind speed, obstacle displacement value, and mechanical response feedback value. An environmental state data structure is established, and the fields include timestamp, variable name, and variable value. When recording the event triggering timing, an event monitoring mechanism is used, and the trigger monitoring field is set to the event ID and the corresponding system response status code. The timestamp and event status update field are bound to construct an event record list. The multi-state discrete sequence alignment algorithm is used to analyze the relationship between event state changes and environmental feedback. The main sequence is set to event The state change trajectory of the event state is the auxiliary sequence, and the change trajectory of the environmental variables is the auxiliary sequence. The two sequences are aligned by timestamp, and the interpolation method is used to synchronously map the asynchronous time points to the unified timeline. The interpolation function is set to a linear interpolation function with an interpolation interval of 1 second. The state feedback comparison is performed using a combination window of event state and environmental variables. The width of each combination window is 5 seconds and the step size is 1 second. The state change rate in the window and the difference between the environmental parameters are numerically fitted, and the least squares fitting method is used to estimate the causal relationship weight. When adjusting the control parameters, the weight value is range-calibrated. The control parameter structure field includes the target node number, control value name, and numerical adjustment amplitude to generate environment and event feedback data.

[0089] Based on the environment and event feedback data, a genetic algorithm is used to recalculate the task path corresponding to each event. The initial population size is 50 paths. Each path is represented as a sequence of node number permutations and combinations. The chromosome structure is a one-dimensional array with an array length equal to the number of task nodes. The fitness function is set as the inverse of the sum of the total execution time of the path and the path conflict weight. The path execution time is calculated by accumulating the estimated execution time field of the task node. The path conflict weight is obtained by weighted accumulation of the conflict factors in the task node and the feedback data. The selection operation adopts the roulette wheel method, the crossover operation adopts the partial mapping crossover method, and the crossover rate is set to 0.8. The mutation operation adopts the position exchange method, and the mutation rate is set to 0. .05, after executing 50 rounds of iteration, the path with the highest fitness is extracted, the order of nodes in the path is adjusted by the exchange sorting algorithm, and the position is fine-tuned after sorting by the weight field. The task priority is adjusted by the priority remapping strategy, and the priority weight is set according to the task importance field in the feedback data. The path sequence order is readjusted according to the weight value. When replanning the path, the starting node and the ending node remain unchanged, and the intermediate node sequence is updated and sorted. The execution order is optimized by reconstructing the node dependency relationship after sorting, and the DAG directed acyclic graph is used to update the task before and after dependency relationship. The position of the dependency conflict node is postponed to generate a dynamic task path planning solution.

[0090] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A rescue robot event-driven control method, characterized in that: The following steps are involved: S1: Identify various events in the rescue environment, obtain the triggering conditions and time series data of each event, use the SPFA algorithm to classify event types, spatially partition the environment, record the event triggering probability corresponding to each partition, combine the event response strategy and weight to perform state mapping, and generate an event state mapping model; S2: Based on the event state mapping model, calculate the response delay, task completion rate and energy consumption parameters of the event state, evaluate each response path according to the weight, optimize the path selection, and generate an event response optimization model; S3: Based on the event response optimization model, the Dijkstra algorithm is used to compare the trigger probability of the event state with the response path cost to perform path selection and judgment. The adaptive response path is selected by comparing the path cost flow, and the event response path solution is generated by combining the state mapping information. S4: Based on the event response path plan, determine whether the current event conflicts with an event that has already occurred. If a conflict occurs, sort the conflicting events by priority, roll back the response path with a lower priority, adjust and optimize the response sequence, and generate an event conflict resolution path. S5: Based on the event conflict resolution path, obtain environmental change information and event triggering timing, adjust the control strategy in combination with real-time feedback, recalculate the adaptive task path, and generate a dynamic task path planning solution through path adjustment and task replanning.

2. The event-driven control method for a rescue robot according to claim 1, characterized in that: The specific steps of generating the event state mapping model are: Identify various events in the rescue environment, match event names and numbers, mark trigger locations, read trigger time nodes, and extract status parameters. After data formatting, field column cleaning, numerical anomalies removal, and linear sorting of time series, an event trigger data set is generated. Based on the event triggering dataset, the SPFA algorithm is used to extract the event type field, perform statistical coding on the tag value, obtain the corresponding spatial coordinates, set the boundaries of the spatial partitions, mark the partition numbers, construct a distribution matrix and the correspondence between event type and spatial location, and generate a spatial area event distribution map; Based on the spatial area event distribution map, the number of event records in each partition is counted, the average trigger frequency is extracted, and a comparison is made according to the historical response action list. The weight value is verified, and a mapping relationship between the response status value and the weight is established. At the same time, the partition number is bound to the mapping relationship to generate an event status mapping model.

3. The event-driven control method for a rescue robot according to claim 2, characterized in that: The SPFA algorithm, according to the formula in: Represents the spatial region from the source event node to the target node The shortest path distance, Indicates the path from the source event node to the current node The shortest path distance, Representation node With node The event label weight value between represents the event tag semantic similarity adjustment coefficient, Representation node and The timestamp difference of the corresponding events, The weight coefficient representing the influence of time difference, Representation node With node The cosine value of the spatial angle between The weight coefficient representing the consistency of spatial propagation direction, Representation node The event density coefficient of the spatial region, represents the event density suppression coefficient.

4. The event-driven control method for a rescue robot according to claim 2, characterized in that: The SPFA algorithm first takes any starting node in the graph to be processed as the source point, initializes the shortest path from the source point to itself to 0, sets the path values ​​of the remaining nodes to positive infinity, and initializes a queue to store the nodes to be relaxed. After the source point is added to the queue, the current node is taken out of the queue in turn, and the target nodes pointed to by all its outgoing edges are traversed. The path is judged according to the path relaxation principle. If the path exists, the path value of the target node is updated. If the target node is not in the queue, the queue is added to continue processing, and the number of times each node is queued is recorded. If the number of times any node is queued exceeds the total number of nodes, it means that there is a negative cycle in the graph, and the algorithm is terminated. Under the condition that there is no negative weight cycle in the graph, the calculation of the path of each node is gradually completed.

5. The event-driven control method for a rescue robot according to claim 1, characterized in that: The specific steps for generating the event response optimization model are: Based on the event state mapping model, the response delay, task completion rate, and energy consumption parameters corresponding to each event state are calculated. By weighted evaluation of the parameters in each response path, an adaptive path is selected, and its path cost is compared with the response performance to generate an event response parameter evaluation value. Based on the event response parameter evaluation value, different response paths are prioritized, and based on the cost and weight of each path, the cost flow of path selection is analyzed, the path selection strategy is optimized, and combined with the path evaluation, an event response optimization model is generated.

6. The event-driven control method for a rescue robot according to claim 1, characterized in that: The specific steps for generating the event response path plan are: Based on the event response optimization model, the trigger probability and response path cost of each event state are obtained, and the Dijkstra algorithm is used to calculate the adaptability of the path. By comparing the responsiveness and cost flow of the path, the execution effect of each path is evaluated, and whether the cost flow of each path meets the task requirements is determined, and a path selection evaluation result is generated; Based on the path selection evaluation results, the adaptive response path is screened out by comparing the cost flow of the path with the trigger probability, analyzing whether the path conforms to the current environmental changes and task objectives, adjusting the optimized path, and generating the adaptive response path; Based on the adaptive response path, each path is compared with the environmental state mapping information, the execution feasibility of the path is evaluated, the environmental changes and task requirements that may be encountered during the execution of the path are checked, and an event response path plan is generated.

7. The event-driven control method for a rescue robot according to claim 6, characterized in that: The Dijkstra algorithm, according to the formula in: Indicates the starting node to the node The minimum path evaluation cost value, Indicates the starting node to the node The known minimum path evaluation cost value, A set of event nodes representing completed path cost updates, Representation node To Node The basic path cost value, Represents a path segment arrive The risk correction factor, Representation node To Node The probability of successful event triggering, Indicates the adjustment coefficient of the trigger probability, Represents a path segment arrive responsiveness score, Represents the weight adjustment coefficient of responsiveness, represents the structural accessibility coefficient.

8. The event-driven control method for a rescue robot according to claim 6, characterized in that: The Dijkstra algorithm first selects the source node in the graph, initializes the path distance to 0, initializes the path distances of all other nodes to positive infinity, and constructs a set of unvisited nodes to record the nodes to be processed. In each round, a node is selected from the set as the current processing node based on the distance value and marked as visited. Then, all adjacent nodes are traversed. If the distance from the node to the adjacent node is less than the currently recorded distance of the adjacent node, the distance value of the adjacent node is updated, and the predecessor node of the adapted path is recorded until all nodes are visited and the adapted path of the target node is confirmed.

9. The event-driven control method for a rescue robot according to claim 1, characterized in that: The specific steps for generating the event conflict resolution path are: Based on the event response optimization model, the trigger probability of each path and the cost data of the response path are obtained, the cost value and trigger conditions between the paths are extracted, the relationship between the cost and the trigger probability is compared, the adaptation path that meets the trigger conditions is screened out, and a comparison result of the path cost and the trigger probability is generated; Based on the comparison results of the path cost and trigger probability, the response time and execution order information of each path are extracted, the adaptability of each path is evaluated, the path selection is optimized, the path priority is adjusted in combination with the state mapping information, the path order is optimized and adjusted, and the event response path plan is generated.

10. The event-driven control method for a rescue robot according to claim 1, characterized in that: The specific steps of generating the dynamic task path planning scheme are: Based on the event conflict resolution path, environmental change information is obtained and the timing of event triggering is recorded. By monitoring real-time environmental data, the state changes of each event and environmental feedback are analyzed, control parameters are adjusted, and environmental and event feedback data are generated; Based on the environment and event feedback data, the task path corresponding to each event is recalculated, and by adjusting the node order and task priority in the path, the path is replanned and the execution order is optimized to generate a dynamic task path planning solution.

Images