Island ant colony path planning method for unmanned trolley

Abstract

The invention relates to the technical field of machine path planning, and discloses an island ant colony path planning method for an unmanned trolley, which comprises the following steps: after generating a grid map, decomposing global path search into path exploration between adjacent islands; non-uniform pheromone initialization is adopted between adjacent islands, local paths between the adjacent islands are searched respectively, and the position of a next grid is selected in the searching process by combining an improved node transition probability with a heuristic function and pheromone concentration; and when global pheromone distribution is updated, an optimal path is rewarded, and a worst path is punished. And combining the local paths of the island pairs to obtain global paths, and selecting the global path with the shortest length as the optimal driving path. According to the method, a segmented island strategy is adopted, the complexity of single search is reduced, the early random probing process is shortened through non-uniform initialization pheromones, the convergence speed is increased, the number of deadlock ants is effectively reduced, and the precision, convergence and stability of path planning are improved.

Description

Technical Field

[0001] This invention belongs to the field of machine path planning technology, specifically relating to an island ant colony path planning method for unmanned vehicles. Background Technology

[0002] With the rapid development of smart factories in China, more and more factories are achieving unmanned operation, especially in some automobile manufacturing plants, where robots have replaced manual labor in processes such as painting and stamping. The use of intelligent vehicles in cargo handling is also becoming increasingly widespread. Path planning is a crucial step in the movement of these intelligent vehicles. Ant Colony Optimization (ACO) is a commonly used path planning algorithm. ACO simulates the mechanism by which real ants communicate using pheromones: ants release pheromones along their paths, and other ants tend to choose paths with higher pheromone concentrations. The pheromones evaporate over time, and shorter paths accumulate more pheromones.

[0003] However, traditional Ant Colony Optimization (ACO) algorithms and some improved versions still have some problems, such as slow convergence speed, low search efficiency, and a tendency to generate a large number of deadlocked ants. The invention "AGV Path Planning Method Based on Improved Ant Colony Algorithm with Constructed Time Cost Function" (Patent Publication No. CN119374605A) introduces a "time cost function" to replace the "path length" in the traditional ant colony algorithm as the evaluation criterion. However, it has limitations; pheromone importance in state transition probabilities, heuristic information parameters, pheromone evaporation factors, and time cost function weights may require readjustment in different application scenarios to ensure effectiveness. Furthermore, it does not specifically address the problems of generating a large number of deadlocked ants and slow convergence speed. In addition, traditional ant colony algorithms exhibit blindness in the early stages of the search, and the pheromone concentration generated is affected by redundant paths. Their global search capability is weak, and they are prone to getting trapped in local optima, thus affecting the accuracy of ant selection. Therefore, it is necessary to improve and optimize the ACO algorithm. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an island ant colony path planning method for unmanned vehicles, which can accelerate the convergence speed in the path planning process of unmanned vehicles, avoid a large number of deadlocked ants, and favor potentially high-quality paths from the beginning of the search, thereby improving the path planning ability of the ant colony algorithm.

[0005] To address the aforementioned technical problems, this invention provides a method for island ant colony path planning for unmanned vehicles, including the unmanned vehicle path planning process:

[0006] S1. Select the grid points that must be passed through in the grid map as island points, and then distribute the initial pheromone values ​​of the grid map in a non-uniform manner using the favorable area initial pheromone distribution method.

[0007] S2. Each ant searches the path between adjacent islands in turn. During the search, it selects the next grid position through pseudo-random state transition rules until it reaches the target island and obtains a local path. Connect the local paths end to end to form the global planning path.

[0008] S3. After all ants have reached the endpoint grid, update the global pheromone distribution according to pheromone update rule one; determine whether the maximum number of iterations has been reached. If not, repeat step S2. If it has been reached, select the shortest global planning path among all ants as the optimal path.

[0009] An improvement to the island ant colony path planning method for unmanned vehicles:

[0010] The grid order for path planning is an island point sequence. ,in, Indicates the index of the starting grid S. The sequence number represents the endpoint grid E, and N is the total number of islands; island point sequence The construction method is as follows:

[0011] With islands Center is the center of the circle, radius All free grid points within the search circle are candidate island points. Each candidate island point The grid with the smallest Euclidean distance to the endpoint grid E is the island. Then with islands Center is the center of the circle, radius Redefine the search circle and candidate island points to locate the islands. This process continues until the search circle contains the endpoint grid E, thus obtaining the island point sequence. ;radius The calculation formula is:

[0012] ;

[0013] in, This represents the Euclidean distance between the starting grid S and the ending grid E.

[0014] A further improvement to the island ant colony path planning method for unmanned vehicles:

[0015] The method for initial pheromone distribution in the favorable region is as follows:

[0016] With adjacent islands and islands The rectangular region defined by the diagonal vertices is the favorable region, and the initial pheromone value is:

[0017] ;

[0018] in, This represents the initial pheromone value of the favorable region. This represents the initial pheromone value of the global map. These represent the grid numbers respectively;

[0019] ;

[0020] in, The side length of a unit grid; For islands and islands Euclidean distance; For islands and grid Euclidean distance; For grid and grid Euclidean distance; For grid and islands Euclidean distance.

[0021] A further improvement to the island ant colony path planning method for unmanned vehicles:

[0022] The pseudo-random state transition rule is as follows:

[0023] In time At that time, the Kth ant wants to move from the current grid. Move to the next grid Node transition probability for:

[0024] ;

[0025] in, and These represent the pheromone concentration factor and the expected heuristic factor, respectively. Represents a grid The collection of surrounding unvisited grids, For heuristic function values:

[0026] ;

[0027] in, It is a scaling factor. and These represent the maximum number of iterations and the current number of iterations, respectively. Represents a grid The Euclidean distance to the endpoint grid E.

[0028] A further improvement to the island ant colony path planning method for unmanned vehicles:

[0029] The first pheromone update rule is:

[0030] ;

[0031] in, This indicates the evaporation of the original pheromone. This means that when ants pass through a path, they add new pheromones to the path. Indicates a reward or punishment item. Indicates adaptive evaporation rate, Indicates the current iteration loop from the grid To grid Reward or penalty value:

[0032] ;

[0033] in, The total amount of pheromones is a constant. It is the length of the optimal path in the current iteration. It is the length of the worst path in the current iteration;

[0034] ;

[0035] in, It is an adjustment factor.

[0036] This invention also provides another method for island ant colony path planning for unmanned vehicles, including the following steps:

[0037] S1. In the grid map, potential deadlock grids are converted into obstacle grids to generate an auxiliary map. Then, grids that must be passed through are selected from the auxiliary map as island points. Then, the auxiliary map is subjected to non-uniform initial pheromone value distribution by a pheromone initialization method based on the dominant region.

[0038] S2. Each ant searches the path between adjacent islands in turn. During the search, it selects the next grid position based on the angle factors between the previous node, the current node and the next node, until it reaches the target island point and obtains a local path. Connect the local paths end to end to form the global planning path.

[0039] S3. After all ants have reached the endpoint grid, update the global pheromone distribution according to pheromone update rule two, and then determine whether the maximum number of iterations has been reached. If not, repeat step S2; if so, select the shortest global path among all ants as the optimal path.

[0040] As an improvement to the island ant colony path planning method for unmanned vehicles in this invention:

[0041] The method for converting the potential deadlock mesh into an obstacle mesh is as follows:

[0042] ;

[0043] in, Represents a grid. These represent the row number and column number of the grid, respectively.

[0044] The grid order for path planning is an island point sequence. ,in, The index of the starting grid S. Indicates the index of the endpoint grid E, the island point sequence. The construction method is as follows:

[0045] With islands Center is the center of the circle, All free grids within a search circle of radius are candidate grids; calculate the sum of the Euclidean distances from each candidate grid to the starting grid S and the ending grid E. ,choose The candidate grid with the smallest value becomes the next island. ; with islands The center is the center of the circle. Redefine the search circle and candidate island points based on the radius to locate the islands. This process continues until the search circle contains the endpoint grid E, thus obtaining the island point sequence. .

[0046] As a further improvement to the island ant colony path planning method for unmanned vehicles of the present invention:

[0047] The method for guiding pheromone initialization based on the dominant region is as follows:

[0048] With adjacent islands and islands The rectangular region with the diagonal vertices as the dominant region ,

[0049] Advantageous areas The local initial pheromone concentration is:

[0050]

[0051] in, , and Passing through islands and islands The parameters of the line equation at the center. Indicates the advantageous area Inner adjacent path nodes and The coordinates of the midpoint;

[0052] The non-uniform initial pheromone distribution of the auxiliary map is as follows:

[0053] ;

[0054] in, Indicates the advantageous area Inner slave node To the node Local initialization of pheromone concentration; Represents any region of the auxiliary map from the node To the node The global pheromone concentration; Indicates the pheromone expansion coefficient; and These represent the maximum and minimum pheromone concentrations in all dominant regions, respectively.

[0055] As a further improvement to the island ant colony path planning method for unmanned vehicles of the present invention:

[0056] The grid selection strategy based on the angle factors between the previous node, the current node, and the next node is as follows:

[0057] In time At that time, the Kth ant wants to move from the current grid. Move to the next grid Node transition probability for:

[0058] ;

[0059] in, Indicates in Time and location and location The amount of pheromones between them and These represent the pheromone concentration factor and the expected heuristic factor, respectively. Indicates position The collection of surrounding unvisited grids, It is a heuristic value:

[0060] ;

[0061] in, Indicates the next position Distance from the endpoint grid E Indicates from node To the node Angle coefficient;

[0062]

[0063] in, Indicates the angle coefficient; Indicates the starting point as a node The endpoint is a node. ; Indicates the starting point as a node The endpoint is a node. ; Representing vectors with vector The angle between them.

[0064] As a further improvement to the island ant colony path planning method for unmanned vehicles of the present invention:

[0065] Pheromones update rule two is:

[0066] After all ants have completed one iteration, the pheromones on the entire auxiliary map will be updated using the following formula:

[0067] ;

[0068] ;

[0069] ;

[0070] Where m is the number of ants. Indicates the volatile term. This indicates the addition of a new pheromone item. Indicates a reward or punishment item. This represents the length of the optimal path in the current iteration. Indicates the length of the worst path. To adapt to the evaporation rate, It is an adjustment factor.

[0071] The beneficial effects of this invention are mainly reflected in:

[0072] 1. This invention adopts a segmented (island) strategy. First, the global path is divided into several "island" segments. Ants search independently between each pair of adjacent islands, reducing the complexity of a single search. Then, each local optimum is combined into a global optimum path, which converges faster. In contrast, the commonly used ant algorithm searches the entire path from the starting point to the end point in one go. This results in a large map size, poor pheromone guidance effect, easy getting stuck in local optima, and high computational cost.

[0073] 2. The present invention uses non-uniform initialization of pheromones in the heuristic information between nodes. Pheromones are initialized between adjacent islands, which indirectly divides the large-scale map into several small units. The pheromones between these small units guide the ants to search for paths, shortening the early random exploration process and quickly concentrating pheromones on the better paths, thus accelerating the convergence speed and effectively reducing the number of deadlocked ants. In addition, in order to ensure global search capability, the algorithm also sets a certain value of initial pheromones in areas other than between adjacent islands.

[0074] 3. In addition to dividing the large-scale map into several small units to speed up the convergence speed, this algorithm also adds extra pheromones to the shortest path and reduces extra pheromones to the worst path during the search process, thereby further accelerating the convergence speed.

[0075] 4. This invention improves the accuracy, efficiency, convergence and stability of ant colony algorithm in path planning in complex environments, and has good application prospects in mobile robots, intelligent logistics and other fields.

[0076] 5. This invention preprocesses the original grid map to construct an auxiliary map, effectively avoiding the risk of potentially deadlocked grids being selected as intermediate islands, and significantly reducing the probability of ants getting stuck in deadlock during path planning. Attached Figure Description

[0077] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0078] Figure 1 This is a flowchart illustrating an island ant colony path planning method for an unmanned vehicle, as described in Example 1.

[0079] Figure 2 This is a schematic diagram of the raster map and selected island points;

[0080] Figure 3 This is a schematic diagram of the non-uniform pheromone distribution in Example 1;

[0081] Figure 4 This is a comparison chart of the optimal path planning in a 20×20 grid map using the path planning method of Example 1, the ant colony algorithm, the artificial potential field ant colony optimization algorithm, the improved ant colony algorithm, and the improved adaptive ant colony optimization algorithm.

[0082] Figure 5 This is a comparison chart of the convergence curves of the path planning method in Example 1 with Ant Colony Optimization (ACO), Artificial Potential Field Ant Colony Optimization (APFACO), Improved Ant Colony Optimization (IACO), and Improved Adaptive Ant Colony Optimization (MAACO) in a 20×20 grid map.

[0083] Figure 6 This is a comparison chart of the number of lost ants in a 20×20 grid map between the path planning method of Example 1 and the Ant Colony Optimization Algorithm (ACO), Artificial Potential Field Ant Colony Optimization Algorithm (APFACO), Improved Ant Colony Optimization Algorithm (IACO), and Improved Adaptive Ant Colony Optimization Algorithm (MAACO).

[0084] Figure 7 This is a comparison chart of the optimal path planning method of Example 1 with Ant Colony Algorithm (ACO), Communication Mechanism Enhanced Ant Colony Algorithm (CMEACO), and Improved Ant Colony Algorithm (IACO) in a 30×30 grid map.

[0085] Figure 8 This is a comparison chart of the convergence curves of the path planning method in Example 1 with Ant Colony Algorithm (ACO), Communication Mechanism Enhanced Ant Colony Algorithm (CMEACO), and Improved Ant Colony Algorithm (IACO) in a 30×30 grid map.

[0086] Figure 9 This is a comparison chart of the number of lost ants in a 30×30 grid map between the path planning method of Example 1 and the Ant Colony Algorithm (ACO), the Communication-Enhanced Ant Colony Algorithm (CMEACO), and the Improved Ant Colony Algorithm (IACO).

[0087] Figure 10 This is a schematic diagram of the free grid and its four adjacent grids in Example 2;

[0088] Figure 11 This is a schematic diagram of the island sequence generated based on the auxiliary map in Example 2;

[0089] Figure 12 This is a schematic diagram of the initial pheromone distribution method in the advantageous region in Example 2;

[0090] Figure 13 This is a diagram showing the angle coefficient distribution of the next node in Example 2;

[0091] Figure 14 This is a flowchart of an island ant colony path planning method for an unmanned vehicle in Example 2;

[0092] Figure 15This is a comparison chart of the optimal path planning method of Example 2 with Ant Colony Algorithm (ACO), Improved Ant Colony Algorithm (IACO), and Improved Ant Colony Algorithm with Communication Mechanism (CMEACO) in a 20×20 grid map.

[0093] Figure 16 This is a comparison chart of the convergence curves of the path planning method in Example 2 with Ant Colony Algorithm (ACO), Improved Ant Colony Algorithm (IACO), and Improved Ant Colony Algorithm with Communication Mechanism (CMEACO) in a 20×20 grid map.

[0094] Figure 17 This is a comparison chart of the number of lost ants in a 20×20 grid map between the path planning method of Example 2 and the Ant Colony Algorithm (ACO), the Improved Ant Colony Algorithm (IACO), and the Improved Ant Colony Algorithm with Communication Mechanism (CMEACO).

[0095] Figure 18 This is a comparison chart of the optimal path planning method of Example 2 with Ant Colony Algorithm (ACO), Communication Mechanism Enhanced Ant Colony Algorithm (CMEACO), and Improved Ant Colony Algorithm (IACO) in a 30×30 grid map.

[0096] Figure 19 This is a comparison chart of the convergence curves of the path planning method in Example 2 with Ant Colony Algorithm (ACO), Communication Mechanism Enhanced Ant Colony Algorithm (CMEACO), and Improved Ant Colony Algorithm (IACO) in a 30×30 grid map.

[0097] Figure 20 This is a comparison chart of the number of lost ants in a 30×30 grid map between the path planning method of Example 2 and the Ant Colony Algorithm (ACO), the Communication Mechanism Enhanced Ant Colony Algorithm (CMEACO), and the Improved Ant Colony Algorithm (IACO).

[0098] Figure 21 This is the control block diagram for the unmanned vehicle. Detailed Implementation

[0099] The present invention will be further described below with reference to specific embodiments, but the scope of protection of the present invention is not limited thereto:

[0100] Example 1: A method for island ant colony path planning for unmanned vehicles, such as... Figures 1-3As shown, based on the traditional ant colony algorithm, this paper addresses the problems of getting trapped in local optima, slow convergence, and strong blind search in traditional ant colony path planning. First, several islands are set up on the map, and the global path search is decomposed into path exploration between adjacent islands. Non-uniform pheromone initialization is used between adjacent islands, and a higher pheromone concentration is preset in key areas of the search space to guide ants to concentrate in advantageous areas in the early stages, improving search efficiency. Then, the ant colony algorithm searches the paths between each adjacent island separately. During the search process, a pseudo-random state transition rule combined with a heuristic function and pheromone concentration is used to greedily select or randomly explore with a certain probability. The pheromone evaporation rate is dynamically adjusted according to the number of iterations to maintain global exploration in the early stages and enhance convergence in the later stages. Furthermore, the best path is rewarded during the search process, and the worst path is penalized. The local paths of each island pair are merged to form the global path. The shortest global path is selected from all global paths as the optimal route planned by the robot from the starting point to the end point, passing through all island points.

[0101] Step 1: Improvements to the Antminer Algorithm

[0102] Step 1.1: Generate a raster map and specify island points

[0103] (1) Raster map

[0104] In a real-world environment, an unmanned vehicle's LiDAR is used for environmental scanning, and a grid map is generated using the Gmapping SLAM algorithm. Free grids are represented by 0, while obstacle grids are represented by 1.

[0105] This invention utilizes a Cartesian coordinate system and a numbered representation to define the position of each grid cell. A raster map for modeling a given environment is shown below. Figure 2 As shown, the diagram primarily consists of a 10×10 grid, serving as an example of path planning for a robotic vehicle. Black grids represent obstacle grids, while white grids are free grids that the robot can traverse. In a Cartesian coordinate system, the center of each grid can be uniquely represented by its absolute coordinates (x, y). Using a numbering system, the grids are numbered sequentially from left to right and top to bottom, starting with 1 and incrementing accordingly. Each grid corresponds to a specific number indicating its location.

[0106] The numbers of all grids in the environment form a set as follows:

[0107] (1)

[0108] in, , and These represent the grid number, the total number of rows in the grid, and the total number of columns in the grid, respectively.

[0109] The set of obstacle meshes is as follows:

[0110] (2)

[0111] in, This indicates the total number of obstacle grids.

[0112] The set of free meshes is:

[0113] (3)

[0114] in, This represents the total number of free grid cells.

[0115] The conversion between the coordinates (x, y) of any grid center and the grid number g is calculated using the following three formulas:

[0116] (4)

[0117] (5)

[0118] (6)

[0119] in, and These represent the length and width of the autonomous vehicle, respectively. mod() is the modulo operation, and ceil() is the floor operation.

[0120] Two free grids are selected as the starting grid S and the ending grid E of the robot unmanned vehicle.

[0121] (2) Calculate the island points on the map

[0122] Island points are the intermediate points (i.e., grid points) that the robot must pass through on its path. The total number of island points on the map is N (including the starting grid point S and the ending grid point E, N≥2). Calculate the island selection radius. The formula is:

[0123] (7)

[0124] in, This represents the Euclidean distance between the starting grid S and the ending grid E.

[0125] Treat the starting point S as the first island. The final destination, E, is considered the last island. With the center point of the starting point S as the center, Define a search circle with radius , and the free grid within the search circle represents the next candidate island points to choose from. z represents the number of free grid cells within the search circle, and the sequence of candidate island points. for:

[0126] (8)

[0127] in, This indicates the total number of candidate islands.

[0128] Calculate each candidate island point separately The Euclidean distance to the destination E is used to determine the island, with the grid corresponding to the smallest Euclidean distance being selected. Then, with islands The center is the center of the circle. Using the radius as the base, repeat the above steps to calculate the island's radius. This continues until the final grid E is included. Within a search circle with radius , the island point sequence is formed as follows:

[0129] (9)

[0130] in, It is the Zth island number. Indicates the index of the starting grid S. This indicates the index of the endpoint grid E. Island selection uses the aforementioned Euclidean distances to progressively approach the endpoint, making path construction more directional and reducing invalid searches.

[0131] In this embodiment, the grid order for the unmanned vehicle's path planning is an island point sequence. .

[0132] by Figure 2 Taking a grid as an example, where 3 and 100 represent the starting and target positions respectively. Assuming the map needs to include 5 islands, the final grid order for path planning would be... .

[0133] This embodiment divides the global path into several "island" segments, and the path between each island pair is used as a local path. The best local path is combined into the global best path, resulting in faster convergence.

[0134] Step 1.2: Non-uniform initialization of pheromone distribution

[0135] In the initial stage of the traditional ACO algorithm, all pheromone values ​​in the environment remain constant. In the early stages of pathfinding, the ant colony exhibits a degree of blindness, leading to a slow convergence speed. This embodiment proposes a non-uniform initial pheromone distribution strategy, namely, a favorable region initial pheromone distribution method. Specifically, it specifies "island pairs" (i.e., adjacent islands) as the initial pheromone distribution. and islands The rectangular area defined by the diagonal vertex of the grid is the favorable area. The initial pheromone value of the favorable area is calculated using formula (10). The initial pheromone values ​​of the favorable areas of other adjacent islands are calculated sequentially using formula (10). The minimum pheromone value of the favorable area is then used as the pheromone value of other areas of the global map. The specific implementation is shown in formula (11).

[0136] (10)

[0137] (11)

[0138] in, and These represent the initial pheromone values ​​for the favorable area and the initial pheromone values ​​for the global map, respectively. Both represent grid numbers; The side length of a unit grid; For islands and islands Euclidean distance; For islands and grid Euclidean distance; For grid and grid Euclidean distance; For grid and islands Euclidean distance.

[0139] Using this method Figure 2 The result of non-uniform initialization pheromone value calculation for the raster map is as follows: Figure 3 As shown in the figure, S represents the starting point, E represents the ending point, the red rectangle represents the favorable area, and different colors represent different initial pheromone values. The lighter the yellow, the higher the initial pheromone value, and the darker the blue, the lower the initial pheromone value.

[0140] This embodiment initializes pheromones non-uniformly in the initial stage of the algorithm, assigning higher pheromones to key regions (i.e., "favorable regions between islands") to prioritize the search, reduce randomness, improve early convergence speed, and shorten the early random trial process.

[0141] Step 1.3: Grid Selection Strategy

[0142] Place all ants in the starting grid S, and use a pseudo-random state transition rule to select the position of the next grid. Repeat this process for all adjacent islands. and Local path search between islands: If the ant successfully starts from the starting point S and reaches each island point in sequence. If the ant stops searching at the endpoint grid E, then the ant has successfully completed the path search. Record the ant's path nodes and path length parameters. If the ant does not successfully reach the endpoint or any intermediate island, then the current ant is considered a deadlock ant, the number of deadlock ants is increased by 1, and an infinite value is used as its path length.

[0143] Path construction is performed for each ant sequentially, that is, in time... At that time, the Kth ant wants to move from the current grid. Move to the next grid Node transition probability It can be calculated using the following formula:

[0144] (12)

[0145] in, and These represent the pheromone concentration factor and the expected heuristic factor, respectively. Represents a grid The collection of surrounding unvisited grids, The heuristic function value is calculated using formula (13):

[0146] (13)

[0147] in, It is a scaling factor. and These represent the maximum number of iterations and the current number of iterations, respectively. Represents a grid The Euclidean distance to the endpoint grid E.

[0148] Step 1.4: Global Pheromone Update

[0149] After an ant colony completes one iteration, the pheromones on the original path evaporate at a certain rate, while ants passing through that path deposit new pheromones. After multiple iterations, the path with the highest pheromone concentration is formed, representing the optimal route. This embodiment proposes an improved pheromone update rule (Pheromone Update Rule 1) aimed at enhancing the pheromone accumulation rate on the optimal path while reducing the pheromone accumulation rate on the suboptimal path. Furthermore, a dynamic adaptive evaporation coefficient is introduced. These adjustments improve the algorithm's convergence rate. The specific calculation method is as follows:

[0150] (14)

[0151] (15)

[0152] (16)

[0153] in, This is the constant for the total amount of pheromones. This indicates the volatile term, representing the volatilization of the original pheromone. This is a pheromone addition item, indicating that ants add new pheromones to the path as they pass through it. Indicates a reward or punishment item. Indicates the current iteration loop from the grid To grid The reward or penalty value can be calculated according to formula (15). This process involves adding additional pheromones on the optimal path to enhance the guiding effect of subsequent iterations, or reducing certain pheromones on the worst path to mitigate their misleading effect on subsequent iterations. It is the length of the optimal path in the current iteration. It is the length of the worst path in the current iteration. The adaptive evaporation rate can be calculated using formula (16), where This is the adjustment factor, with a value ranging from 0 to 1. This formula shows that in the initial stage of the algorithm, due to... The value is relatively large. The pheromone value is very small, resulting in minimal differences in pheromone values ​​across paths, thus enhancing the ant colony's global path planning ability. With... The value gradually decreases. As the pheromone value gradually increases, the differences in pheromone values ​​across different paths become larger. This guides more ants to choose the optimal path, enabling the algorithm to achieve faster convergence in later stages.

[0154] Step 2: Use the improved ant algorithm for path planning. The entire process is as follows: Figure 1 As shown.

[0155] S1. The autonomous vehicle generates a grid map, selects the grids it must pass through as island points, and plans the grid order as the island point sequence. ,in, For island serial numbers, The island number is the starting point grid S. Let N be the island index of the endpoint grid E, and N be the total number of islands.

[0156] S2. Distribute non-uniform initial pheromone values ​​on the raster map using the favorable area initial pheromone distribution method.

[0157] S3, Search Path:

[0158] S3.1. All ants are placed at the starting position, and the tabu list is initialized;

[0159] S3.2, A single ant from the island Start by visiting the adjacent islands in sequence. and islands Local path search between islands ;

[0160] S3.2.1 Local Path Search: From Island Starting from this point, the next grid position is selected based on pseudo-random state transition rules, and the tabu table is updated until an island is reached. Obtain the local path;

[0161] S3.2.2 Connect the local paths end to end to form a global planning path, and save the path sequence and shortest path length of a single ant;

[0162] S3.2.3 If an ant fails to reach any of the intermediate islands or the final grid E, then the current ant is considered a deadlock ant, the number of deadlock ants is increased by 1, and the path length becomes an infinite value.

[0163] S3.3, When all the ants reach the island Then, the global pheromone distribution is updated using pheromone update rule one. Next, it is determined whether the maximum number of iterations has been reached. If not, steps S3.1-S3.3 are repeated until the maximum number of iterations is reached. If it is reached, the shortest global path among all ants is selected as the optimal path and output. This optimal path is the optimal route planned by the robot from the starting point to the end point, passing through all island points.

[0164] Experiment 1:

[0165] To verify the effectiveness of the algorithm, the island ant colony path planning method (hereinafter referred to as INACO) for unmanned vehicles in this embodiment is compared with the Ant Colony Optimization Algorithm (ACO), the Artificial Potential Field Ant Colony Optimization Algorithm (APFACO), the Improved Ant Colony Optimization Algorithm (IACO), and the Improved Adaptive Ant Colony Optimization Algorithm (MAACO) in a 20x20 map environment. The initialization parameters are set as follows: Total number of ants Set to 50. The optimal paths generated by the five algorithms are as follows: Figure 4 As shown, the convergence curves of the five algorithms are as follows: Figure 5 As shown, the number of lost ants for the five algorithms is as follows: Figure 6 As shown in Table 1, the performance-related calculation results of the five algorithms are statistically analyzed.

[0166] Table 1. Statistical analysis of performance-related calculation results for five algorithms

[0167] project Average number of iterations Average path length Optimal path length Standard deviation Average number of lost ants ACO 40.50 30.9706 30.9706 0 1112.7 APFACO 7.7 35.6573 35.0711 - 561.2 IACO 11.4 30.97056 30.97056 0 173.7 MAACO 5.3 30.97056 30.97056 0 74.3 INACO 2.7 30.97056 30.97056 0 50.1

[0168] As shown in the table above, the globally optimal path length in this environment is 30.97056, a path length that ACO, IACO, MAACO, and INACO can all achieve. Regarding the number of convergence iterations, INACO has an average of 2.7 iterations, significantly outperforming ACO, APFACO, IACO, and MAACO. Figure 5 As can be seen, the INACO algorithm proposed in this embodiment has the fastest convergence speed. INACO's average number of lost ants is 50.1, which is the smallest among ACO, APFACO, IACO, MAACO, and INACO. This proves that the INACO algorithm proposed in this embodiment can not only quickly find the optimal path, but also effectively reduce the number of deadlock ants.

[0169] Experiment 2:

[0170] To verify the effectiveness of the algorithm, the INACO algorithm proposed in this embodiment is compared with the Ant Colony Algorithm (ACO), the Communication Mechanism Enhanced Ant Colony Algorithm (CMEACO), and the Improved Ant Colony Algorithm (IACO) in a 30x30 map environment. The initialization parameters are set as follows: , , , , , Total number of ants Set to 50. The optimal paths generated by the four algorithms are as follows: Figure 7 As shown, the convergence curves of the four algorithms are as follows: Figure 8 As shown, the number of lost ants for the four algorithms is as follows: Figure 9 As shown in Table 2, the statistical results of the four algorithms are presented.

[0171] Table 2. Statistical analysis of the relevant calculation results of the four algorithms

[0172] project Average number of iterations Average path length Optimal path length Standard deviation Average number of lost ants ACO 50 49.51271 46.2843 1.6564 1987.7 IACO 22.1 45.7768 44.5269 0.5409 387.3 CMEACO 11.7 44.70 44.5269 0.2043 237.70 INACO 5.2 44.5269 44.5269 0 95.2

[0173] As can be seen from the table above, the globally optimal path length in this environment is 44.5269. ACO can no longer find the optimal path within the specified number of iterations. IACO, CMEACO, and INACO can all achieve this path length, but IACO and CMEACO require an average of 11.7 and 22.1 iterations respectively, while INACO requires only 5.2 iterations. Figure 8As can be seen, the INACO algorithm proposed in this embodiment has the fastest convergence speed. INACO's average number of lost ants is 95.2, which is the smallest among ACO, APFACO, IACO, MAACO, and INACO. This proves that the INACO algorithm proposed in this embodiment can not only quickly find the optimal path, but also effectively reduce the number of deadlock ants.

[0174] Example 2: A method for island ant colony path planning for unmanned vehicles, such as... Figures 10-14 As shown.

[0175] The surrounding environment is scanned using LiDAR to construct an initial grid map. Grids that may cause path deadlocks are then identified and marked as obstacles, generating an auxiliary map. A circular search method is used to identify the locations of intermediate islands in the auxiliary map and map them back to the original grid map. An ant is then placed at the starting grid and navigates through each intermediate island sequentially to reach the destination, forming a complete global path, which is then recorded. By comparing all generated paths, the optimal global path is selected.

[0176] Step 1: Improvements to the Antminer Algorithm

[0177] Step 1.1: Generate a raster map and specify island points

[0178] (1) The method for generating the grid map is the same as in Example 1. Two free grids with a value of 0 are selected from the grid map as the starting grid S and the ending grid E of the robot unmanned vehicle.

[0179] (2) Process the raster map into an auxiliary map.

[0180] To prevent deadlocked meshes from being selected as island meshes, the state of the four neighboring meshes surrounding each free mesh must be evaluated. For example... Figure 10 As shown, if the grid ( If two of the four adjacent grids (representing the row and column numbers of the grid) are obstacles, then the free grid is... This could potentially lead to a deadlock. Therefore, this free grid... It will be converted into an obstacle mesh, denoted by the number 1, as shown in Equation (17). This preprocessing step transforms the potential deadlock mesh into obstacles, thereby generating an auxiliary map.

[0181] (17)

[0182] (3) Calculate island points in the auxiliary map.

[0183] Island points are intermediate points that the path must pass through. The total number of island points on the map is N (including the starting point grid S and the ending point grid E, N≥2). The island selection radius is calculated as follows. :

[0184] (18)

[0185] in, This represents the Euclidean distance between the starting grid S and the ending grid E.

[0186] Using the starting grid S as the first island point The center of the starting grid S is the center of the circle. Define a search circle with radius S. All free grids within the search circle are designated as candidate grids. For each candidate grid, calculate the sum of its Euclidean distances to the starting grid S and the ending grid E. The specific calculation formula is shown below:

[0187] (19)

[0188] in, Indicates candidate grids (indicating the number). The center coordinates are ( , From the starting grid S (center coordinates are ()) to the starting grid S (center coordinates are () , )) and the endpoint grid E (center coordinates are ( , The sum of the Euclidean distances of )) This represents the set of candidate grid indices for the search circle. Selecting grids within the search circle... The candidate grid with the smallest value is designated as the next island. With islands The center of the grid is the center of the circle. Redefine the search circle and the set of candidate mesh indices for the radius. Repeat the above steps until the search circle contains the endpoint grid E, thus determining the order of all islands as follows:

[0189] (20)

[0190] in, Indicates the first The serial numbers of the islands.

[0191] In this embodiment, the grid order for the unmanned vehicle's path planning is the island point sequence. .

[0192] by Figure 2 The raster map generates an auxiliary map, which in turn generates an island search map, such as... Figure 11 As shown, the red grid (grid 3) and the cyan grid (grid 100) represent the starting and ending points, respectively. Assuming the map needs to include 5 islands, the remaining island grids (excluding the starting and ending points) are marked in blue. The black obstacle grids marked by the blue rectangles are obstacles added during the conversion of the original map into an auxiliary map. Therefore, the final path planning order is... By generating locally optimal paths between adjacent islands sequentially using path planning methods, and connecting all locally optimal paths from the starting point to the ending point, the globally optimal path for the entire auxiliary map can be obtained.

[0193] Step 1.2: Initialize pheromone distribution non-uniformly.

[0194] In the initial stage of the Ant Colony Algorithm (ACO), pheromone values ​​are uniformly distributed throughout the environment. Therefore, ants exhibit a degree of randomness in their path exploration, leading to a slower convergence speed. To address this issue, a non-uniform initial pheromone distribution strategy is proposed: a dominant region-guided pheromone initialization method. Its computational principle is as follows: Figure 12 As shown.

[0195] Numbered by adjacent islands and The rectangular region with diagonal vertices is defined as the dominant region. According to the island and Determine the equation of the line passing through the center points using the coordinates of the center points:

[0196] (twenty one)

[0197] in , and These represent the relevant parameters of the linear equation, and the rectangular region. Inner adjacent path nodes and midpoint coordinates The calculation formula is shown in formula (22). The local initial pheromone concentration of this area is calculated using formula (23). Subsequently, the local initial pheromone concentrations of other dominant areas in the environment are calculated sequentially. Finally, the non-uniform initial pheromone distribution of the entire map can be calculated using formula (24):

[0198] (twenty two)

[0199] (twenty three)

[0200] (twenty four)

[0201] in, Represents a rectangular area Inner slave node To the node Local initialization of pheromone concentration; Represents any region of the auxiliary map from the node To the node The global pheromone concentration; Indicates the pheromone expansion coefficient; and These represent the maximum and minimum pheromone concentrations in all dominant regions, respectively. Their expressions are as follows:

[0202] (25)

[0203] Step 1.3: Grid selection strategy.

[0204] Proceed to all adjacent islands in sequence and Local path search between islands: If the ant successfully starts from the starting point S and reaches each island point in sequence... If the ant stops searching at the destination E, then the ant has successfully completed the path search. Record the ant's path nodes and path length parameters. If the ant does not successfully reach the destination or any island in between, then the current ant is considered a deadlock ant, the number of deadlock ants is increased by 1, and an infinite value is used as its path length.

[0205] In time At that time, the Kth ant wants to move from the current grid. Move to the next grid The transition probability can be calculated using the following formula. for:

[0206] (26)

[0207] in, Indicates in Time and location and location The amount of pheromones between them and These represent the pheromone concentration factor and the expected heuristic factor, respectively. This represents the set of unvisited grid cells around position i. The heuristic value is calculated using formula (27):

[0208] (27)

[0209] in, Indicates the next position Distance from the finish line This represents the angle selection coefficient.

[0210] From node To the node Angular coefficient distribution as follows Figure 13 As shown, it can be calculated using formula (28).

[0211] (28)

[0212] in, Indicates the angle coefficient; Indicates the starting point as a node The endpoint is a node. Vectors; nodes Represents nodes in the path The previous node; node Indicates the current position; Indicates the starting point as a node The endpoint is a node. Vectors; nodes This indicates the node that will be selected in the path planning process; Representing vectors with vector The angle between them.

[0213] This embodiment adopts a grid selection strategy based on the angle factors between the previous node, the current node, and the next node. The improved heuristic function combines the distance between the current position and the target position, as well as the influence of the path turning angle, to guide the path selection through the angle coefficient and encourage ants to choose smoother paths with more reasonable turns.

[0214] Step 1.4: Global Pheromone Update

[0215] Pheromones on the original path evaporate at a certain rate, while ants passing through the path deposit new pheromones. After multiple iterations, the path with the highest pheromone concentration is formed, representing the optimal route. This paper proposes an improved pheromone update rule (Pheromone Update Rule 2) to enhance the pheromone accumulation rate on the optimal path while reducing the accumulation rate on the suboptimal path. Furthermore, a dynamic adaptive evaporation coefficient is introduced. These adjustments improve the convergence rate of the algorithm. The specific calculation method is as follows:

[0216] (29)

[0217] (30)

[0218] (31)

[0219] After all ants have completed one iteration, the pheromones on the entire map will be updated using formula (29), where m is the number of ants. This indicates the volatile phase, reflecting the volatilization of the original pheromones. This indicates newly added pheromone items, referring to the pheromones added after the ant traverses the path. The reward or penalty term can be calculated according to formula (30). This mechanism enhances the guidance of subsequent iterations by adding extra pheromones on the optimal path, while reducing the pheromone on the worst path to reduce its misleading effect. This represents the length of the optimal path in the current iteration. Indicates the length of the worst path. The adaptive evaporation rate can be calculated using formula (31), where This is an adjustment coefficient, ranging from 0 to 1. As can be seen from this formula, in the early stages of the algorithm, a larger... Leading to smaller The pheromone differences between paths are small, enabling the ant colony to maintain a strong global path planning ability; and as The decrease, As the value of pheromone gradually increases, the difference in pheromone between paths gradually increases, thereby guiding more ants toward the optimal path and promoting faster convergence of the algorithm in the later stages.

[0220] Step 2: Use the improved ant colony algorithm for path planning. The entire process is as follows: Figure 14 As shown.

[0221] S1. The unmanned vehicle generates a grid map and converts potential deadlock grids into obstacle grids to generate an auxiliary map;

[0222] S2. Select the grid points that must be traversed from the auxiliary map as island points. The grid order for path planning is the island point sequence. ,in, The island's serial number. The island number is the starting point grid S. N represents the island index of the endpoint grid E, and N is the total number of islands;

[0223] S3. A non-uniform initial pheromone value distribution is applied to the auxiliary map using a pheromone initialization method based on advantageous regions:

[0224] S4, Search Path:

[0225] S4.1. All ants are placed at the starting position, and the tabu list is initialized;

[0226] S4.2, A single ant from the island Start by visiting the adjacent islands in sequence. and Local path search between islands ;

[0227] S4.2.1 Local Path Search: From Island Starting from the beginning, a grid selection strategy based on the angle factors between the previous node, the current node, and the next node is used to select the next grid position, and the tabu table is updated until an island is reached. Obtain the local path;

[0228] S4.2.2 Connect the local paths end to end to form a global planning path, and record the length and grid numbers traversed;

[0229] S4.2.3 If an ant fails to reach the destination grid E or any island in between, then the current ant is considered a deadlock ant, the number of deadlock ants is increased by 1, and the path length becomes an infinite value.

[0230] S4.3 When all the ants reach the island Then, the global pheromone distribution is updated using pheromone update rule two. Then, it is determined whether the maximum number of iterations has been reached. If not, steps S4.1-S4.3 are repeated until the maximum number of iterations is reached. If it is reached, the shortest global path among all ants is selected as the optimal path.

[0231] Experiment 3:

[0232] To verify the effectiveness of the algorithm, the island ant colony path planning method (hereinafter referred to as EACI) for an unmanned vehicle in this embodiment is compared with the Ant Colony Algorithm (ACO), the Improved Ant Colony Algorithm (IACO), and the Improved Ant Colony Algorithm with Communication Mechanism (CMEACO) in a 20x20 map environment. The initialization parameters are set as follows: , , , , , , , and The algorithm generates an optimal path, for example... Figure 15 As shown, all four algorithms successfully captured the path from the starting point to the ending point. The convergence curves of the optimal path for the four algorithms are shown in the figure. Figure 16 As shown in the figure, the comparison curves of the number of lost ants for the four algorithms are as follows: Figure 17 As shown in Table 3, the statistical results of the four algorithms are presented.

[0233] Table 3. Statistical analysis of the relevant calculation results of the four algorithms

[0234] project Average number of iterations Average path length Optimal path length Average number of lost ants Total turning angle (°) ACO 40.5 30.97 30.97 1246.0 675 IACO 10.5 30.97 30.97 172.30 675 CMEACO 6.4 30.97 30.97 165.10 585 EACI 1 30.97 30.97 9.85 675

[0235] Of the four algorithms, EACI exhibits the fastest convergence speed; however, its total rotation angle is 675°, the same as ACO and IACO, and slightly larger than CMEACO. Figure 17 As shown, the number of deadlock ants decreases with increasing iterations in all algorithms. Notably, EACI generates the fewest deadlock ants and decreases the fastest, further highlighting its efficiency in path optimization. Table 3 shows that all algorithms successfully found the shortest path of length 30.97. Most notably, EACI exhibits the fewest average deadlock ants and the lowest average number of iterations among the four algorithms, at 9.85 (9.85 < 165.10 < 172.3 < 1246.0) and 1 (1 < 6.4 < 10.5 < 40.5), respectively. Specific data are shown in the table. Overall, the experimental results validate the excellent performance of EACI.

[0236] Experiment 4:

[0237] To verify the effectiveness of the algorithm, the island ant colony path planning method (hereinafter referred to as EACI) for an unmanned vehicle in this embodiment is compared with the Ant Colony Algorithm (ACO), the Communication Mechanism Enhanced Ant Colony Algorithm (CMEACO), and the Improved Ant Colony Algorithm (IACO) in a 30x30 map environment. The initialization parameters are set as follows: , , , , , , , and Comparison of the optimal production path Figure 18 As shown, all four algorithms successfully found the path from the starting point to the ending point. The convergence curves of the four algorithms are shown in the figure. Figure 19 As shown in the figure, the comparison curves of the number of lost ants for the four algorithms are as follows: Figure 20 As shown in Table 4, the statistical results of the four algorithms are presented.

[0238] Table 4. Statistical analysis of the results of the four algorithms in a 30x30 map environment.

[0239] project Average number of iterations Average path length Optimal path length Average number of lost ants Total turning angle (°) ACO 77.10 45.89 45.36 3112.40 675 IACO 21.90 45.22 44.53 415.30 810 CMEACO 11.7 44.70 44.53 237.70 630 EACI 1.4 44.53 44.53 27.5 765

[0240] The results show that EACI achieves the shortest path length and the fastest convergence speed among all algorithms. Furthermore, as... Figure 20As shown, the number of deadlock ants gradually decreases with increasing iteration count across all algorithms. Notably, EACI generates the fewest deadlock ants and decreases the fastest. According to the experimental results in Table 4, the shortest path length generated by EACI is 44.53, consistent with IACO and CMEACO, but 0.83 shorter than ACO. Furthermore, the average path length of EACI is also 44.53, similar to CMEACO, but 0.69 and 1.36 shorter than IACO and ACO, respectively. The two parameters showing the most significant changes in Table 4 are the average number of iterations and the average number of lost ants. EACI's average number of iterations is 1.4, approximately 1 / 11 of CMEACO, 1 / 21 of IACO, and 1 / 77 of ACO. Regarding the average number of lost ants, EACI's value is 27.5, 210.2 fewer than CMEACO, 384.3 fewer than IACO, and 3084.9 fewer than ACO. The total turning angle of the path is 765°, slightly higher than CMEACO and ACO, but lower than IACO. This observation indicates that EACI still has room for improvement in optimizing the total turning angle. Overall, the above data fully demonstrates the superior performance exhibited by EACI.

[0241] Example 3:

[0242] The unmanned vehicle is equipped with an industrial computer. Control handles, multi-line LiDAR, 2D LiDAR, depth cameras, inertial measurement units (IMUs), motor drivers, and encoders are all connected to the industrial computer via signal transmission. Figure 21 As shown, environmental information is acquired through multi-line LiDAR and 2D LiDAR to model an environmental grid map. While navigating along the planned path, the two LiDARs can detect obstacles in real time. A depth camera, working in conjunction with the multi-line and 2D LiDARs, detects obstacles. If an obstacle is encountered, the autonomous vehicle stops until it moves away before resuming its navigation. During movement, the autonomous vehicle is primarily powered by a battery, controlling the left and right motors via motor drivers. The left and right motors are equipped with left and right encoders respectively, allowing for speed measurement. The inertial measurement unit (IMU) is a combination of accelerometers, gyroscopes, and a digital motion processor; the acquired data is used to calculate odometer readings and other data.

[0243] The working process of the unmanned vehicle:

[0244] (1) Start the unmanned vehicle. The staff can use the control handle to control the movement of the vehicle. The multi-line lidar and 2D lidar collect information and generate a two-dimensional grid map.

[0245] (2) After the grid map is established, the unmanned vehicle does not need to be controlled by staff. Only a starting grid S and an ending grid E are needed for the unmanned vehicle. The unmanned vehicle will use the optimal path obtained by the path planning method of island ant colony in Example 1 or Example 2 as the driving route.

[0246] (3) Control the movement of the left and right motors to travel along the planned route. The left and right encoders measure the speed of the left and right motors respectively. With the data from the inertial measurement unit (IMU), the unmanned vehicle calculates the distance traveled and its current position on the grid map.

[0247] (4) During the journey, the unmanned vehicle collects information about the surrounding environment in real time through multi-line lidar, 2D lidar and depth camera. When it encounters a newly generated obstacle (an obstacle not present in the two-dimensional grid map established in step (1)), it stops moving until the obstacle is moved away. Then the vehicle continues to move along the navigation path until it reaches the endpoint grid E.

[0248] Finally, it should be noted that the above examples are merely some specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments and many variations are possible. All variations that can be directly derived or conceived by those skilled in the art from the disclosure of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A method for island ant colony path planning for unmanned vehicles, characterized in that... The path planning process for autonomous vehicles includes the following steps: S1. Select the grid points that must be passed through in the grid map as island points, and then distribute the initial pheromone values ​​of the grid map in a non-uniform manner using the favorable area initial pheromone distribution method. S2. Each ant searches the path between adjacent islands in turn. During the search, it selects the next grid position through pseudo-random state transition rules until it reaches the target island and obtains a local path. Connect the local paths end to end to form the global planning path. S3. After all ants have reached the endpoint grid, update the global pheromone distribution according to pheromone update rule one; determine whether the maximum number of iterations has been reached. If not, repeat step S2. If it has been reached, select the shortest global planning path among all ants as the optimal path.

2. The island ant colony path planning method for an unmanned vehicle according to claim 1, characterized in that: The grid order for path planning is an island point sequence. ,in, Indicates the index of the starting grid S. The sequence number represents the endpoint grid E, and N is the total number of islands; island point sequence The construction method is as follows: With islands Center is the center of the circle, radius All free grid points within the search circle are candidate island points. Each candidate island point The grid with the smallest Euclidean distance to the endpoint grid E is the island. Then with islands Center is the center of the circle, radius Redefine the search circle and candidate island points to locate the islands. This process continues until the search circle contains the endpoint grid E, thus obtaining the island point sequence. ;radius The calculation formula is: ； in, This represents the Euclidean distance between the starting grid S and the ending grid E.

3. The island ant colony path planning method for an unmanned vehicle according to claim 2, characterized in that: The method for initial pheromone distribution in the favorable region is as follows: With adjacent islands and islands The rectangular region defined by the diagonal vertices is the favorable region, and the initial pheromone value is: ； in, This represents the initial pheromone value of the favorable region. This represents the initial pheromone value of the global map. These represent the grid numbers respectively; ； in, The side length of a unit grid; For islands and islands Euclidean distance; For islands and grid Euclidean distance; For grid and grid Euclidean distance; For grid and islands Euclidean distance.

4. The island ant colony path planning method for an unmanned vehicle according to claim 3, characterized in that: The pseudo-random state transition rule is as follows: In time At that time, the Kth ant wants to move from the current grid. Move to the next grid Node transition probability for: ； in, and These represent the pheromone concentration factor and the expected heuristic factor, respectively. Represents a grid The collection of surrounding unvisited grids, For heuristic function values: ； in, It is a scaling factor. and These represent the maximum number of iterations and the current number of iterations, respectively. Represents a grid The Euclidean distance to the endpoint grid E.

5. The island ant colony path planning method for an unmanned vehicle according to claim 4, characterized in that: The first pheromone update rule is: ； in, This indicates the evaporation of the original pheromone. This means that when ants pass through a path, they add new pheromones to the path. Indicates a reward or punishment item. Indicates adaptive evaporation rate, Indicates the current iteration loop from the grid To grid Reward or penalty value: ； in, The total amount of pheromones is a constant. It is the length of the optimal path in the current iteration. It is the length of the worst path in the current iteration; ； in, It is an adjustment factor.

6. A method for island ant colony path planning for an unmanned vehicle, characterized in that... The path planning process for autonomous vehicles includes the following steps: S1. In the grid map, potential deadlock grids are converted into obstacle grids to generate an auxiliary map. Then, grids that must be passed through are selected from the auxiliary map as island points. Then, the auxiliary map is subjected to non-uniform initial pheromone value distribution by a pheromone initialization method based on the dominant region. S2. Each ant searches the path between adjacent islands in turn. During the search, it selects the next grid position based on the angle factors between the previous node, the current node and the next node, until it reaches the target island point and obtains a local path. Connect the local paths end to end to form the global planning path. S3. After all ants have reached the endpoint grid, update the global pheromone distribution according to pheromone update rule two, and then determine whether the maximum number of iterations has been reached. If not, repeat step S2; if so, select the shortest global path among all ants as the optimal path.

7. The island ant colony path planning method for an unmanned vehicle according to claim 6, characterized in that: The method for converting the potential deadlock mesh into an obstacle mesh is as follows: ； in, Represents a grid. These represent the row number and column number of the grid, respectively. The grid order for path planning is an island point sequence. ,in, The index of the starting grid S. Indicates the index of the endpoint grid E, the island point sequence. The construction method is as follows: With islands Center is the center of the circle, All free grids within a search circle of radius are candidate grids; calculate the sum of the Euclidean distances from each candidate grid to the starting grid S and the ending grid E. ,choose The candidate grid with the smallest value becomes the next island. ; with islands The center is the center of the circle. Redefine the search circle and candidate island points based on the radius to locate the islands. This process continues until the search circle contains the endpoint grid E, thus obtaining the island point sequence. .

8. The island ant colony path planning method for an unmanned vehicle according to claim 7, characterized in that: The method for guiding pheromone initialization based on the dominant region is as follows: With adjacent islands and islands The rectangular region with the diagonal vertices as the dominant region , Advantageous areas The local initial pheromone concentration is: in, , and Passing through islands and islands The parameters of the line equation at the center. Indicates the advantageous area Inner adjacent path nodes and The coordinates of the midpoint; The non-uniform initial pheromone distribution of the auxiliary map is as follows: ； in, Indicates the advantageous area Inner slave node To the node Local initialization of pheromone concentration; Represents any region of the auxiliary map from the node To the node The global pheromone concentration; Indicates the pheromone expansion coefficient; and These represent the maximum and minimum pheromone concentrations in all dominant regions, respectively.

9. The island ant colony path planning method for an unmanned vehicle according to claim 8, characterized in that: The grid selection strategy based on the angle factors between the previous node, the current node, and the next node is as follows: In time At that time, the Kth ant wants to move from the current grid. Move to the next grid Node transition probability for: ； in, Indicates in Time and location and location The amount of pheromones between them and These represent the pheromone concentration factor and the expected heuristic factor, respectively. Indicates position The collection of surrounding unvisited grids, It is a heuristic value: ； in, Indicates the next position Distance from the endpoint grid E Indicates from node To the node Angle coefficient; in, Indicates the angle coefficient; Indicates the starting point as a node The endpoint is a node. ; Indicates the starting point as a node The endpoint is a node. ; Representing vectors with vector The angle between them.

10. The island ant colony path planning method for an unmanned vehicle according to claim 9, characterized in that: Pheromones update rule two is: After all ants have completed one iteration, the pheromones on the entire auxiliary map will be updated using the following formula: ； ； ； Where m is the number of ants. Indicates the volatile term. This indicates the addition of a new pheromone item. Indicates a reward or punishment item. This represents the length of the optimal path in the current iteration. Indicates the length of the worst path. To adapt to the evaporation rate, It is an adjustment factor.

Images