A*hybrid ant colony path planning method for simulating flame spreading grid map

Abstract

The invention belongs to the technical field of path planning, and particularly relates to an A * hybrid ant colony path planning method for simulating a flame spreading grid. Comprising the following steps: interactively setting a plurality of fire source positions in an obstacle environment of an initial static grid simulation map, and selecting a starting point; obtaining terrain analysis data and fuel model data of the map; calculating the spreading rate value of each basic grid, and constructing a conversion rule among a non-combustion state, a combustion state and a burnout state to simulate flame spreading; the map information is input into a function formed by nesting a simplified ant colony algorithm into a traditional A * algorithm, and the optimal path for the robot to go to a plurality of target points is planned; simulating the movement of the robot to dynamically present the optimal global path; the flame spreading model and the robot dynamic path planning model are input into a map function together, optimal path planning of the simulation robot in the dynamically changing flame spreading environment is achieved, and the environment and robot moving conditions are displayed in a basic grid map in real time. According to the invention, the optimal path planning for the robot to go to a plurality of target fire sources to implement the fire extinguishing activity under the obstacle and dynamic flame environment can be realized.

Description

Technical Field

[0001] The invention belongs to the technical field of path planning, in particular to an A* hybrid ant colony path planning method for simulating flame spreading grids. Background Art

[0002] The method of constructing flame spread grid maps is one of the key technologies in wildfire management, disaster prevention and control, and ecological research. It involves multidisciplinary intersection and is closely related to the progress of computer technology, physical models, and geographic information science.

[0003] Early research relied on field observations and laboratory combustion experiments to establish empirical relationships between fire spread rate and wind speed, slope, and fuel type. For example, Byram (1959) proposed a formula for fire line intensity, which laid the foundation for the conservation of energy for fire spread. Richard Rothermel (1972) of the U.S. Forest Service proposed the first physics-based fire spread model, calculating fire line velocity based on fuel properties (moisture content, density) and environmental parameters (wind speed, slope), which became the core algorithm for subsequent grid simulations.

[0004] With the rise of numerical methods and computer simulation, scholars such as Green (1983) applied cellular automata to fire spread simulation. Each grid cell was updated based on the neighboring state (burned / unburned), which simplified the calculation of complex propagation processes. The popularization of geographic information systems (GIS) (such as ArcGIS) promoted the grid processing of spatial data. The FARSITE software developed by Finney (1998) combined the Rothermel model with GIS to realize the grid prediction of dynamic fire spread for the first time, becoming an industry standard tool.

[0005] With the advent of the era of big data and intelligence, flame spread models tend to be real-time and precise. McGrattan et al. (2000) introduced tools such as FDS (Fire Dynamics Simulator) for high-fidelity simulation of turbulence, thermal radiation and combustion chemistry, supporting refined modeling of meter-level grids; satellite data such as MODIS, Sentinel-2 and drone data can realize real-time updates of combustible material humidity and fire line location, improving the dynamic calibration capability of grid models, and deep learning such as CNN (convolutional neural network) and graph neural network can be used to learn propagation laws from historical fire data, replacing traditional empirical formulas. Research based on these two is in full swing, such as the fire spread prediction AI model developed by the University of California team (2020), Radke et al. (2021) using CNN to predict flame spread from remote sensing data; and digital twin platforms based on high-precision grids such as LANDFIRE combined with IoT sensors support real-time fire situation simulation and emergency resource scheduling.

[0006] Current path planning algorithms are categorized into global path planning for known environments and local path planning for unknown environments, depending on whether they have prior knowledge of the environment. Global planning is a static form of planning that primarily provides guidance and constraints for local path planning, while local planning is a dynamic form of planning used for real-time obstacle avoidance in unknown environments.

[0007] Italian scholars M. Dorigo et al. (1991) first proposed the ant colony algorithm (ACO), building upon which a robust optimization algorithm has gradually developed. The classic ACO transforms problem solving into a path search process by simulating the pheromone interactions of ants. It utilizes positive feedback and volatility mechanisms to balance global exploration with local optimization. While it is a typical example of a multi-objective global path planning algorithm, its application in obstacle avoidance environments has certain limitations.

[0008] Peter Hart et al. (1968) proposed the A* algorithm, which achieves obstacle avoidance path finding by calculating the total evaluation value from the starting point to the target node (combining the actual cost and heuristic estimation) and updating the reachability of the node by considering obstacles during the search process. Its use conditions are limited to the middle between the starting point and the target node.

[0009] Current research on firefighting robots focuses primarily on the mechanical structure and obstacle-crossing capabilities of the robots themselves, but lacks research on the intelligent capabilities of forest firefighting robots. Mature autonomous driving algorithms are highly dependent on structured roads and cannot adapt well to unstructured roads in forest environments, especially those that dynamically spread fires. Therefore, obstacle detection and avoidance in forest environments are an urgent issue that needs to be addressed. Summary of the Invention

[0010] To solve the above problems, the present invention provides an A* hybrid ant colony path planning method that simulates a flame spread grid, which can achieve optimal path planning for a robot to multiple target fire sources in an obstacle and dynamic flame environment.

[0011] The technical solution of the present invention is described as follows in conjunction with the accompanying drawings:

[0012] The present invention provides an A* hybrid ant colony path planning method for simulating a flame spread grid, comprising the following steps:

[0013] Interactively set the obstacle environment and multiple fire source locations of the initial static mesh simulation map and select the starting point to build the map model;

[0014] Obtain terrain analysis data and fuel model data from the map model;

[0015] Calculate the spread rate value of each basic grid, establish the conversion rules between the unburned state, the burned state and the burnt-out state to simulate the flame spread, that is, build a flame spread model;

[0016] The relevant information in the map model is input into the function of the simplified ant colony algorithm nested in the traditional A* algorithm to plan the optimal path for the robot to reach multiple target points, which is the robot dynamic path planning model;

[0017] The flame spread model and the robot dynamic path planning model are input into the map model together. The robot movement is simulated to dynamically present the optimal global path on the grid map drawn based on the map model.

[0018] The optimal path planning of the simulated robot in a dynamically changing flame spreading environment is achieved, and the environment and robot movement are displayed in real time on the basic grid map.

[0019] Furthermore, the specific method of interactively setting multiple fire source locations of the obstacle environment of the initial static grid simulation map and selecting the starting point, that is, building the map model, is as follows:

[0020] S11. Create an initial grid map matrix, store attribute information of each grid in the map, and preset the map size and initial map generation method;

[0021] S12. After the map is initialized, obstacles, fire sources, and starting points are defined in the map through the guidance of the graphical interface and prompt box, and different area attributes are represented by different values ​​at corresponding coordinates in the map matrix;

[0022] S13. Distinguish the locations of obstacles, fire points, and starting points through different colors in the graphical interface, and update the map display in real time.

[0023] Furthermore, the specific method of obtaining the terrain analysis data and fuel model data of the map is as follows:

[0024] S21. Download the latest version of terrain and fuel model data from the LANDFIRE website before the fire or use your own environmental data.

[0025] S22. Assign the acquired data to each blank basic grid with terrain features and combustible attributes; the combustible attributes are used to describe the combustion characteristics of combustibles in the spatial environment where the basic grid is located.

[0026] Furthermore, the specific method of calculating the spread rate value of each basic grid and constructing the transition rules between the unburned state, the burned state and the burnt-out state to simulate the flame spread, that is, constructing the flame spread model, is as follows:

[0027] S31. Use the empirical model for forest fire spread prediction to calculate the initial spread rate R0 as follows:

[0028] R0=a*T+b*V+c*(100-H)+d

[0029] Where T is temperature; V is wind speed level; H is relative humidity; a, b, c, and d are empirical parameters with values ​​of 0.03, 0.05, 0.01, and -0.3, respectively.

[0030] The initial spreading rate R0 is corrected according to different fuel models, terrain slope and wind vector as follows:

[0031] R=R0*K f *K s *K w

[0032] Where R is the corrected forest fire spread rate; K f is the correction factor for combustible type; K s is the terrain slope correction coefficient, and the calculation formula is:

[0033]

[0034] Where, is the tangent value of the slope; α is the sign correction coefficient of the slope angle, when the slope is positive, α = 1, otherwise α = -1;

[0035] K w is the wind vector correction coefficient, and the calculation formula is:

[0036]

[0037] Where, v w is the wind speed; θ is the angle between the wind direction and the fire spread direction.

[0038] S32. Based on the initial state values ​​stored in the generated initial grid map, the combustible material in the unburned state is set to 1, the ignition point in the burning state is set to 2, and the feasible area in the burnt-out state is set to 0;

[0039] Traverse the initial map matrix and find the coordinates of the combustibles that are not burned;

[0040] If the current position has the ability to spread and there are burning states in the adjacent 8 basic grids, considering the accidental factors, the current position will be converted from the unburned state to the burning state after Δt1 time with probability P1=1-∈1, where ∈1 is a small probability accidental event, which is taken as 0.01;

[0041] definition Where R is the spreading rate of cells with spreading ability; L is the geographical distance between the neighboring cells and the central cell. Due to the regular space division strategy, all cells are closely adjacent square grids, so the geographical distance between cells and the cell side length are the same;

[0042] The base grid coordinates in the burning state are entered into the temporary matrix firetree=[x1,y1;x2,y2;...;x n ,y n ], and the time is counted into the ignition time set T = {t1, t2, ..., t n};

[0043] The burning grid stored in the firetree is converted from the burning state to the burnt-out state in sequence after Δt2 time with probability P2=1-∈2, considering the accidental factor, where ∈2 is a small probability accidental event, which is taken as 0.01;

[0044] The forest cell, i.e. the grid state, is calculated based on the cellular automaton iteration mode, and the temporary matrix firetree of the basic grid combustion state and the ignition time T={t1,t2,...,t n}Real-time updates are carried out to simulate the spatiotemporal spread of forest fires.

[0045] Furthermore, the specific method of inputting the relevant information in the map model into the function of the simplified ant colony algorithm nested in the traditional A* algorithm to plan the optimal path for the robot to multiple target points, that is, the robot dynamic path planning model, is as follows:

[0046] S41, taking the fire source as the target point, nesting the A* algorithm to perform obstacle avoidance optimal path planning for the starting point and the target point separately, and cyclically solving the minimum distance d between the starting point and all target points. ij ;

[0047] S42, construct the solution space, place all ants at the starting point, and each ant moves according to the probability Select the next target point until all target points are visited, and calculate the total length of the path passed by each ant;

[0048] S43. Each time all ants complete the path, the pheromone concentration is updated, the old pheromone is volatilized, and new information is added to record the optimal path in the current cycle;

[0049] S44. Repeat the ant's path construction record and pheromone update until the maximum number of iterations is reached, and return the optimal path.

[0050] Furthermore, the specific method of S41 is as follows:

[0051] S411. Introduce a grid map matrix and input the starting point and target point for this local path planning to perform path operations on the map; each entry in the matrix represents a grid, and the state is either traversable or intraversable;

[0052] S412, initializing variables and node cost attributes;

[0053] S413, add the starting point to the set of nodes to be examined, OpenList, assign the actual cost G to 0, and use the heuristic estimate H;

[0054] S414, traverse the OpenList, find the node with the smallest F generation value, use it as the current node to be processed, and move this node to the examined node set CloseList;

[0055] S415, judging the situation of the eight adjacent squares of the current square and taking corresponding measures;

[0056] S416, looping process until the end point is found or there is no feasible path, and then stop searching;

[0057] S417. If the end point is found, trace back from the end point along the parent node to the starting point, save the final searched path to PathList, and return G as the minimum distance d ij .

[0058] Furthermore, the specific method of S412 is as follows:

[0059] S4121. The actual cost G from the starting point to the current node is determined by the following formula:

[0060]

[0061] S4122. The heuristic estimate H from the current node to the target node is determined by the Chebyshev distance, as shown below:

[0062] H(n)=max(|x n -x target |,|y n -y target |)

[0063] Among them, x n ,y n are the horizontal and vertical coordinates of the current node respectively; x target ,y target are the horizontal and vertical coordinates of the target node respectively;

[0064] S4123. The evaluation function of a node consists of the actual cost G and the heuristic estimate H, as follows:

[0065] F(n)=G(n)+H(n).

[0066] Furthermore, the specific method of S415 is as follows:

[0067] S4151, the node is unreachable or in the CloseList, ignored;

[0068] S4152: If the node is not in the OpenList, add it to the OpenList and record the G, H values ​​and parent node.

[0069] S4153: The node is already in the OpenList. If this path is better, that is, the G value is smaller, set the current square as the new parent node and update the G and H values.

[0070] Furthermore, the probability in S42 The method for determining is as follows:

[0071]

[0072] Where α and β are the parameters of pheromone and heuristic factor respectively, α≥0, β≤0; S is the set of nodes that ant k can choose in the next step.

[0073] Furthermore, the method for updating the pheromone concentration in S43 is as follows:

[0074]

[0075] Where ρ is the volatility coefficient; τ ij (t) and τ ij (t+1) are the pheromone concentrations on the path in the previous and current cycles, respectively; represents the pheromone concentration left on the path by the kth ant in this cycle; Q is a constant, which is 1.

[0076] Furthermore, the flame spread model and the robot dynamic path planning model are input into the map model together, and the specific method of simulating the robot movement to dynamically present the optimal global path in the grid map drawn based on the map model is as follows:

[0077] S51, inputting the flame spread model and the robot dynamic path planning model into the map model together, so as to modify the map model information during the operation of the two models and share the latest map information in real time;

[0078] S52, controlling the dynamic display of the shortest path by setting the time interval between each operation of the robot dynamic planning model, i.e., the delay function, to simulate the robot moving in the grid map at a set speed;

[0079] S53. The simulation of the robot and the simulation of the flame spread are performed synchronously through the parallel pool, so that the robot can update the global path planning in real time in the dynamic environment and move according to the optimal path selection.

[0080] The beneficial effects of the present invention are:

[0081] 1. This invention utilizes a grid structure to expand the multiple expressions of spatial environment attributes, making the grid model a new data structure for describing the spatial environment. It also allows for trade-offs in environmental data without affecting the general trend of flame spread, thereby better meeting the needs of path planning algorithm testing.

[0082] 2. The present invention embeds the traditional A* algorithm in the classic ant colony algorithm and combines it with the use of parallel pools to achieve optimal path planning from a certain starting point to multiple target points by continuously adjusting the planned path based on effective avoidance of environmental obstacles (obstacles and fire spread areas) and real-time analysis of environmental changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0084] Figure 1 It is a schematic diagram of the process of the present invention;

[0085] Figure 2 A flowchart for building an environmental grid map under human-computer interaction conditions;

[0086] Figure 3 A grid map matrix that stores different square attributes with corresponding numbers;

[0087] Figure 4 Set up a grid map interface diagram for visual interaction;

[0088] Figure 5 A schematic diagram of a grid map that uses colors to distinguish attributes;

[0089] Figure 6 A flow chart for simulating flame spread in order to construct the transition rules between the unburned, burned and burnt-out states;

[0090] Figure 7 It is a schematic diagram of the conversion rules between the unburned state, the burned state and the burnt-out state;

[0091] Figure 8 The overall framework diagram of the optimization algorithm for multi-objective planning in a dynamic obstacle avoidance environment by nesting the ant colony algorithm with the A* algorithm;

[0092] Figure 9 A visual example of the A* hybrid ant colony algorithm for obstacle avoidance multi-objective planning;

[0093] Figure 10 Schematic diagram of the parallel pool algorithm simulating robot movement as flames spread while updating path planning in real time. DETAILED DESCRIPTION

[0094] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It will be understood that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present invention, not all structures.

[0095] See Figure 1 This embodiment provides an A* hybrid ant colony path planning method for simulating a flame spread grid, comprising the following steps:

[0096] S1. Interactively set the obstacle environment and multiple fire source locations of the initial static mesh simulation map and select the starting point to build the map model. The details are as follows:

[0097] S11. Create an initial grid map matrix, store the attribute information of each grid in the map, and preset the map size and initial map generation method (input obstacle coordinates, set vegetation coverage to randomly generate initial obstacles, and customize the grid map);

[0098] S12. After the map is initialized, the user is guided by the graphical interface and prompt box to interactively define obstacles, fire sources, and starting points in the map. Different values ​​at the corresponding coordinates in the map matrix represent different area attributes (such as Figure 3 The figure shows the element distribution of the map matrix in one embodiment, where different element values ​​represent obstacle 0, feasible area 1, fire source 2, and starting point 3).

[0099] S13, see Figure 4 and Figure 5 The map is displayed through a graphical interface and grid lines are superimposed to enhance visualization. The grid map initialization process is intuitively displayed, and different colors are used to distinguish (obstacles: black, feasible areas: white, fire sources: red, starting points: green) so that users can intuitively see the locations of obstacles, fire points and starting points, and update the map display in real time.

[0100] The following table shows the corresponding information of different areas on the map:

[0101]

[0102] S2. Obtain terrain analysis data and fuel model data of the map, as follows:

[0103] S21. Download the latest version of terrain and fuel model data from the LANDFIRE website before the fire or use your own environmental data.

[0104] S22. Assign the acquired data to each blank basic grid with terrain features and combustible attributes; the combustible attributes are used to describe the combustion characteristics of combustibles in the spatial environment where the basic grid is located.

[0105] S3. Calculate the spread rate value of each basic grid, establish the conversion rules between the unburned state, the burned state and the burnt-out state to simulate the flame spread, that is, build a flame spread model, as follows:

[0106] S31. Based on meteorological conditions and fuel model data, the forest fire spread prediction empirical model is used to calculate the initial spread rate R0 in m / min as follows:

[0107] R0=a*T+b*V+c*(100-H)+d

[0108] Where T is the temperature in °C; V is the wind speed level; H is the relative humidity; a, b, c, and d are empirical parameters with values ​​of 0.03, 0.05, 0.01, and -0.3, respectively.

[0109] The initial spreading rate R0 is corrected according to different fuel models, terrain slope and wind vector as follows:

[0110] R=R0*K f *K s *K w

[0111] Where R is the corrected forest fire spread rate; K f is the correction factor for the combustible type;

[0112]

[0113] K s is the terrain slope correction coefficient, and the calculation formula is:

[0114]

[0115] Where, is the tangent value of the slope; α is the sign correction coefficient of the slope angle, when the slope is positive, α = 1, otherwise α = -1;

[0116] K w is the wind vector correction coefficient, and the calculation formula is:

[0117]

[0118] Where, v wis the wind speed, in m / min; θ is the angle between the wind direction and the fire spread direction.

[0119] S32. Based on the initial state values ​​stored in the generated initial grid map, the combustible material (in an unburned state) is 1, the ignition point (in a burning state) is 2, and the feasible area (the burnt state is considered passable) is 0;

[0120] Traverse the initial map matrix and find the coordinates of the combustibles that are not burned;

[0121] If the current position has the ability to spread and there are burning states in the adjacent 8 basic grids, considering the accidental factors, the current position will be converted from the unburned state to the burning state after Δt1 time with probability P1=1-∈1, where ∈1 is a small probability accidental event, which is taken as 0.01;

[0122] definition Where R is the spreading rate of cells with spreading ability; L is the geographical distance between the neighboring cells and the central cell. Due to the regular space division strategy, all cells are closely adjacent square grids, so the geographical distance between cells and the cell side length are the same;

[0123] The base grid coordinates in the burning state are entered into the temporary matrix firetree=[x1,y1;x2,y2;...;x n ,y n ], and the time is counted into the ignition time set T = {t1, t2, ..., t n};

[0124] Considering the accidental factors, the burning grid stored in the firetree is converted from the unburned state to the burning state with probability P2=1-∈2 after Δt1 time, where ∈2 is a small probability accidental event, which is set to 0.01;

[0125] The forest cell, i.e. the grid state, is calculated based on the cellular automaton iteration mode, and the temporary matrix firetree of the basic grid combustion state and the ignition time T={t1,t2,...,t n}Real-time updates are carried out to simulate the spatiotemporal spread of forest fires.

[0126] S4. Input the relevant information in the map model into the function of the simplified ant colony algorithm nested in the traditional A* algorithm to plan the optimal path for the robot to reach multiple target points, namely the robot dynamic path planning model, as follows:

[0127] S41, taking the fire source as the target point, nesting the A* algorithm to perform obstacle avoidance optimal path planning for the starting point and the target point separately, and cyclically solving the minimum distance d between the starting point and all target points. ij, as follows:

[0128] S411. Introduce the grid map matrix and input the starting point and target point for this local path planning to operate the path on the map. Each entry in the matrix represents a grid, and its status is either walkable or unwalkable.

[0129] S412: Initialize variables and node cost attributes, where the node cost attributes are determined by the following steps:

[0130] S4121. The actual cost G from the starting point to the current node is determined by the following formula:

[0131]

[0132] S4122. The heuristic estimate H from the current node to the end point is determined by the Chebyshev distance, as shown below:

[0133] H(n)=max(|x n -x target |,|y n -y target |)

[0134] Where x n ,y n are the horizontal and vertical coordinates of the current node respectively; x target ,y target are the horizontal and vertical coordinates of the target node respectively;

[0135] S4123. The evaluation function of a node consists of the actual cost G and the heuristic estimate H, as follows:

[0136] F(n)=G(n)+H(n).

[0137] S413, add the starting point to the set of nodes to be examined, OpenList, assign the actual cost G to 0, and use the heuristic estimate H;

[0138] S414, traverse the OpenList, find the node with the smallest F generation value, use it as the current node to be processed (parent), and move this node to the examined node set CloseList;

[0139] S415: Determine the status of the eight adjacent squares (child nodes) of the current square and make corresponding processing, as follows:

[0140] S4151, the node is unreachable (obstacle or fire) or it is in the CloseList, ignore it;

[0141] S4152: If the node is not in the OpenList, add it to the OpenList and record its G, H values ​​and parent node.

[0142] S4153: The node is already in the OpenList. If this path is better, that is, the G value is smaller, set the current square as the new parent node and update the G and H values.

[0143] S416, loop processing until the end point is found (OpenList contains the target node) or there is no feasible path (OpenList is empty) and stop searching;

[0144] S417. If the end point is found, trace back from the end point along the parent node to the starting point, save the final searched path to PathList, and return G as the minimum distance d ij .

[0145] S42, construct the solution space, place all ants at the starting point, and each ant moves according to the probability Select the next target point until all target points are visited, and calculate the total length of the path passed by each ant;

[0146] Among them, the probability of each ant selecting the next node j from node i is Determined by the following formula:

[0147]

[0148] Where α and β are the parameters of pheromone and heuristic factor respectively (α≥0, β≤0), and S is the set of nodes that ant k can choose in the next step.

[0149] S43. Each time all ants complete the path, the pheromone concentration is updated, the old pheromone is volatilized, and new information is added to record the optimal path in the current cycle;

[0150] Among them, the method for updating pheromone concentration is as follows:

[0151]

[0152] Where ρ is the volatility coefficient; τ ij (t) and τ ij (t+1) are the pheromone concentrations on the path in the previous and current cycles, respectively; represents the pheromone concentration left on the path by the kth ant in this cycle; Q is a constant, which is 1.

[0153] S44. Repeat the ant's path construction record and pheromone update until the maximum number of iterations is reached, and return the optimal path.

[0154] like Figure 9The figure shows a visualization diagram of an embodiment of the A* hybrid ant colony algorithm for obstacle avoidance multi-objective planning. In the figure, red represents the fire source; blue represents the optimal path; and green represents the starting point.

[0155] S5. See Figure 10 In the figure, red is the spreading flame; blue is the trajectory of the robot simulation; green is the starting point;

[0156] The flame spread model and the robot dynamic path planning model are input into the map model together. The optimal global path is dynamically presented on the grid map drawn based on the map model by simulating the robot movement, as follows:

[0157] S51, controlling the dynamic display of the shortest path through a delay function, simulating the robot moving in the grid map at a set speed;

[0158] S52, using a parallel pool to synchronize the robot simulation and the flame spread simulation, so that the robot can update the global path planning in real time in a dynamic environment and move according to the optimal path selection;

[0159] S53. The simulation of the robot and the simulation of the flame spread are performed synchronously through the parallel pool, so that the robot can update the global path planning in real time in the dynamic environment and move according to the optimal path selection.

[0160] S6. Display the real-time dynamics in the initial basic grid map.

[0161] In summary, the present invention embeds the traditional A* algorithm in the classic ant colony algorithm and combines the use of parallel pools to achieve optimal path planning from a certain starting point to multiple target points by continuously adjusting the planned path based on effective avoidance of environmental obstacles (obstacles and fire spread areas) and real-time analysis of environmental changes.

[0162] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An A* hybrid ant colony path planning method for simulating flame spread grids, characterized in that: The following steps are involved: Interactively set the obstacle environment and multiple fire source locations of the initial static mesh simulation map and select the starting point to build the map model; Obtain terrain analysis data and fuel model data for maps; Calculate the spread rate value of each basic grid, establish the conversion rules between the unburned state, the burned state and the burnt-out state to simulate the flame spread, that is, build a flame spread model; The relevant information in the map model is input into the function of the simplified ant colony algorithm nested in the traditional A* algorithm to plan the optimal path for the robot to reach multiple target points, which is the robot dynamic path planning model; The flame spread model and the robot dynamic path planning model are input into the map model together. The robot movement is simulated to dynamically present the optimal global path on the grid map drawn based on the map model. The optimal path planning of the simulated robot in a dynamically changing flame spreading environment is achieved, and the environment and robot movement are displayed in real time on the basic grid map.

2. The A* hybrid ant colony path planning method for simulating flame spread grid according to claim 1, characterized in that: The specific method of interactively setting the locations of multiple fire sources in the obstacle environment of the initial static grid simulation map and selecting the starting point to build the map model is as follows: S11. Create an initial grid map matrix, store attribute information of each grid in the map, and preset the map size and initial map generation method; S12. After the map is initialized, obstacles, fire sources, and starting points are defined in the map through the guidance of the graphical interface and prompt box, and different area attributes are represented by different values ​​at corresponding coordinates in the map matrix; S13. Distinguish the locations of obstacles, fire points, and starting points through different colors in the graphical interface, and update the map display in real time.

3. The A* hybrid ant colony path planning method for simulating flame spread grid according to claim 1, characterized in that: The specific method of obtaining the terrain analysis data and fuel model data of the map is as follows: S21. Download the latest version of terrain and fuel model data from the LANDFIRE website before the fire or use your own environmental data. S22. Assign the acquired data to each blank basic grid with terrain features and combustible attributes; the combustible attributes are used to describe the combustion characteristics of combustibles in the spatial environment where the basic grid is located.

4. The A* hybrid ant colony path planning method for simulating flame spread grid according to claim 1, characterized in that: The specific method of calculating the spread rate value of each basic grid and constructing the transition rules between the unburned state, the burned state and the burnt-out state to simulate the flame spread, that is, constructing the flame spread model, is as follows: S31. Use the empirical model for forest fire spread prediction to calculate the initial spread rate R0 as follows: R0=a*T+b*V+c*(100-H)+d Where T is temperature; V is wind speed level; H is relative humidity; a, b, c, and d are empirical parameters with values ​​of 0.03, 0.05, 0.01, and -0.3, respectively. The initial spreading rate R0 is corrected according to different fuel models, terrain slope and wind vector as follows: R=R0*K f *K s *K w Where R is the corrected forest fire spread rate; K f is the correction factor for combustible type; K s is the terrain slope correction coefficient, and the calculation formula is: Where, is the tangent value of the slope; α is the sign correction coefficient of the slope angle, when the slope is positive, α = 1, otherwise α = -1; K w is the wind vector correction coefficient, and the calculation formula is: Where, v w is the wind speed; θ is the angle between the wind direction and the fire spread direction. S32. Based on the initial state values ​​stored in the generated initial grid map, the combustible material in the unburned state is set to 1, the ignition point in the burning state is set to 2, and the feasible area in the burnt-out state is set to 0; Traverse the initial map matrix and find the coordinates of the combustibles that are not burned; If the current position has the ability to spread and there are burning states in the adjacent 8 basic grids, considering the accidental factors, the current position will be converted from the unburned state to the burning state after Δt1 time with probability P1=1-∈1, where ∈1 represents a small probability accidental event and is taken as 0.01; definition Where R is the spreading rate of cells with spreading ability; L is the geographical distance between the neighboring cells and the central cell. Due to the regular space division strategy, all cells are closely adjacent square grids, so the geographical distance between cells and the cell side length are the same; The base grid coordinates in the burning state are entered into the temporary matrix firetree=[x1,y1;x2,y2;...;x n ,y n ], and the time is counted into the ignition time set T = {t1, t2, ..., t n }; The burning grid stored in the firetree is converted from an unburned state to a burning state after Δt1, taking into account the accidental factors, with a probability of P2 = 1-∈2, where ∈2 represents a small probability accidental event, which is taken as 0.01; The forest cell, i.e. the grid state, is calculated based on the cellular automaton iteration mode, and the temporary matrix firetree of the basic grid combustion state and the ignition time T={t1,t2,...,t n }Real-time updates are carried out to simulate the spatiotemporal spread of forest fires.

5. The A* hybrid ant colony path planning method for simulating flame spread grid according to claim 1, characterized in that: The specific method of inputting the relevant information in the map model into the function of the simplified ant colony algorithm nested in the traditional A* algorithm to plan the optimal path for the robot to multiple target points, that is, the robot dynamic path planning model, is as follows: S41, taking the fire source as the target point, nesting the A* algorithm to perform obstacle avoidance optimal path planning for the starting point and the target point separately, and cyclically solving the minimum distance d between the starting point and all target points. ij ; S42, construct the solution space, place all ants at the starting point, and each ant moves according to the probability Select the next target point until all target points are visited, and calculate the total length of the path passed by each ant; S43. Each time all ants complete the path, the pheromone concentration is updated, the old pheromone is volatilized, and new information is added to record the optimal path in the current cycle; S44. Repeat the ant's path construction record and pheromone update until the maximum number of iterations is reached, and return the optimal path.

6. The A* hybrid ant colony path planning method for simulating flame spread grid according to claim 5, characterized in that: The specific method of S41 is as follows: S411. Introduce a grid map matrix and input the starting point and target point for this local path planning to perform path operations on the map; each entry in the matrix represents a grid, and the state is either traversable or intraversable; S412, initializing variables and node cost attributes; S413, add the starting point to the set of nodes to be examined, OpenList, assign the actual cost G to 0, and use the heuristic estimate H; S414, traverse the OpenList, find the node with the smallest F generation value, use it as the current node to be processed, and move this node to the examined node set CloseList; S415, judging the situation of the eight adjacent squares of the current square and taking corresponding measures; S416, looping process until the end point is found or there is no feasible path, and then stop searching; S417. If the end point is found, trace back from the end point along the parent node to the starting point, save the final searched path to PathList, and return G as the minimum distance d ij .

7. The A* hybrid ant colony path planning method for simulating flame spread grid according to claim 6, characterized in that: The specific method of S412 is as follows: S4121. The actual cost G from the starting point to the current node is determined by the following formula: S4122. The heuristic estimate H from the current node to the end point is determined by the Chebyshev distance, as shown below: H(n)=max(|x n -x target |,|y n -y target |) Among them, x n ,y n are the horizontal and vertical coordinates of the current node respectively; x target ,y target are the horizontal and vertical coordinates of the target node respectively; S4123. The evaluation function of a node consists of the actual cost G and the heuristic estimate H, as follows: F(n)=G(n)+H(n).

8. The A* hybrid ant colony path planning method for simulating flame spread grid according to claim 6, characterized in that: The specific method of S415 is as follows: S4151, the node is unreachable or in the CloseList, ignored; S4152: If the node is not in the OpenList, add it to the OpenList and record the G, H values ​​and parent node. S4153: The node is already in the OpenList. If this path is better, that is, the G value is smaller, set the current square as the new parent node and update the G and H values.

9. The A* hybrid ant colony path planning method for simulating flame spread grid according to claim 5, characterized in that: The probability in S42 The method for determining is as follows: Where α and β are the parameters of pheromone and heuristic factor respectively, α≥0, β≤0; S is the set of nodes that ant k can choose in the next step.

10. The A* hybrid ant colony path planning method for simulating flame spread grid according to claim 5, characterized in that: The method for updating the pheromone concentration in S43 is as follows: Where ρ is the volatility coefficient; τ ij (t) and τ ij (t+1) are the pheromone concentrations on the path in the previous and current cycles, respectively; represents the pheromone concentration left on the path by the kth ant in this cycle; Q is a constant, which is 1.

11. The A* hybrid ant colony path planning method for simulating flame spread grid according to claim 1, characterized in that: The flame spread model and the robot dynamic path planning model are input into the map model together. The specific method for simulating robot movement and dynamically presenting the optimal global path on the grid map drawn based on the map model is as follows: S51, inputting the flame spread model and the robot dynamic path planning model into the map model together, so as to modify the map model information during the operation of the two models and share the latest map information in real time; S52, controlling the dynamic display of the shortest path by setting the time interval between each operation of the robot dynamic planning model, i.e., the delay function, to simulate the robot moving in the grid map at a set speed; S53. The simulation of the robot and the simulation of the flame spread are performed synchronously through the parallel pool, so that the robot can update the global path planning in real time in the dynamic environment and move according to the optimal path selection.

Images