A rescue route planning method and system for a flood drainage robot

Abstract

The present application belongs to the technical field of flood drainage rescue, and discloses a rescue route planning method and system of a flood drainage robot; comprising the following modules: a data acquisition module, used for collecting environmental data of a water accumulation area in an underground space in real time. The present application can calculate and monitor the total completion time of each flood drainage robot executing its optimized route in real time through the setting of a water accumulation amount prediction model. Once a significant imbalance in task progress between robots is detected, an intelligent support mechanism is started. Instead of simply instructing a nearby robot to go and help, the mechanism includes a complete decision chain, optimally segments the remaining task sequence of the flood drainage robot in need of support, strips part of the task and reallocates it to the supportable flood drainage robot, then re-plans a conflict-free space-time path for both, and continuously monitors during the support process to dynamically determine the exit timing of the support.

Description

Technical Field

[0001] This invention relates to the field of flood control and rescue technology, and more specifically, to a method and system for planning rescue routes for a flood control robot. Background Technology

[0002] Urban underground spaces such as underground parking garages, subway tunnels, and integrated utility tunnels are highly susceptible to flooding during heavy rainfall, which seriously threatens the safety of people's lives and property, the safety of urban operations, and the integrity of infrastructure. Timely and efficient drainage and rescue have become major challenges in the field of emergency management. In recent years, intelligent equipment such as drainage robots and unmanned boats have begun to be used in underground drainage and rescue due to their advantage of being able to operate in high-risk environments.

[0003] In terms of path planning for flood drainage and rescue operations, existing methods mostly focus on static optimal path search. For example, some solutions plan the shortest path for a single robot, but do not fully consider the impact of multi-robot collaboration, dynamic task allocation, and dwell time at work points on the overall route efficiency. When facing multiple scattered flood points, simple "traveling salesman problem"-style optimization cannot adapt to differences in urgency and changes in robot status, easily leading to an unbalanced situation where some robots are overloaded while others are idle. Existing flood drainage robot systems generally lack effective dynamic collaboration and replanning capabilities. The rescue environment is highly uncertain, such as the expansion of flood area, the emergence of new flood points, and sudden blockage of channels. Once the current system plans a path, it is often difficult to make effective adjustments based on real-time information during task execution. When a robot encounters unexpected delays, there is a lack of mechanisms for other robots that have completed their tasks or have lighter workloads to provide intelligent support, causing the overall task completion time to be dragged down by individual robots. Summary of the Invention

[0004] To address the problems in the background art, this invention proposes a method and system for planning rescue routes for a flood drainage robot.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a rescue route planning system for a flood drainage robot, comprising the following modules: The data acquisition module is used to collect environmental data in real time from waterlogged areas within underground spaces. The urgency assessment module is used to generate urgency scores for all waterlogged areas based on environmental data, and to classify each waterlogged area according to the urgency scores. The flood control and rescue route planning module is used to plan the initial flood control and rescue routes based on all high-emergency and medium-emergency flood areas. The matching degree prediction module is used to acquire historical matching degree data and build a matching degree prediction model. The matching degree prediction model is used to predict the matching degree between each low-emergency water accumulation area and each initial drainage and rescue route. The drainage and rescue route optimization module is used to add each low-emergency waterlogged area to the corresponding initial drainage and rescue route based on the predicted matching degree to obtain the optimized drainage and rescue route. The water accumulation prediction module is used to acquire historical water accumulation data and build a water accumulation prediction model, and then use the water accumulation prediction model to predict the water accumulation in each water accumulation area. The support scheduling module is used to schedule drainage robots for support. It calculates the drainage completion time based on the water volume of each waterlogged area, obtains the predicted total completion time for each optimized drainage rescue route based on the drainage completion time, obtains the average total completion time based on the total completion time, and schedules available drainage robots to support the drainage robots that need support based on the average total completion time.

[0006] Furthermore, the waterlogged area refers to an area within the underground space that has accumulated water and requires drainage and rescue operations. The environmental data includes water accumulation data and data on the property of trapped personnel in the waterlogged area. The water accumulation data includes water depth, water area, and water rise rate. The data on the property of trapped personnel includes the density of trapped personnel in the waterlogged area and the risk level of critical assets.

[0007] Furthermore, the process of generating an emergency score for all flooded areas based on environmental data includes: The urgency score S for the flooded area is:

[0008] In the formula, , , , and These are weighting coefficients, obtained through training based on historical data.

[0009] Furthermore, the process of classifying each flooded area based on its urgency score includes: Based on historical emergency rating data, set appropriate first and second emergency rating thresholds. The historical emergency rating data refers to the data set of emergency ratings for previous flooded areas. Compare the emergency rating of each flooded area with the two emergency rating thresholds to mark the flooded areas as flooded areas of different levels. When the urgency score is less than the first urgency score threshold, the flooded area is marked as a low-urgency flooded area. When the first urgency score threshold is less than or equal to the urgency score, which is less than or equal to the second urgency score threshold, the flooded area is marked as a severe urgency flooded area. When the urgency score is greater than the second urgency score threshold, the flooded area is marked as a high-urgency flooded area.

[0010] Furthermore, the process of planning the initial drainage and rescue routes based on all high-emergency and medium-emergency flooding areas includes: All high-emergency flood areas are globally sorted according to their urgency scores from highest to lowest. x drainage robots are initially deployed at the x high-emergency flood areas closest to the underground space entrance. Each drainage robot starts from its current position and selects the high-emergency flood area that is not yet covered by any drainage robot path from the set of high-emergency flood areas that minimizes the weighted sum of the broken-line distance from the current position to the high-emergency flood area and the reciprocal of the urgency score of the high-emergency flood area as the next target to visit. The weight parameters in the weighted sum calculation formula are obtained by training with historical data to achieve a balance between proximity and urgency priority. After the robot visits the target high-emergency flood area, it marks it as covered and updates its current position to the high-emergency flood area. This expansion process is repeated until all high-emergency flood areas are assigned to the path of a drainage robot. After completing the allocation of high-emergency flood areas, all medium-emergency flood areas are processed in a similar manner. At this point, each medium-emergency flood area is regarded as a task to be inserted. The path length increment generated by inserting it into any two consecutive nodes in the existing path of each drainage robot is calculated. The drainage robot and insertion position with the smallest increment are selected. At the same time, an urgency penalty factor is introduced to ensure that the insertion operation will not excessively delay the access to subsequent high-emergency flood areas on the path. In each medium-emergency flood area addition or insertion operation, the shortest safe path that can actually pass between two points is calculated in real time based on the pre-constructed three-dimensional topology map of underground space, and the complete route sequence of the drainage robot is dynamically updated. Output an initial flood relief route for each drainage robot, starting near the entrance and sequentially traversing all high-emergency and medium-emergency flooded areas assigned to it, with each step based on the actual connecting roads.

[0011] Furthermore, the process of acquiring historical matching data and constructing a matching prediction model, and then using this model to predict the matching degree between each low-emergency flooding area and each initial drainage and rescue route, includes: The matching degree refers to the degree of matching between each low-emergency flooding area and each initial drainage and rescue route; Factors affecting matching include: walking distance, low urgency score, route coherence index, communication signal strength, and cooperation potential value; The walking distance refers to the linear distance from the initial flood relief route of the drainage robot to the low-emergency flood area; The low urgency rating refers to the urgency rating of the low urgency flooded area; The route coherence index refers to the smoothness of the initial drainage and rescue route after adding the low-emergency flood area, which is the original length of the initial drainage and rescue route divided by the length of the route after adding it. Communication signal strength refers to the expected communication quality after the drainage robot arrives at the low-emergency flood area; Synergy potential refers to the degree to which a low-emergency flooding area is close to other initial drainage and relief routes; Obtain historical matching data for a single low-emergency flood area. The historical matching data includes the walking distance, low emergency score, route coherence index, communication signal strength, cooperation potential value, and historical matching degree of the single low-emergency flood area. Based on the walking distance, low urgency score, route coherence index, communication signal strength, collaborative potential value and corresponding historical matching degree of the corresponding low-emergency water accumulation area in different historical matching degree data, a matching degree prediction set is generated and divided into the first training set and the first test set. Construct a first convolutional neural network, using walking distance, low urgency score, route coherence index, communication signal strength, and collaborative potential value from different historical matching degree data in the first training set as input data for the first convolutional neural network, and using the corresponding historical matching degree in the first training set as output data for the first convolutional neural network. The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network that outputs a first test error threshold less than or equal to the preset first test error threshold is used as the matching degree prediction model. The walking distance, low urgency score, route coherence index, communication signal strength, and cooperation potential value of each low-emergency flood area are input into the matching degree prediction model to obtain the predicted matching degree of each low-emergency flood area.

[0012] Furthermore, the process of optimizing the drainage and rescue routes by adding each low-emergency flooding area to the corresponding initial drainage and rescue route based on the predicted matching degree includes: All low-emergency waterlogging areas are sorted across the entire region according to their predicted matching degree. A matching degree threshold is set, and low-emergency waterlogging areas with predicted matching degree greater than the matching degree threshold are processed first to ensure that low-emergency waterlogging areas with high matching degree are given priority for reasonable allocation. The selection of the insertion location for low-emergency flooding areas adopts the principle of minimum interference. The optimal insertion point is found on the route topology map that minimizes the increase in the total path and does not disrupt the access order of high and medium-emergency flooding areas. At the same time, a time buffer mechanism is adopted to ensure that the insertion of low-emergency flooding areas will not delay the completion of subsequent emergency tasks as much as possible. For the remaining unassigned low-emergency flood areas, a second round of collaborative insertion optimization is performed, allowing two adjacent low-emergency flood areas to be packaged as a combined task and inserted into adjacent positions on the same route. This reduces the back-and-forth movement of the drainage robots. Global conflict detection and resolution are performed on all optimized routes to ensure that the paths of multiple drainage robots do not overlap in space and time. Finally, x optimized drainage and rescue routes are generated, which include all high and medium-emergency flood areas and reasonably allocated low-emergency flood areas.

[0013] Furthermore, the process of acquiring historical water accumulation data and constructing a water accumulation prediction model, and then using this model to predict the water accumulation in each waterlogged area, includes: The water volume refers to the amount of water in the waterlogged area. Factors affecting water accumulation include: water depth, water area, ground slope, seepage rate, external water inflow rate, and drainage outlet condition; The ground slope corresponding to the water accumulation area was obtained from the underground space ground database; The seepage rate corresponding to the water accumulation area was obtained from the underground space surface database; The external water inlet rate was measured by an ultrasonic flow meter at the inlet. The camera was used to capture the situation at the drain outlet. Obtain historical water accumulation data for a single waterlogged area. The historical water accumulation data includes the water depth, waterlogged area, ground slope, seepage rate, external water inflow rate, and drainage outlet status of the single waterlogged area, as well as the historical water accumulation volume of the single waterlogged area. Based on the water depth, water area, ground slope, seepage rate, external water inflow rate, and drainage outlet status of the corresponding water accumulation area in different historical water accumulation data, a water accumulation prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The water depth, water area, ground slope, seepage rate, external water inflow rate and drainage outlet status in the different historical water volume data in the second training set are used as the input data of the second convolutional neural network, and the corresponding historical water volume in the second training set is used as the output data of the second convolutional neural network. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network that outputs a second test error threshold less than or equal to the preset second test error threshold is used as the water accumulation prediction model. The water depth, water area, ground slope, seepage rate, external water inflow rate, and drainage outlet status of each waterlogged area are input into the water volume prediction model to obtain the predicted water volume for each waterlogged area.

[0014] Furthermore, the process of dispatching drainage robots for support includes: The predicted water volume for each waterlogged area is obtained based on the water volume prediction model. The drainage completion time for each waterlogged area is obtained based on the predicted water volume, natural drainage rate, and maximum drainage rate of the drainage robot. The total completion time for each drainage robot to complete the drainage rescue is obtained based on the drainage completion time of each waterlogged area, the average travel speed of the drainage robot, and the degree of optimization of the drainage rescue route. The average total completion time is obtained based on the total completion time of all optimized drainage rescue routes. Drainage robots with a total completion time greater than the average total completion time are set as drainage robots that need support, and drainage robots with a total completion time less than the average total completion time are set as drainage robots that can be supported. For robots requiring support for flood drainage, select the nearest available robot from the cluster of available flood drainage robots. Analyze the remaining optimized flood drainage and rescue routes of the robots requiring support and find an optimal task split point. Usually, the waterlogged areas at the back of the route, especially those with low urgency or long operation time, are cut off in whole or in part and redistributed to available flood drainage robots. The optimal task split point is when the time for the two flood drainage robots to complete their remaining tasks is rebalanced. The two flood drainage robots replan their conflict-free routes to the newly assigned task points and coordinate their operation time windows to avoid meeting in the same narrow space. For large waterlogged areas, a collaborative mode of simultaneous entry from opposite sides can be designed. During the support mission, the progress of both parties is continuously monitored. Once the risk of delay for the drainage robots that need support is eliminated or the original mission is completed, the drainage robots that can be supported are directed to withdraw in an orderly manner and continue to execute their own suspended original mission sequence.

[0015] A method for planning rescue routes for a flood drainage robot includes the following steps: S1: Real-time collection of environmental data from various waterlogged areas in underground space, normalization of the environmental data, generation of an emergency score for each waterlogged area based on the processed data, and classification of the waterlogged areas into three emergency levels: high, medium, and low, based on a preset emergency score threshold. S2: Sort all high-urgency areas by score, select the nearest and most urgent area as the starting point for each drainage robot, and allocate all high and medium-urgency areas in sequence to form the initial drainage and rescue route covering all high and medium-urgency waterlogged areas. S3: Construct a matching degree prediction model to predict the matching degree between each low-emergency waterlogging area and each initial drainage and rescue route. Based on the matching degree, insert the low-emergency waterlogging area into the optimal position of the initial drainage and rescue route to generate an optimized drainage and rescue route. S4: Construct a water accumulation prediction model to predict the water accumulation in each water accumulation area. Based on the maximum drainage rate of the drainage robot and the natural drainage rate of the water accumulation area, calculate the drainage completion time of each water accumulation area and the predicted total completion time of each optimized drainage and rescue route. S5: Calculate the average total completion time of all drainage robots, select the nearest available drainage robot for the drainage robot that needs support, divide the subsequent tasks of the drainage robot that needs support to the available drainage robots, and replan the non-conflicting paths between the two.

[0016] The technical effects and advantages of the rescue route planning method and system for a flood drainage robot of the present invention are as follows: (1) By setting up a water accumulation prediction model and a support scheduling module, the accuracy and reliability of water accumulation and subsequent drainage time prediction are significantly improved. Based on accurate time prediction, the total completion time of each drainage robot executing its optimized route can be calculated and monitored in real time. Once a significant imbalance in the task progress between robots is detected, the intelligent support mechanism is immediately activated. This mechanism is not simply an instruction for neighboring robots to go and help, but contains a complete decision chain. It optimally divides the remaining task sequence of the drainage robots that need support, strips out some tasks and redistributes them to the drainage robots that can support them, and then replans a conflict-free spatiotemporal path for both parties. It continuously monitors during the support process and dynamically determines the timing of exiting the support, thereby improving the success rate and completion speed of the entire drainage and rescue mission in the face of various uncertainties.

[0017] (2) By setting an urgency score, the waterlogged area is divided into high, medium and low urgency waterlogged areas, providing a clear and scientific action guide for rescue. For the multi-drainage robot system, a priority coverage of high and medium urgency waterlogged areas and intelligent matching of low urgency areas are designed. Based on the urgency, each robot plans an initial drainage rescue route covering all high and medium urgency areas. Based on the matching degree prediction model, low urgency areas are dynamically inserted into each robot route with the principle of minimum interference to form an optimized drainage rescue route. This not only ensures that the most critical area is responded to the fastest, but also makes full use of the drainage robot's travel gap to handle the secondary area through intelligent matching, avoiding the robot being idle or moving blindly. This significantly shortens the total processing time of all waterlogged points at the global level and achieves a qualitative leap in rescue efficiency. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] Reference Figure 1 A rescue route planning system for a flood drainage robot includes the following modules: The data acquisition module is used to collect environmental data in real time from waterlogged areas within underground spaces. The urgency assessment module is used to generate urgency scores for all waterlogged areas based on environmental data, and to classify each waterlogged area according to the urgency scores. The flood control and rescue route planning module is used to plan the initial flood control and rescue routes based on all high-emergency and medium-emergency flood areas. The matching degree prediction module is used to acquire historical matching degree data and build a matching degree prediction model. The matching degree prediction model is used to predict the matching degree between each low-emergency water accumulation area and each initial drainage and rescue route. The drainage and rescue route optimization module is used to add each low-emergency waterlogged area to the corresponding initial drainage and rescue route based on the predicted matching degree to obtain the optimized drainage and rescue route. The water accumulation prediction module is used to acquire historical water accumulation data and build a water accumulation prediction model, and then use the water accumulation prediction model to predict the water accumulation in each water accumulation area. The support scheduling module is used to schedule drainage robots for support. It calculates the drainage completion time based on the water volume of each waterlogged area, obtains the predicted total completion time for each optimized drainage rescue route based on the drainage completion time, obtains the average total completion time based on the total completion time, and schedules available drainage robots to support the drainage robots that need support based on the average total completion time.

[0021] It should be further explained that, in the specific implementation process, the waterlogged area refers to the area in the underground space where there is water accumulation and drainage and rescue are required. The environmental data includes water accumulation data of the waterlogged area and data on the property of trapped personnel. The water accumulation data includes water depth, water area and water rise rate. The data on the property of trapped personnel includes the density of trapped personnel in the waterlogged area and the risk level of key assets. The depth of the accumulated water was measured using an ultrasonic level sensor.

[0022] In the formula, To normalize the water depth, This represents the measured depth of the accumulated water. The maximum water depth is 3m. The area of ​​water accumulation was obtained through camera and image recognition.

[0023] In the formula, To normalize the water accumulation area, This represents the measured area of ​​accumulated water. This is the theoretical maximum water accumulation area, specifically 500. ; The rate of water rise is calculated using historical data from the level sensor, i.e., the water rise height divided by the rise time.

[0024] In the formula, To normalize the rate of water level rise, This represents the actual rate of water level rise. This represents the theoretical maximum rate of water rise, specifically 1 m / min. The density of trapped personnel was determined using infrared thermal imaging and AI identification.

[0025] In the formula, To normalize the density of trapped personnel, This represents the actual density of trapped personnel. The theoretical maximum density of trapped people is 10 people per 100 square meters; The risk level K of critical assets is obtained from the underground space asset registration database. If there are no critical assets, the risk level is 0; for general equipment, it is 0.3; for important equipment, it is 0.7; and for core facilities, it is 1.0.

[0026] It should be further explained that, in the specific implementation process, the process of generating an emergency score for all flooded areas based on environmental data includes: The urgency score S for the flooded area is:

[0027] In the formula, , , , and These are weighting coefficients, obtained from training on historical data, and set to 0.15, 0.15, 0.2, 0.35, and 0.15 respectively. If the water accumulation data for a certain waterlogged area is: It is 1.5m. 100 , It is 0.4 m / min. If the ratio is 5 people per 100 square meters and K is 0.7, then the emergency score S for this flooded area is 0.465.

[0028] It should be further explained that, in the specific implementation process, the process of classifying each flooded area according to the urgency score includes: Based on historical emergency rating data, set appropriate first and second emergency rating thresholds. The historical emergency rating data refers to the data set of emergency ratings for previous flooded areas. Compare the emergency rating of each flooded area with the two emergency rating thresholds to mark the flooded areas as flooded areas of different levels. When the urgency score is less than the first urgency score threshold, the flooded area is marked as a low-urgency flooded area. When the first urgency score threshold is less than or equal to the urgency score, which is less than or equal to the second urgency score threshold, the flooded area is marked as a severe urgency flooded area. When the urgency score is greater than the second urgency score threshold, the flooded area is marked as a high-urgency flooded area.

[0029] It should be further explained that, in the specific implementation process, the initial drainage and rescue route planning based on all high-emergency and medium-emergency flood areas includes: All high-emergency flood areas are globally sorted according to their urgency scores from highest to lowest. x drainage robots are initially deployed at the x high-emergency flood areas closest to the underground space entrance. Each drainage robot starts from its current position and selects the high-emergency flood area that is not yet covered by any drainage robot path from the set of high-emergency flood areas that minimizes the weighted sum of the broken-line distance from the current position to the high-emergency flood area and the reciprocal of the urgency score of the high-emergency flood area as the next target to visit. The weight parameters in the weighted sum calculation formula are obtained by training with historical data to achieve a balance between proximity and urgency priority. After the robot visits the target high-emergency flood area, it marks it as covered and updates its current position to the high-emergency flood area. This expansion process is repeated until all high-emergency flood areas are assigned to the path of a drainage robot. After completing the allocation of high-emergency flood areas, all medium-emergency flood areas are processed in a similar manner. At this point, each medium-emergency flood area is regarded as a task to be inserted. The path length increment generated by inserting it into any two consecutive nodes in the existing path of each drainage robot is calculated. The drainage robot and insertion position with the smallest increment are selected. At the same time, an urgency penalty factor is introduced to ensure that the insertion operation will not excessively delay the access to subsequent high-emergency flood areas on the path. In each medium-emergency flood area addition or insertion operation, the shortest safe path that can actually pass between two points is calculated in real time based on the pre-constructed three-dimensional topology map of underground space, and the complete route sequence of the drainage robot is dynamically updated. Output an initial flood relief route for each drainage robot, starting near the entrance and sequentially traversing all high-emergency and medium-emergency flooded areas assigned to it, with each step based on the actual connecting roads.

[0030] It should be further explained that, in the specific implementation process, the process of acquiring historical matching degree data and constructing a matching degree prediction model, and using the matching degree prediction model to predict the matching degree between each low-emergency flooding area and each initial drainage and rescue route, includes: The matching degree refers to the degree of matching between each low-emergency flooding area and each initial drainage and rescue route; Factors affecting matching include: walking distance, low urgency score, route coherence index, communication signal strength, and cooperation potential value; The walking distance refers to the linear distance from the initial flood relief route of the drainage robot to the low-emergency flood area; The low urgency rating refers to the urgency rating of the low urgency flooded area; The route coherence index refers to the smoothness of the initial drainage and rescue route after adding the low-emergency flood area, which is the original length of the initial drainage and rescue route divided by the length of the route after adding it. Communication signal strength refers to the expected communication quality after the drainage robot arrives at the low-emergency flood area; Synergy potential refers to the degree to which a low-emergency flooding area is close to other initial drainage and relief routes; Obtain historical matching data for a single low-emergency flood area. The historical matching data includes the walking distance, low emergency score, route coherence index, communication signal strength, cooperation potential value, and historical matching degree of the single low-emergency flood area. Based on the walking distance, low urgency score, route coherence index, communication signal strength, collaborative potential value and corresponding historical matching degree of the corresponding low-emergency water accumulation area in different historical matching degree data, a matching degree prediction set is generated and divided into the first training set and the first test set. Construct a first convolutional neural network, using walking distance, low urgency score, route coherence index, communication signal strength, and collaborative potential value from different historical matching degree data in the first training set as input data for the first convolutional neural network, and using the corresponding historical matching degree in the first training set as output data for the first convolutional neural network. The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network that outputs a first test error threshold less than or equal to the preset first test error threshold is used as the matching degree prediction model. The walking distance, low urgency score, route coherence index, communication signal strength and cooperation potential value of each low-emergency waterlogging area are input into the matching degree prediction model to obtain the predicted matching degree of each low-emergency waterlogging area. In an embodiment of the present invention, the predicted matching degree of all low-emergency waterlogging areas is obtained through a matching degree prediction model. The predicted matching degree is related to walking distance, low emergency score, route coherence index, communication signal strength and cooperation potential value. The length of the walking distance directly affects the prediction matching degree. The longer the walking distance, the longer the drainage robot needs to walk and the longer it takes, resulting in a lower prediction matching degree. Therefore, the walking distance and the matching degree are negatively correlated. The low urgency score directly affects the prediction match rate. The higher the low urgency score, the greater the urgency of the low urgency waterlogging area and the greater the prediction match rate. Therefore, the low urgency score and the match rate are positively correlated. The magnitude of the route coherence index directly affects the prediction matching degree. The larger the route coherence index, the smoother the route after adding the low emergency water accumulation area, and the greater the prediction matching degree. Therefore, the route coherence index and the matching degree are positively correlated. The strength of the communication signal directly affects the prediction matching degree. The stronger the communication signal, the better the communication signal in low-emergency water accumulation areas, which makes it easier to control the drainage robot and the greater the prediction matching degree. Therefore, the communication signal strength and the matching degree are positively correlated. The size of the collaborative potential value directly affects the magnitude of the prediction matching degree. The larger the collaborative potential value, the more likely multiple drainage robots can work together to handle the same low-emergency waterlogging area, and the greater the prediction matching degree. Therefore, the collaborative potential value and the matching degree are positively correlated.

[0031] It should be further explained that, in the specific implementation process, the process of adding each low-emergency flooding area to the corresponding initial drainage and rescue route based on the predicted matching degree to obtain the optimized drainage and rescue route includes: All low-emergency waterlogging areas are sorted across the entire region according to their predicted matching degree. A matching degree threshold is set, and low-emergency waterlogging areas with predicted matching degree greater than the matching degree threshold are processed first to ensure that low-emergency waterlogging areas with high matching degree are given priority for reasonable allocation. The selection of the insertion location for low-emergency flooding areas adopts the principle of minimum interference. The optimal insertion point is found on the route topology map that minimizes the increase in the total path and does not disrupt the access order of high and medium-emergency flooding areas. At the same time, a time buffer mechanism is adopted to ensure that the insertion of low-emergency flooding areas will not delay the completion of subsequent emergency tasks as much as possible. For the remaining unassigned low-emergency flood areas, a second round of collaborative insertion optimization is performed, allowing two adjacent low-emergency flood areas to be packaged as a combined task and inserted into adjacent positions on the same route. This reduces the back-and-forth movement of the drainage robots. Global conflict detection and resolution are performed on all optimized routes to ensure that the paths of multiple drainage robots do not overlap in space and time. Finally, x optimized drainage and rescue routes are generated, which include all high and medium-emergency flood areas and reasonably allocated low-emergency flood areas.

[0032] It should be further explained that, in the specific implementation process, the process of obtaining historical water accumulation data and constructing a water accumulation prediction model, and then using the water accumulation prediction model to predict the water accumulation in each water accumulation area, includes: The water volume refers to the amount of water in the waterlogged area. Factors affecting water accumulation include: water depth, water area, ground slope, seepage rate, external water inflow rate, and drainage outlet condition; The ground slope corresponding to the water accumulation area was obtained from the underground space ground database; The seepage rate corresponding to the water accumulation area was obtained from the underground space surface database; The external water inlet rate was measured by an ultrasonic flow meter at the inlet. The camera was used to capture the situation at the drain outlet. Obtain historical water accumulation data for a single waterlogged area. The historical water accumulation data includes the water depth, waterlogged area, ground slope, seepage rate, external water inflow rate, and drainage outlet status of the single waterlogged area, as well as the historical water accumulation volume of the single waterlogged area. Based on the water depth, water area, ground slope, seepage rate, external water inflow rate, and drainage outlet status of the corresponding water accumulation area in different historical water accumulation data, a water accumulation prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The water depth, water area, ground slope, seepage rate, external water inflow rate and drainage outlet status in the different historical water volume data in the second training set are used as the input data of the second convolutional neural network, and the corresponding historical water volume in the second training set is used as the output data of the second convolutional neural network. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network that outputs a second test error threshold less than or equal to the preset second test error threshold is used as the water accumulation prediction model. The water depth, water area, ground slope, seepage rate, external water inflow rate, and drainage outlet status of each waterlogged area are input into the water volume prediction model to obtain the predicted water volume for each waterlogged area. In an embodiment of the present invention, the predicted water volume of all waterlogged areas is obtained through a water volume prediction model. The predicted water volume is related to the water depth, water area, ground slope, seepage rate, external water inflow rate, and drainage outlet status. The depth of the water directly affects the predicted amount of water accumulation. The greater the water depth, the greater the predicted amount of water accumulation. Therefore, the water depth and the amount of water accumulation are positively correlated. The size of the water accumulation area directly affects the predicted water volume. The larger the water accumulation area, the larger the predicted water volume. Therefore, the water accumulation area and the water volume are positively correlated. The slope of the ground directly affects the predicted amount of water accumulation. The steeper the slope, the easier it is for water to flow to lower areas, and the smaller the predicted amount of water accumulation. Therefore, the slope of the ground and the amount of water accumulation are negatively correlated. The magnitude of the seepage rate directly affects the predicted amount of water accumulation. The greater the seepage rate, the faster the water naturally decreases, and the smaller the predicted amount of water accumulation. Therefore, the seepage rate and the amount of water accumulation are negatively correlated. The rate of external water inflow directly affects the predicted amount of water accumulation. The higher the rate of external water inflow, the faster the water accumulation grows and the larger the predicted amount of water accumulation. Therefore, the rate of external water inflow and the amount of water accumulation are positively correlated. The condition of the drain outlet directly affects the predicted amount of water accumulation. The better the condition of the drain outlet, the smoother the drainage, the more water is drained, and the smaller the predicted amount of water accumulation. Therefore, the condition of the drain outlet is negatively correlated with the amount of water accumulation.

[0033] It should be further explained that, in the specific implementation process, the process of dispatching drainage robots for support includes: The predicted water volume for each waterlogged area is obtained based on the water volume prediction model. The drainage completion time for each waterlogged area is obtained based on the predicted water volume, natural drainage rate, and maximum drainage rate of the drainage robot. The total completion time for each drainage robot to complete the drainage rescue is obtained based on the drainage completion time of each waterlogged area, the average travel speed of the drainage robot, and the degree of optimization of the drainage rescue route. The average total completion time is obtained based on the total completion time of all optimized drainage rescue routes. Drainage robots with a total completion time greater than the average total completion time are set as drainage robots that need support, and drainage robots with a total completion time less than the average total completion time are set as drainage robots that can be supported. For robots requiring support for flood drainage, select the nearest available robot from the cluster of available flood drainage robots. Analyze the remaining optimized flood drainage and rescue routes of the robots requiring support and find an optimal task split point. Usually, the waterlogged areas at the back of the route, especially those with low urgency or long operation time, are cut off in whole or in part and redistributed to available flood drainage robots. The optimal task split point is when the time for the two flood drainage robots to complete their remaining tasks is rebalanced. The two flood drainage robots replan their conflict-free routes to the newly assigned task points and coordinate their operation time windows to avoid meeting in the same narrow space. For large waterlogged areas, a collaborative mode of simultaneous entry from opposite sides can be designed. During the support mission, the progress of both parties is continuously monitored. Once the risk of delay for the drainage robots that need support is eliminated or the original mission is completed, the drainage robots that can be supported are directed to withdraw in an orderly manner and continue to execute their own suspended original mission sequence.

[0034] A method for planning rescue routes for a flood drainage robot includes the following steps: S1: Real-time collection of environmental data from various waterlogged areas in underground space, normalization of the environmental data, generation of an emergency score for each waterlogged area based on the processed data, and classification of the waterlogged areas into three emergency levels: high, medium, and low, based on a preset emergency score threshold. S2: Sort all high-urgency areas by score, select the nearest and most urgent area as the starting point for each drainage robot, and allocate all high and medium-urgency areas in sequence to form the initial drainage and rescue route covering all high and medium-urgency waterlogged areas. S3: Construct a matching degree prediction model to predict the matching degree between each low-emergency waterlogging area and each initial drainage and rescue route. Based on the matching degree, insert the low-emergency waterlogging area into the optimal position of the initial drainage and rescue route to generate an optimized drainage and rescue route. S4: Construct a water accumulation prediction model to predict the water accumulation in each water accumulation area. Based on the maximum drainage rate of the drainage robot and the natural drainage rate of the water accumulation area, calculate the drainage completion time of each water accumulation area and the predicted total completion time of each optimized drainage and rescue route. S5: Calculate the average total completion time of all drainage robots, select the nearest available drainage robot for the drainage robot that needs support, divide the subsequent tasks of the drainage robot that needs support to the available drainage robots, and replan the non-conflicting paths between the two.

[0035] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0036] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A rescue route planning system for a flood drainage robot, characterized in that, Includes the following modules: The data acquisition module is used to collect environmental data in real time from waterlogged areas within underground spaces. The urgency assessment module is used to generate urgency scores for all waterlogged areas based on environmental data, and to classify each waterlogged area according to the urgency scores. The flood control and rescue route planning module is used to plan the initial flood control and rescue routes based on all high-emergency and medium-emergency flood areas. The matching degree prediction module is used to acquire historical matching degree data and build a matching degree prediction model. The matching degree prediction model is used to predict the matching degree between each low-emergency water accumulation area and each initial drainage and rescue route. The drainage and rescue route optimization module is used to add each low-emergency waterlogged area to the corresponding initial drainage and rescue route based on the predicted matching degree to obtain the optimized drainage and rescue route. The water accumulation prediction module is used to acquire historical water accumulation data and build a water accumulation prediction model, and then use the water accumulation prediction model to predict the water accumulation in each water accumulation area. The support scheduling module is used to schedule drainage robots for support. It calculates the drainage completion time based on the water volume of each waterlogged area, obtains the predicted total completion time for each optimized drainage rescue route based on the drainage completion time, obtains the average total completion time based on the total completion time, and schedules available drainage robots to support the drainage robots that need support based on the average total completion time.

2. The rescue route planning system for the drainage robot according to claim 1, characterized in that, The flooded area refers to an area in the underground space where there is water accumulation and drainage and rescue operations are required. The environmental data includes water accumulation data and data on the property of trapped personnel in the flooded area. The water accumulation data includes water depth, water area, and water rise rate. The data on the property of trapped personnel includes the density of trapped personnel in the flooded area and the risk level of critical assets.

3. The rescue route planning system for the drainage robot according to claim 2, characterized in that, The process of generating an emergency score for all flooded areas based on environmental data includes: The urgency score S for the flooded area is: In the formula, , , , and These are weighting coefficients, obtained through training based on historical data.

4. The rescue route planning system for the drainage robot according to claim 3, characterized in that, The process of classifying each flooded area based on its urgency rating includes: Based on historical emergency rating data, set appropriate first and second emergency rating thresholds. The historical emergency rating data refers to the data set of emergency ratings for previous flooded areas. Compare the emergency rating of each flooded area with the two emergency rating thresholds to mark the flooded areas as flooded areas of different levels. When the urgency score is less than the first urgency score threshold, the flooded area is marked as a low-urgency flooded area. When the first urgency score threshold is less than or equal to the urgency score, which is less than or equal to the second urgency score threshold, the flooded area is marked as a severe urgency flooded area. When the urgency score is greater than the second urgency score threshold, the flooded area is marked as a high-urgency flooded area.

5. The rescue route planning system for the drainage robot according to claim 4, characterized in that, The process of planning the initial drainage and rescue routes based on all high-emergency and medium-emergency flood areas includes: All high-emergency flood areas are globally sorted according to their urgency scores from highest to lowest. x drainage robots are initially deployed at the x high-emergency flood areas closest to the underground space entrance. Each drainage robot starts from its current position and selects the high-emergency flood area that is not yet covered by any drainage robot path from the set of high-emergency flood areas that minimizes the weighted sum of the broken-line distance from the current position to the high-emergency flood area and the reciprocal of the urgency score of the high-emergency flood area as the next target to visit. The weight parameters in the weighted sum calculation formula are obtained by training with historical data to achieve a balance between proximity and urgency priority. After the robot visits the target high-emergency flood area, it marks it as covered and updates its current position to the high-emergency flood area. This expansion process is repeated until all high-emergency flood areas are assigned to the path of a drainage robot. After completing the allocation of high-emergency flood areas, all medium-emergency flood areas are processed in a similar manner. At this point, each medium-emergency flood area is regarded as a task to be inserted. The path length increment generated by inserting it into any two consecutive nodes in the existing path of each drainage robot is calculated. The drainage robot and insertion position with the smallest increment are selected. At the same time, an urgency penalty factor is introduced to ensure that the insertion operation will not excessively delay the access to subsequent high-emergency flood areas on the path. In each medium-emergency flood area addition or insertion operation, the shortest safe path that can actually pass between two points is calculated in real time based on the pre-constructed three-dimensional topology map of underground space, and the complete route sequence of the drainage robot is dynamically updated. Output an initial flood relief route for each drainage robot, starting near the entrance and sequentially traversing all high-emergency and medium-emergency flooded areas assigned to it, with each step based on the actual connecting roads.

6. The rescue route planning system for the drainage robot according to claim 5, characterized in that, The process of acquiring historical matching data and constructing a matching prediction model, and then using this model to predict the matching degree between each low-emergency flooding area and each initial drainage and rescue route, includes: The matching degree refers to the degree of matching between each low-emergency flooding area and each initial drainage and rescue route; Factors affecting matching include: walking distance, low urgency score, route coherence index, communication signal strength, and cooperation potential value; Obtain historical matching data for a single low-emergency flood area. The historical matching data includes the walking distance, low emergency score, route coherence index, communication signal strength, cooperation potential value, and historical matching degree of the single low-emergency flood area. Based on the walking distance, low urgency score, route coherence index, communication signal strength, collaborative potential value and corresponding historical matching degree of the corresponding low-emergency water accumulation area in different historical matching degree data, a matching degree prediction set is generated and divided into the first training set and the first test set. Construct a first convolutional neural network, using walking distance, low urgency score, route coherence index, communication signal strength, and collaborative potential value from different historical matching degree data in the first training set as input data for the first convolutional neural network, and using the corresponding historical matching degree in the first training set as output data for the first convolutional neural network. The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network that outputs a first test error threshold less than or equal to the preset first test error threshold is used as the matching degree prediction model. The walking distance, low urgency score, route coherence index, communication signal strength, and cooperation potential value of each low-emergency flood area are input into the matching degree prediction model to obtain the predicted matching degree of each low-emergency flood area.

7. The rescue route planning system for the drainage robot according to claim 6, characterized in that, The process of optimizing drainage and rescue routes by adding each low-emergency flooding area to the corresponding initial drainage and rescue route based on the predicted matching degree includes: All low-emergency waterlogging areas are sorted across the entire region according to their predicted matching degree. A matching degree threshold is set, and low-emergency waterlogging areas with predicted matching degree greater than the matching degree threshold are processed first to ensure that low-emergency waterlogging areas with high matching degree are given priority for reasonable allocation. The selection of the insertion location for low-emergency flooding areas adopts the principle of minimum interference. The optimal insertion point is found on the route topology map that minimizes the increase in the total path and does not disrupt the access order of high and medium-emergency flooding areas. At the same time, a time buffer mechanism is adopted to ensure that the insertion of low-emergency flooding areas will not delay the completion of subsequent emergency tasks as much as possible. For the remaining unassigned low-emergency flood areas, a second round of collaborative insertion optimization is performed, allowing two adjacent low-emergency flood areas to be packaged as a combined task and inserted into adjacent positions on the same route. This reduces the back-and-forth movement of the drainage robots. Global conflict detection and resolution are performed on all optimized routes to ensure that the paths of multiple drainage robots do not overlap in space and time. Finally, x optimized drainage and rescue routes are generated, which include all high and medium-emergency flood areas and reasonably allocated low-emergency flood areas.

8. The rescue route planning system for the drainage robot according to claim 7, characterized in that, The process of acquiring historical water accumulation data and building a water accumulation prediction model, and then using the water accumulation prediction model to predict the water accumulation in each water accumulation area, includes: The water volume refers to the amount of water in the waterlogged area. Factors affecting water accumulation include: water depth, water area, ground slope, seepage rate, external water inflow rate, and drainage outlet condition; Obtain historical water accumulation data for a single waterlogged area. The historical water accumulation data includes the water depth, waterlogged area, ground slope, seepage rate, external water inflow rate, and drainage outlet status of the single waterlogged area, as well as the historical water accumulation volume of the single waterlogged area. Based on the water depth, water area, ground slope, seepage rate, external water inflow rate, and drainage outlet status of the corresponding water accumulation area in different historical water accumulation data, a water accumulation prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The water depth, water area, ground slope, seepage rate, external water inflow rate and drainage outlet status in the different historical water volume data in the second training set are used as the input data of the second convolutional neural network, and the corresponding historical water volume in the second training set is used as the output data of the second convolutional neural network. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network that outputs a second test error threshold less than or equal to the preset second test error threshold is used as the water accumulation prediction model. The water depth, water area, ground slope, seepage rate, external water inflow rate, and drainage outlet status of each waterlogged area are input into the water volume prediction model to obtain the predicted water volume for each waterlogged area.

9. The rescue route planning system for the drainage robot according to claim 8, characterized in that, The process of dispatching drainage robots for support includes: The predicted water volume for each waterlogged area is obtained based on the water volume prediction model. The drainage completion time for each waterlogged area is obtained based on the predicted water volume, natural drainage rate, and maximum drainage rate of the drainage robot. The total completion time for each drainage robot to complete the drainage rescue is obtained based on the drainage completion time of each waterlogged area, the average travel speed of the drainage robot, and the degree of optimization of the drainage rescue route. The average total completion time is obtained based on the total completion time of all optimized drainage rescue routes. Drainage robots with a total completion time greater than the average total completion time are set as drainage robots that need support, and drainage robots with a total completion time less than the average total completion time are set as drainage robots that can be supported. For robots requiring support for flood drainage, select the nearest available robot from the cluster of available flood drainage robots. Analyze the remaining optimized flood drainage and rescue routes of the robots requiring support and find an optimal task split point. Usually, the waterlogged areas at the back of the route, especially those with low urgency or long operation time, are cut off in whole or in part and redistributed to available flood drainage robots. The optimal task split point is when the time for the two flood drainage robots to complete their remaining tasks is rebalanced. The two flood drainage robots replan their conflict-free routes to the newly assigned task points and coordinate their operation time windows to avoid meeting in the same narrow space. For large waterlogged areas, a collaborative mode of simultaneous entry from opposite sides can be designed. During the support mission, the progress of both parties is continuously monitored. Once the risk of delay for the drainage robots that need support is eliminated or the original mission is completed, the drainage robots that can be supported are directed to withdraw in an orderly manner and continue to execute their own suspended original mission sequence.

10. A method for planning rescue routes for a flood drainage robot, implemented based on the rescue route planning system for a flood drainage robot as described in any one of claims 1-9, characterized in that, Includes the following steps: S1: Real-time collection of environmental data from various waterlogged areas in underground space, normalization of the environmental data, generation of an emergency score for each waterlogged area based on the processed data, and classification of the waterlogged areas into three emergency levels: high, medium, and low, based on a preset emergency score threshold. S2: Sort all high-urgency areas by score, select the nearest and most urgent area as the starting point for each drainage robot, and allocate all high and medium-urgency areas in sequence to form the initial drainage and rescue route covering all high and medium-urgency waterlogged areas. S3: Construct a matching degree prediction model to predict the matching degree between each low-emergency waterlogging area and each initial drainage and rescue route. Based on the matching degree, insert the low-emergency waterlogging area into the optimal position of the initial drainage and rescue route to generate an optimized drainage and rescue route. S4: Construct a water accumulation prediction model to predict the water accumulation in each water accumulation area. Based on the maximum drainage rate of the drainage robot and the natural drainage rate of the water accumulation area, calculate the drainage completion time of each water accumulation area and the predicted total completion time of each optimized drainage and rescue route. S5: Calculate the average total completion time of all drainage robots, select the nearest available drainage robot for the drainage robot that needs support, divide the subsequent tasks of the drainage robot that needs support to the available drainage robots, and replan the non-conflicting paths between the two.

Images