AI-driven building demolition strategy generation method, device, equipment and storage medium

Abstract

The application provides an AI-driven building demolition strategy generation method, device, equipment and storage medium, the application obtains building basic data of a target block; a building fire risk layer is generated by performing fire risk assessment; an initial population composed of multiple demolition strategies is randomly generated based on constraint conditions; the population is input into a fire simulation system, fire rescue simulation and personnel evacuation simulation are respectively performed based on the fire risk layer, and rescue and evacuation efficiency indexes of each strategy are calculated; the indexes are used as quantitative values of optimization targets, a multi-objective evolutionary algorithm is used for iterative optimization of the population, and an optimal solution set considering rescue efficiency and evacuation efficiency is generated. The application deeply integrates multi-objective optimization and multi-agent simulation, realizes intelligent decision-making of significantly improving block fire safety with the least demolition amount on the premise of protecting historical buildings, and provides scientific and systematic technical support for protection and updating of historical and cultural blocks.

Description

Technical Field

[0001] This invention relates to the field of fire protection planning technology, and in particular to an AI-driven method, apparatus, equipment and storage medium for generating building demolition strategies. Background Technology

[0002] Historic and cultural districts carry the historical memory and cultural heritage of a city. Their buildings are mostly wooden structures with low fire resistance, and coupled with high building density, mixed functions, and dense population, the fire hazards are extremely prominent. In the event of a fire, the narrow and winding streets not only obstruct the passage of fire trucks but also severely hinder the evacuation of people. While protecting the historical context, appropriately demolishing some buildings is an effective way to improve the fire safety of these districts. However, current technologies mainly focus on individual building maintenance, fire risk prediction, or the location of fire-fighting facilities, and have not yet addressed the automatic generation and optimization of building demolition plans. Traditional methods rely on manual experience, making it difficult to achieve a scientific balance between cultural heritage protection and fire safety improvement. They lack consideration for both rescue and evacuation efficiency, often resulting in demolition plans that are incomplete and fail to achieve overall optimization.

[0003] In summary, the problems existing in the current technology urgently need to be solved. Summary of the Invention

[0004] This invention provides an AI-driven method, apparatus, equipment, and storage medium for generating building demolition strategies, which addresses the shortcomings of existing technologies and significantly improves the overall fire safety level of neighborhoods.

[0005] This invention provides an AI-driven method for generating building demolition strategies, comprising: Obtain basic building data for the target block; Based on the building data and the pre-built fire risk assessment system, a fire risk assessment is conducted on the target block to generate a building fire risk layer. Based on preset constraints, a demolition strategy population consisting of multiple demolition strategies is randomly generated, where each demolition strategy is a set of building identifiers to be demolished that satisfy the constraints. The current demolition strategy population is input into the preset fire simulation system. Based on the building fire risk layer, fire rescue simulation and personnel evacuation simulation are performed respectively, and the rescue efficiency index and evacuation efficiency index corresponding to each demolition strategy are calculated. Using the rescue efficiency index and the evacuation efficiency index as the quantitative values ​​of the optimization objectives, a multi-objective evolutionary algorithm is used to iteratively optimize the population of demolition schemes to generate a target optimal solution set for the demolition strategy. Each solution in the target optimal solution set corresponds to a building demolition scheme that takes into account both rescue efficiency and evacuation efficiency.

[0006] According to the present invention, an AI-driven method for generating building demolition strategies is provided, wherein the basic building data includes a building vector surface layer and corresponding attribute information, and the attribute information includes at least building identification, cultural heritage protection level, number of building floors, street-facing attributes, building function, fire resistance performance and population data.

[0007] According to an AI-driven building demolition strategy generation method provided by the present invention, the step of conducting a fire risk assessment of the target block based on the building's basic data and a pre-built fire risk assessment system to generate a building fire risk layer specifically includes: A fire risk assessment system comprising primary and secondary indicators is constructed, and the weights of the primary and secondary indicators are determined using the analytic hierarchy process (AHP). The building data is converted into spatial raster data and then weighted and superimposed according to the index weights to generate a street fire risk raster map. The values ​​of the street fire risk raster are extracted to the building point layer to obtain the fire risk value of each building, thus forming a building fire risk layer.

[0008] According to the AI-driven building demolition strategy generation method provided by the present invention, the step of randomly generating a population of demolition strategies consisting of multiple demolition strategies based on preset constraints specifically includes: Set the maximum number of demolitions; Several groups of building identifiers to be demolished are randomly generated. The number of building identifiers to be demolished in each group does not exceed the maximum number of demolished buildings, and the building identifiers to be demolished in each group meet the filtering rules for non-demolishable buildings. Each group of building markers to be demolished is treated as a demolition strategy, and all demolition strategies constitute the initial demolition strategy population. The constraints include screening rules for non-demolishable buildings.

[0009] According to an AI-driven building demolition strategy generation method provided by the present invention, the fire rescue simulation includes: In the fire simulation system, a simulation environment is built, and the simulated plots corresponding to the buildings to be demolished are marked as passable areas based on the input demolition strategy. Identify and mark road plots with a width less than a preset threshold as areas inaccessible to fire trucks; Create a fire truck agent to calculate the maximum reachable range of the fire truck under the current demolition strategy, which serves as the primary indicator of rescue efficiency. Based on the building fire risk layer, simulated fire points are randomly generated, and the shortest path from the starting point to the fire point is planned. The number of times the fire truck successfully reaches the fire point is counted as a second rescue efficiency indicator.

[0010] According to an AI-driven building demolition strategy generation method provided by the present invention, the personnel evacuation simulation includes: Based on the simulated ignition point, determine the affected buildings within the radius, count the total population in the affected buildings, and generate the corresponding number of affected personnel agents. For each disaster-affected agent, the shortest evacuation route to the nearest safe area is calculated. The calculation employs a dynamic replanning path search algorithm to avoid disaster-affected areas along the route. The simulation of the movement of disaster-affected intelligent agents along the shortest evacuation path is shown, and the movement speed of the intelligent agents is affected in real time by the population density along the path. The states of all agents are statistically analyzed, and the longest evacuation time, average evacuation time, and number of unsuccessful evacuations are calculated as indicators of evacuation efficiency.

[0011] According to the AI-driven method for generating building demolition strategies provided by the present invention, the multi-objective evolutionary algorithm is the NSGA-II algorithm, and the step of iteratively optimizing the demolition strategy population using the multi-objective evolutionary algorithm specifically includes: The current generation of the demolition strategy population is merged with the offspring population generated through crossover mutation to form a joint population; For each individual in the joint population, a rapid non-dominated ranking is performed based on its rescue efficiency index and evacuation efficiency index to obtain different non-dominated levels. Calculate the crowding distance for each individual within the same non-dominated hierarchy; Based on the level of non-dominance and the size of the crowding distance, elite individuals with the same population size as the parent population are selected from the joint population to serve as the new generation of demolition strategy population.

[0012] The present invention also provides an AI-driven building demolition strategy generation device, comprising: The data acquisition module is used to acquire basic building data for the target block. The layer generation module is used to conduct a fire risk assessment of the target block based on the building basic data and a pre-built fire risk assessment system, so as to generate a building fire risk layer. The strategy generation module is used to randomly generate a demolition strategy population consisting of multiple demolition strategies based on preset constraints, wherein each demolition strategy is a set of building identifiers to be demolished that satisfy the constraints. The strategy simulation module is used to input the current demolition strategy population into the preset fire simulation system, and to perform fire rescue simulation and personnel evacuation simulation based on the building fire risk layer, and calculate the rescue efficiency index and evacuation efficiency index corresponding to each demolition strategy. The strategy optimization module is used to use the rescue efficiency index and the evacuation efficiency index as quantitative values ​​of the optimization objectives, and to use a multi-objective evolutionary algorithm to iteratively optimize the population of demolition schemes to generate a target optimal solution set for the demolition strategy. Each solution in the target optimal solution set corresponds to a building demolition scheme that takes into account both rescue efficiency and evacuation efficiency.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the AI-driven building demolition strategy generation method as described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the AI-driven building demolition strategy generation method as described above.

[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the AI-driven building demolition strategy generation method described above.

[0016] This invention provides an AI-driven method, apparatus, device, and storage medium for generating building demolition strategies. By acquiring basic building data of a target street block and conducting fire risk assessment, a building fire risk layer is generated, providing a scientific basis for generating ignition points in subsequent simulations and making the assessment results more closely reflect the actual fire risk distribution. Secondly, based on preset constraints, a population of multiple demolition strategies is randomly generated. Each strategy is a set of building identifiers that satisfy the non-demolition rule, ensuring that cultural heritage protection requirements are considered from the initial design stage. Subsequently, the population is input into a fire simulation system to conduct fire rescue simulations and personnel evacuation simulations, calculating the rescue efficiency index and evacuation efficiency index corresponding to each strategy, achieving a high-fidelity quantitative assessment of the demolition scheme's performance in real-world scenarios. Finally, using the rescue and evacuation efficiency indices as quantitative values ​​for optimization objectives, a multi-objective evolutionary algorithm is used to iteratively optimize the population, generating an optimal solution set that balances rescue and evacuation efficiency. The entire method deeply integrates multi-objective optimization and multi-agent simulation, realizing a leap from human experience-based decision-making to data-driven intelligent decision-making. It can significantly improve the overall fire safety level of the block with minimal demolition while maximizing the protection of historical buildings, providing scientific, systematic and operable technical support for the protection and renewal of historical and cultural blocks. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the AI-driven building demolition strategy generation method provided by the present invention. Figure 2 This invention provides a building fire risk layer for an AI-driven building demolition strategy generation method. Figure 3 This is a fire rescue simulation diagram of the AI-driven building demolition strategy generation method provided by the present invention; Figure 4 This is a schematic diagram of personnel evacuation simulation based on the AI-driven building demolition strategy generation method provided by the present invention. Figure 5 This is a schematic diagram of the structure of the AI-driven building demolition strategy generation device provided by the present invention; Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] To address the problems in existing technologies, this invention proposes an AI-driven method for generating building demolition strategies, aiming to significantly improve the overall fire safety level of a neighborhood. The following describes this AI-driven building demolition strategy generation method, as follows: Figure 1 As shown, including but not limited to the following steps: Step 110: Obtain basic building data for the target block.

[0021] Step 110: Obtain basic building data for the target block. In this embodiment, the target block is taken as an example, covering an area of ​​approximately 30m × 30m in the Guangfu Middle Street area. The obtained basic building data includes a building vector surface layer, which should contain the following fields: building identifier (used to uniquely identify each building), cultural relic protection level, number of floors, whether it is a street-facing building, number of residents in the building, building function type, and building fire resistance performance. This data forms the basis for subsequent fire risk assessment and demolition strategy generation.

[0022] Step 120: Based on the building data and the pre-built fire risk assessment system, conduct a fire risk assessment on the target block to generate a building fire risk layer.

[0023] Step 120: Based on the building's basic data and a pre-constructed fire risk assessment system, a building fire risk layer is generated. In this embodiment, a multi-level indicator system is constructed, including factors such as the street environment, disaster-causing facilities, fire protection facilities and management, and the weights are determined using the analytic hierarchy process (AHP). Using a GIS platform, each factor is converted into raster data and weighted overlaid to obtain a street fire risk raster map. Finally, the raster values ​​are extracted to each building point to form a risk layer for each building. This layer will serve as the probability basis for the generation of fire points in the simulation.

[0024] Step 130: Based on preset constraints, randomly generate a demolition strategy population consisting of multiple demolition strategies, where each demolition strategy is a set of building identifiers to be demolished that satisfy the constraints.

[0025] Step 130: Randomly generate an initial demolition strategy population based on preset constraints. These constraints are set according to historical and cultural preservation requirements, such as prohibiting the demolition of protected historical buildings, high-rise buildings, and buildings facing the street. After setting the maximum number of buildings to be demolished, several sets of building identifiers to be demolished are randomly generated, each set satisfying the constraints, thus forming the initial population.

[0026] Step 140: Input the current population into the fire simulation system to simulate fire rescue and personnel evacuation.

[0027] In step 140, during the fire rescue simulation, the passable area is updated according to the demolition strategy, roads with insufficient width are marked as inaccessible, and the reachability of fire trucks and the fire-fighting success rate are calculated using the A* algorithm. During the personnel evacuation simulation, the affected population is determined based on the fire location, and the evacuation path is calculated using the A* algorithm with dynamic replanning, taking into account the impact of population density on movement speed, and evacuation time and success rate are statistically analyzed. The simulation results for each strategy are quantified into rescue efficiency indicators and evacuation efficiency indicators.

[0028] Step 150: Using the rescue efficiency index and the evacuation efficiency index as the quantitative values ​​of the optimization objectives, a multi-objective evolutionary algorithm is used to iteratively optimize the population of demolition schemes to generate a target optimal solution set for the demolition strategy. Each solution in the target optimal solution set corresponds to a building demolition scheme that takes into account both rescue efficiency and evacuation efficiency.

[0029] In step 150, using the aforementioned index as the optimization objective value, the NSGA-II multi-objective evolutionary algorithm is employed to iteratively optimize the population. Through operations such as non-dominated sorting, crowding calculation, and selection crossover mutation, the algorithm approximates the Pareto front generation by generation, ultimately outputting a set of optimal demolition scheme solutions that balance rescue and evacuation efficiency. Experiments demonstrate that the scheme generated by this method significantly outperforms the manual scheme and the non-demolition scheme in key indicators such as evacuation time and rescue success rate.

[0030] As a further optional embodiment, the building basic data includes a building vector surface layer and corresponding attribute information, the attribute information including at least building identification, cultural heritage protection level, number of building floors, street-facing attribute, building function, fire resistance performance and population data.

[0031] In this embodiment, the building vector surface layer is stored in Shapefile or GeoJSON format to accurately depict the planar outline and spatial location of each building. The associated attribute information includes at least the following fields: Building Identifier, a unique identifier used to index and trace each building throughout the optimization process; Cultural Heritage Protection Level, which records the building's cultural heritage protection level, such as "Municipal Cultural Heritage Protection," "Historical Building," "Traditional Style Building Clue," or "Non-Protected Building," which is the core basis for constraint screening in step 130; Building Stories, used to determine whether it is a "High-Rise Building," typically setting 4 stories or more as non-demolishable based on the actual situation of the block; Street Facing, indicating whether the building is adjacent to an external expressway or main road, as street-facing buildings are usually included in the protection scope and not demolished due to their interface value; Building Function, describing the building's usage type, such as residential, catering, retail, warehousing, etc., which is used to assign values ​​to the "Building Function" index in step 120 fire risk assessment; Fire Resistance, reflecting the building's structural materials and fire resistance rating, such as frame structure, brick-concrete structure, brick-wood structure, etc., which is an important input for fire risk calculation; Population Data, recording the number of permanent or instantaneous residents in the building, used for the statistics of the number of disaster victims in step 140 personnel evacuation simulation. The completeness and accuracy of the above data provide a solid data foundation for the subsequent refined calculation of fire risk assessment and the scientific decision-making of demolition strategies.

[0032] As a further optional embodiment, the step of conducting a fire risk assessment of the target block based on the building infrastructure data and a pre-built fire risk assessment system to generate a building fire risk layer specifically includes: A fire risk assessment system comprising primary and secondary indicators is constructed, and the weights of the primary and secondary indicators are determined using the analytic hierarchy process (AHP). The building data is converted into spatial raster data and then weighted and superimposed according to the index weights to generate a street fire risk raster map. The values ​​of the street fire risk raster are extracted to the building point layer to obtain the fire risk value of each building, thus forming a building fire risk layer.

[0033] In this embodiment, a fire risk assessment system comprising primary and secondary indicators is first constructed (as shown in Table 1). The primary indicators cover three dimensions: neighborhood environment, disaster-causing facilities, and fire protection facilities and management. Each primary indicator is further subdivided into secondary indicators, such as building density, building height, building function, fire resistance performance, cultural heritage protection level, population density, distance to garbage stations, distance to non-motorized vehicle parking areas, accessibility of fire hydrants, accessibility of evacuation areas, and distance to bulletin boards. By introducing the analytic hierarchy process (AHP), domain experts are invited to conduct pairwise comparisons of the indicators at each level, constructing a judgment matrix and calculating the weights of each indicator to ensure the scientific rigor and objectivity of the assessment system.

[0034] Secondly, the building data is converted into spatial raster data and then weighted and overlaid according to the indicator weights to generate a street fire risk raster map (see Table 2). Specifically, using the spatial analysis tools of the GIS platform, the vector data corresponding to each indicator (such as building points, facility points, and road lines) is converted into continuously distributed raster layers. The value of each raster cell represents the risk score (1-5 points) of that location on that indicator. Subsequently, based on the indicator weights determined by the analytic hierarchy process (AHP), the raster layers are weighted and summed to obtain a raster map that comprehensively reflects the spatial distribution of street fire risk. Areas with higher risk are shown with higher values ​​on the map.

[0035] Finally, the values ​​of the street fire risk raster are extracted to the building point layer to obtain the fire risk value of each building, forming a building fire risk layer (see reference). Figure 2 In practice, the "Value Extraction to Point" tool in GIS is used to extract the raster values ​​of each building's location from the fire risk raster map, taking the building point layer as input, and attaching them as attribute fields to the building layer. The resulting building fire risk layer quantitatively associates each building with its fire risk, providing a direct probabilistic basis for the random generation of simulated fire points in subsequent step 140.

[0036] The fire risk factor weighting table and the assignment standards for building function and cultural relic protection level indicators are shown below:

[0037] Table 1: Fire Risk Factor Weighting Table

[0038] Table 2: Value Assignment Standards for Building Function and Cultural Relics Protection Level Indicators As a further optional embodiment, the step of randomly generating a population of demolition strategies consisting of multiple demolition strategies based on preset constraints specifically includes: Set the maximum number of demolitions; Several groups of building identifiers to be demolished are randomly generated. The number of building identifiers to be demolished in each group does not exceed the maximum number of demolished buildings, and the building identifiers to be demolished in each group meet the filtering rules for non-demolishable buildings. Each group of building markers to be demolished is treated as a demolition strategy, and all demolition strategies constitute the initial demolition strategy population. The constraints include screening rules for non-demolishable buildings.

[0039] In this embodiment, the maximum number of demolitions is first set. Based on the protection requirements of the historical and cultural district, the actual renovation budget, and the social impact assessment, the planner pre-sets an upper limit value X for the number of demolitions (e.g., X=10 or 15). This value is used to control the aggressiveness of each demolition strategy, avoiding damage to the district's appearance due to excessive demolition, and also providing boundary constraints for the search space of subsequent optimization algorithms.

[0040] Secondly, several sets of building identifiers to be demolished are randomly generated. In the Python environment, based on the basic building data obtained in step 110, the program first filters all buildings according to the screening rules for non-demolishable buildings, generating a candidate pool of demolishable buildings. The screening rules for non-demolishable buildings include, but are not limited to: cultural relics protection buildings (such as municipal-level cultural relics protection units, district-registered protection units, historical buildings, etc.) cannot be demolished; high-rise buildings with more than 4 floors cannot be demolished; and street-facing buildings adjacent to external expressways or main roads cannot be demolished. From the candidate pool, the program generates several sets of building identifiers to be demolished through random sampling. The number of building identifiers in each set does not exceed the preset maximum number of demolition identifiers X. Random sampling can be carried out using uniform distribution random sampling, or a certain bias can be introduced (such as a tendency to select buildings with higher fire risk) to improve the quality of the initial population, but it must be ensured that each set of samples meets the constraints.

[0041] Finally, each group of building identifiers to be demolished is treated as a demolition strategy, and all demolition strategies together constitute the initial demolition strategy population. In this embodiment, the size N of the initial population is set by the planner (e.g., N=50), meaning 50 different random demolition strategies are generated at once. Each strategy is mathematically represented as a list of building identifiers in the form {ID_3, ID_7, ID_15}, representing the specific building to be demolished. These strategies will be input one by one into the fire simulation system for evaluation in subsequent steps, and continuously iterated and optimized through a multi-objective evolutionary algorithm, ultimately converging to a Pareto optimal solution set that balances rescue and evacuation efficiency.

[0042] As a further optional embodiment, the fire rescue simulation includes: In the fire simulation system, a simulation environment is built, and the simulated plots corresponding to the buildings to be demolished are marked as passable areas based on the input demolition strategy. Identify and mark road plots with a width less than a preset threshold as areas inaccessible to fire trucks; Create a fire truck agent to calculate the maximum reachable range of the fire truck under the current demolition strategy, which serves as the primary indicator of rescue efficiency. Based on the building fire risk layer, simulated fire points are randomly generated, and the shortest path from the starting point to the fire point is planned. The number of times the fire truck successfully reaches the fire point is counted as a second rescue efficiency indicator.

[0043] Reference Figure 3 In this embodiment, firstly, a simulation environment is constructed within the fire simulation system. This embodiment uses the Netlogo platform to construct a multi-agent simulation environment, importing and mapping the building vector surface layer obtained in step 110 to patch units in the simulation world. Each patch intersecting with a building surface is marked as a "building" state. When a specific demolition strategy (i.e., a set of building identifiers to be demolished) is input, the system iterates through each building identifier in the strategy, changing the state of all patches occupied by the corresponding building from "building" to "accessible," thereby simulating the open space or passageway formed after the building demolition.

[0044] Secondly, the system identifies and marks areas inaccessible to fire trucks. After marking buildings for demolition, the system checks the width of all patches marked "roads" (i.e., patches not occupied by any buildings). Specifically, a buffer zone with a radius of 1 meter (corresponding to 1 unit in the simulation) is created for each road patch. The system then checks whether this buffer zone contains other building patches. If no more building patches appear in the buffer zone, the road width is determined to be greater than or equal to 2 meters, and it is marked as "accessible by fire trucks." Conversely, if other building patches exist in the buffer zone, the road width is determined to be less than 2 meters, and it is marked as "inaccessible by fire trucks." This logic is based on the minimum width requirement for fire truck passage.

[0045] Next, the maximum reachability of the fire truck is calculated. The system creates a fire truck agent and sets its initial location (usually a main entrance to the block or a pre-defined fire station location). Starting from this location, a breadth-first search is performed using the A* pathfinding algorithm, allowing the agent to move only to road patches marked as "fire truck reachable". After the search is complete, all reachable patches constitute the fire truck reachability under the current demolition strategy. The number or area of ​​patches covered by this range is used as the first rescue efficiency indicator, denoted as "fire truck reachability". This indicator directly reflects the degree to which the demolition plan improves the connectivity of the block's road network.

[0046] Finally, a batch firefighting simulation experiment was conducted. Based on the building fire risk layer generated in step 120, the system randomly generated simulated fire points according to the fire risk value of each building as a probability weight. For each generated fire point, the A* pathfinding algorithm was called again to plan the shortest rescue path, starting from the current location of the fire truck and ending at the location of the fire point. If a feasible path exists and the entire path is within the reachable area, the rescue is considered successful, and the path length is recorded; otherwise, the rescue is considered a failure. The above simulation was repeated a preset number of times (e.g., 100 or 200 times), and the total number of times the fire truck successfully reached the fire point was counted as the second rescue efficiency indicator, denoted as "number of successful rescues". This indicator reflects the overall rescue reliability of the demolition plan in the face of random fire events.

[0047] As a further optional embodiment, the personnel evacuation simulation includes: Based on the simulated ignition point, determine the affected buildings within the radius, count the total population in the affected buildings, and generate the corresponding number of affected personnel agents. For each disaster-affected agent, the shortest evacuation route to the nearest safe area is calculated. The calculation employs a dynamic replanning path search algorithm to avoid disaster-affected areas along the route. The simulation of the movement of disaster-affected intelligent agents along the shortest evacuation path is shown, and the movement speed of the intelligent agents is affected in real time by the population density along the path. The states of all agents are statistically analyzed, and the longest evacuation time, average evacuation time, and number of unsuccessful evacuations are calculated as indicators of evacuation efficiency.

[0048] Reference Figure 4 First, the affected area is determined and intelligent agents representing affected individuals are generated. In each simulation experiment, a fire point is randomly set based on the building fire risk layer generated in step 120. All buildings within a radius R centered on this fire point are identified as affected buildings (in this embodiment, R is 5 meters, corresponding to an actual physical distance of 5 meters). The system reads the population field from the building basic data obtained in step 110 and counts the total population in all affected buildings. Subsequently, a corresponding number of intelligent agents representing affected individuals are generated on the road patch closest to the affected buildings. Each agent represents a resident who needs to be evacuated, and its location, target state, and other attributes are initialized.

[0049] Secondly, the shortest evacuation route from the affected individuals to the safe zone is calculated. Safe zones fall into two categories: one is refuge areas around the neighborhood (such as squares, parks, and refuge buildings), and the other is safe roads (such as main passageways wider than 4 meters that are unaffected by the fire). The system first calculates the Euclidean distance between each affected individual and all safe zones, selecting the closest safe zone as the evacuation target point. Subsequently, the A* path search algorithm with dynamic replanning is used to calculate the shortest evacuation route for each individual. The unique feature of this algorithm is that during path planning, the system continuously monitors whether the path passes through the current fire's influence area or areas already affected by the fire. If the path passes through a disaster area, a replanning mechanism is triggered to recalculate the path, bypassing the dangerous area, until a completely safe and feasible path is found or it is determined that there is no way to proceed.

[0050] Next, the movement process of the agents of disaster victims is simulated. After the path is determined, the agents begin to move along the path. The movement speed is not constant, but is affected in real time by the population density on the current path. Specifically, the system counts the number of agents on each road patch at regular time steps, calculates the local population density, and adjusts the movement speed of the agents on that patch according to a preset speed-density relationship function: the higher the density, the lower the movement speed, thus simulating the crowd congestion effect. When an agent moves to a distance of less than a threshold (e.g., 1 meter) from the boundary of the safe zone, it is marked as "evacuated" and removed from the simulation; if an agent cannot move for several consecutive time steps (e.g., trapped by congestion or fire) and the cumulative time exceeds the preset maximum evacuation time threshold, it is marked as "trapped".

[0051] Finally, the evacuation efficiency metrics are calculated. After a simulation experiment ends (all agents are marked as "evacuated" or "trapped"), the system calculates three core metrics: longest evacuation time (the maximum time taken by all successfully evacuated agents), average evacuation time (the arithmetic mean of the times taken by all successfully evacuated agents), and number of agents who failed to evacuate (the number of agents marked as "trapped"). For the same demolition strategy, the system repeats the simulation experiment multiple times (e.g., 50 or 100 times), randomly generating different ignition points in each experiment. The statistical mean of multiple experiments is taken as the quantified value of the evacuation efficiency metric for this strategy and input into the multi-objective optimization algorithm in step 150 for iterative evolution. Figure 4 For simulation results, Figure 4 (Left) Evacuation of disaster victims to the main road. Figure 4 (Right) The disaster-stricken people are evacuated to a refuge point.

[0052] As a further optional embodiment, the multi-objective evolutionary algorithm is the NSGA-II algorithm, and the step of iteratively optimizing the demolition strategy population using the multi-objective evolutionary algorithm specifically includes: The current generation of the demolition strategy population is merged with the offspring population generated through crossover mutation to form a joint population; For each individual in the joint population, a rapid non-dominated ranking is performed based on its rescue efficiency index and evacuation efficiency index to obtain different non-dominated levels. Calculate the crowding distance for each individual within the same non-dominated hierarchy; Based on the level of non-dominance and the size of the crowding distance, elite individuals with the same population size as the parent population are selected from the joint population to serve as the new generation of demolition strategy population.

[0053] In this embodiment, the current generation of the dismantling strategy population is first merged with the offspring population generated through crossover and mutation to form a joint population. During each generation of optimization, individuals in the current parent population (size N) are first sorted using non-dominated ranking and crowding distance calculations. Then, a tournament selection method is used to select superior individuals from the parent generation for a mating pool. Individuals in the mating pool are randomly paired, and simulated binary crossover and polynomial mutation operations are performed to generate an offspring population of size N. The parent and offspring populations are then merged to obtain a joint population of size 2N. The purpose of this operation is to retain superior genes from the parent generation while introducing new characteristics into the offspring, providing a larger candidate space for subsequent elite selection.

[0054] Secondly, a fast non-dominated ranking is performed on each individual in the joint population. In this embodiment, each individual corresponds to a demolition strategy, and its fitness is composed of multiple rescue efficiency indicators (such as fire truck reachability and number of successful rescues) and evacuation efficiency indicators (such as longest evacuation time, average evacuation time, and number of unsuccessful evacuees) calculated in step 140. The core of the fast non-dominated ranking is to compare the dominance relationships between individuals: if all indicators of individual A are not inferior to those of individual B, and at least one indicator is strictly superior to that of individual B, then A is said to dominate B. By traversing all individuals in the joint population, individuals not dominated by any other individual are classified into the first non-dominated level (F1). Then, individuals in F1 are temporarily removed, and the remaining individuals not dominated by any other individual are classified into the second level (F2), and so on, until all individuals are classified into levels. The higher the level (such as F1), the better the overall performance of the individual.

[0055] Next, the crowding distance for each individual within the same non-dominated level is calculated. Crowding distance measures the distribution density of individuals on the same frontier surface to maintain population diversity. For each non-dominated level, individuals are sorted according to their objective function values. For each objective function, the difference between the individual's two adjacent individuals on that objective is calculated, and the differences across all objectives are summed and normalized to obtain the individual's crowding distance. The crowding distance for boundary individuals is set to infinity to ensure their retention. A larger crowding distance indicates a sparser solution in the individual's region, better diversity, and priority retention in subsequent selection.

[0056] Finally, based on the level of non-dominated hierarchy and the size of the crowding distance, elite individuals with the same population size as the parent generation are selected from the joint population to form the next generation of the demolition strategy population. The selection rule is as follows: prioritize retaining all individuals at the top of the non-dominated hierarchy; if the population size N is still not reached after retaining all individuals at a certain level, then individuals with larger crowding distances are selected from the next lower level to fill the gaps, until the population size reaches N. This elite retention strategy ensures the convergence (prioritizing the retention of high-quality solutions) and diversity (retaining sparsely distributed solutions) of the algorithm, effectively avoiding premature convergence and local optima problems.

[0057] Repeat the above sub-steps until the preset maximum number of iterations G is reached (in this embodiment, G is 100 or 200). As the iterations proceed, the overall quality of the population continuously improves, and the individuals in the first non-dominated level (F1) gradually approach the true Pareto optimal frontier. The final output of all individuals in the F1 level constitutes the target optimal solution set, and each solution corresponds to an optimal building demolition scheme that achieves different trade-offs between rescue efficiency and evacuation efficiency, which planners can choose according to actual needs.

[0058] The following describes the AI-driven building demolition strategy generation device provided by the present invention, such as... Figure 5 As shown, the AI-driven building demolition strategy generation device described below and the AI-driven building demolition strategy generation method described above can be referred to in correspondence.

[0059] An AI-driven building demolition strategy generation device includes: Data acquisition module 510 is used to acquire basic building data of the target block; The layer generation module 520 is used to conduct a fire risk assessment on the target block based on the building basic data and a pre-built fire risk assessment system, so as to generate a building fire risk layer. The strategy generation module 530 is used to randomly generate a demolition strategy population consisting of multiple demolition strategies based on preset constraints, wherein each demolition strategy is a set of building identifiers to be demolished that satisfy the constraints. The strategy simulation module 540 is used to input the current demolition strategy population into the preset fire simulation system, perform fire rescue simulation and personnel evacuation simulation based on the building fire risk layer, and calculate the rescue efficiency index and evacuation efficiency index corresponding to each demolition strategy. The strategy optimization module 550 is used to use the rescue efficiency index and the evacuation efficiency index as quantitative values ​​of the optimization objectives, and to use a multi-objective evolutionary algorithm to iteratively optimize the demolition scheme population to generate a target optimal solution set for the demolition strategy. Each solution in the target optimal solution set corresponds to a building demolition scheme that takes into account both rescue efficiency and evacuation efficiency.

[0060] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communications interface 620, and the memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute an AI-driven building demolition strategy generation method, which includes: Obtain basic building data for the target block; Based on the building data and the pre-built fire risk assessment system, a fire risk assessment is conducted on the target block to generate a building fire risk layer. Based on preset constraints, a demolition strategy population consisting of multiple demolition strategies is randomly generated, where each demolition strategy is a set of building identifiers to be demolished that satisfy the constraints. The current demolition strategy population is input into the preset fire simulation system. Based on the building fire risk layer, fire rescue simulation and personnel evacuation simulation are performed respectively, and the rescue efficiency index and evacuation efficiency index corresponding to each demolition strategy are calculated. Using the rescue efficiency index and the evacuation efficiency index as the quantitative values ​​of the optimization objectives, a multi-objective evolutionary algorithm is used to iteratively optimize the population of demolition schemes to generate a target optimal solution set for the demolition strategy. Each solution in the target optimal solution set corresponds to a building demolition scheme that takes into account both rescue efficiency and evacuation efficiency.

[0061] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0062] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the AI-driven building demolition strategy generation method provided by the above methods, the method comprising: Obtain basic building data for the target block; Based on the building data and the pre-built fire risk assessment system, a fire risk assessment is conducted on the target block to generate a building fire risk layer. Based on preset constraints, a demolition strategy population consisting of multiple demolition strategies is randomly generated, where each demolition strategy is a set of building identifiers to be demolished that satisfy the constraints. The current demolition strategy population is input into the preset fire simulation system. Based on the building fire risk layer, fire rescue simulation and personnel evacuation simulation are performed respectively, and the rescue efficiency index and evacuation efficiency index corresponding to each demolition strategy are calculated. Using the rescue efficiency index and the evacuation efficiency index as the quantitative values ​​of the optimization objectives, a multi-objective evolutionary algorithm is used to iteratively optimize the population of demolition schemes to generate a target optimal solution set for the demolition strategy. Each solution in the target optimal solution set corresponds to a building demolition scheme that takes into account both rescue efficiency and evacuation efficiency.

[0063] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements an AI-driven building demolition strategy generation method provided by the methods described above, the method comprising: Obtain basic building data for the target block; Based on the building data and the pre-built fire risk assessment system, a fire risk assessment is conducted on the target block to generate a building fire risk layer. Based on preset constraints, a demolition strategy population consisting of multiple demolition strategies is randomly generated, where each demolition strategy is a set of building identifiers to be demolished that satisfy the constraints. The current demolition strategy population is input into the preset fire simulation system. Based on the building fire risk layer, fire rescue simulation and personnel evacuation simulation are performed respectively, and the rescue efficiency index and evacuation efficiency index corresponding to each demolition strategy are calculated. Using the rescue efficiency index and the evacuation efficiency index as the quantitative values ​​of the optimization objectives, a multi-objective evolutionary algorithm is used to iteratively optimize the population of demolition schemes to generate a target optimal solution set for the demolition strategy. Each solution in the target optimal solution set corresponds to a building demolition scheme that takes into account both rescue efficiency and evacuation efficiency.

[0064] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0065] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An AI-driven method for generating building demolition strategies, characterized in that, include: Obtain basic building data for the target block; Based on the building data and the pre-built fire risk assessment system, a fire risk assessment is conducted on the target block to generate a building fire risk layer. Based on preset constraints, a demolition strategy population consisting of multiple demolition strategies is randomly generated, where each demolition strategy is a set of building identifiers to be demolished that satisfy the constraints. The current demolition strategy population is input into the preset fire simulation system. Fire rescue simulation and personnel evacuation simulation are performed based on the building fire risk layer. The rescue efficiency index and evacuation efficiency index corresponding to each demolition strategy are calculated. Using the rescue efficiency index and the evacuation efficiency index as the quantitative values ​​of the optimization objectives, a multi-objective evolutionary algorithm is used to iteratively optimize the population of demolition schemes to generate a target optimal solution set for the demolition strategy. Each solution in the target optimal solution set corresponds to a building demolition scheme that takes into account both rescue efficiency and evacuation efficiency.

2. The AI-driven building demolition strategy generation method according to claim 1, characterized in that, The building basic data includes a building vector surface layer and corresponding attribute information. The attribute information includes at least building identification, cultural heritage protection level, number of building floors, street frontage, building function, fire resistance performance, and population data.

3. The AI-driven building demolition strategy generation method according to claim 1, characterized in that, The step of conducting a fire risk assessment of the target block based on the building infrastructure data and a pre-built fire risk assessment system to generate a building fire risk layer specifically includes: A fire risk assessment system comprising primary and secondary indicators is constructed, and the weights of the primary and secondary indicators are determined using the analytic hierarchy process (AHP). The building data is converted into spatial raster data and then weighted and superimposed according to the index weights to generate a street fire risk raster map. The values ​​of the street fire risk raster are extracted to the building point layer to obtain the fire risk value of each building, thus forming a building fire risk layer.

4. The AI-driven building demolition strategy generation method according to claim 1, characterized in that, The step of randomly generating a population of demolition strategies consisting of multiple demolition strategies based on preset constraints specifically includes: Set the maximum number of demolitions; Several groups of building identifiers to be demolished are randomly generated. The number of building identifiers to be demolished in each group does not exceed the maximum number of demolished buildings, and the building identifiers to be demolished in each group meet the filtering rules for non-demolishable buildings. Each group of building markers to be demolished is treated as a demolition strategy, and all demolition strategies constitute the initial demolition strategy population. The constraints include screening rules for non-demolishable buildings.

5. The AI-driven building demolition strategy generation method according to claim 1, characterized in that, The fire rescue simulation includes: In the fire simulation system, a simulation environment is built, and the simulated plots corresponding to the buildings to be demolished are marked as passable areas based on the input demolition strategy. Identify and mark road plots with a width less than a preset threshold as areas inaccessible to fire trucks; Create a fire truck agent to calculate the maximum reachable range of the fire truck under the current demolition strategy, which serves as the primary indicator of rescue efficiency. Based on the building fire risk layer, simulated fire points are randomly generated, and the shortest path from the starting point to the fire point is planned. The number of times the fire truck successfully reaches the fire point is counted as a second rescue efficiency indicator.

6. The AI-driven building demolition strategy generation method according to claim 1, characterized in that, The personnel evacuation simulation includes: Based on the simulated ignition point, determine the affected buildings within the radius, count the total population in the affected buildings, and generate the corresponding number of affected personnel agents. For each disaster-affected agent, the shortest evacuation route to the nearest safe area is calculated. The calculation employs a dynamic replanning path search algorithm to avoid disaster-affected areas along the route. The simulation of the movement of disaster-affected intelligent agents along the shortest evacuation path is shown, and the movement speed of the intelligent agents is affected in real time by the population density along the path. The states of all agents are statistically analyzed, and the longest evacuation time, average evacuation time, and number of unsuccessful evacuations are calculated as indicators of evacuation efficiency.

7. The AI-driven building demolition strategy generation method according to claim 1, characterized in that, The multi-objective evolutionary algorithm is the NSGA-II algorithm. The step of iteratively optimizing the demolition strategy population using the multi-objective evolutionary algorithm specifically includes: The current generation of the demolition strategy population is merged with the offspring population generated through crossover mutation to form a joint population; For each individual in the joint population, a rapid non-dominated ranking is performed based on its rescue efficiency index and evacuation efficiency index to obtain different non-dominated levels. Calculate the crowding distance for each individual within the same non-dominated hierarchy; Based on the level of non-dominance and the size of the crowding distance, elite individuals with the same population size as the parent population are selected from the joint population to serve as the new generation of demolition strategy population.

8. An AI-driven building demolition strategy generation device, characterized in that, include: The data acquisition module is used to acquire basic building data for the target block. The layer generation module is used to conduct a fire risk assessment of the target block based on the building basic data and a pre-built fire risk assessment system, so as to generate a building fire risk layer. The strategy generation module is used to randomly generate a demolition strategy population consisting of multiple demolition strategies based on preset constraints, wherein each demolition strategy is a set of building identifiers to be demolished that satisfy the constraints. The strategy simulation module is used to input the current demolition strategy population into the preset fire simulation system, and to perform fire rescue simulation and personnel evacuation simulation based on the building fire risk layer, and calculate the rescue efficiency index and evacuation efficiency index corresponding to each demolition strategy. The strategy optimization module is used to use the rescue efficiency index and the evacuation efficiency index as quantitative values ​​of the optimization objectives, and to use a multi-objective evolutionary algorithm to iteratively optimize the population of demolition schemes to generate a target optimal solution set for the demolition strategy. Each solution in the target optimal solution set corresponds to a building demolition scheme that takes into account both rescue efficiency and evacuation efficiency.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the AI-driven building demolition strategy generation method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the AI-driven building demolition strategy generation method as described in any one of claims 1 to 7.

Images