Air-ground robot mountainous area path planning method based on multi-strategy improved ant colony algorithm

Abstract

The invention discloses an air-ground robot mountainous area path planning method based on a multi-strategy improved ant colony algorithm. The method comprises the following steps: establishing a mountainous area three-dimensional environment model; an air-ground robot is initialized; non-uniform initial pheromones in normal distribution are generated; designing a heuristic function considering the air-ground bimodal and the motion direction; constructing a state transition probability which simultaneously considers motion smoothness, safety and grid importance; perfecting a pheromone updating mechanism; and a multi-strategy improved ant colony algorithm is adopted to complete path planning and obtain an optimal path. By using the method, the planning efficiency of the air-ground robot can be effectively improved, and the feasibility and practicability of path planning are enhanced. The open space robot mountainous area path planning method based on the multi-strategy improved ant colony algorithm can be widely applied to the technical field of robot path planning.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of robot path planning, in particular to a mountain path planning method for air-ground robots based on a multi-strategy improved ant colony algorithm. BACKGROUND

[0002] Air-ground robots have the ability to freely switch between air and ground motion modes, and are an important direction of multi-modal robot research. In terms of function, air-ground robots can independently complete exploration tasks in complex composite environments such as the sky, ground, canyon, and pit. In terms of structure, air-ground robots usually use adjustable active nodes, which can adaptively adjust the motion configuration according to environmental changes, making the overall structure more compact than traditional single-modal platforms such as unmanned vehicles, unmanned aerial vehicles, or unmanned underwater vehicles. In terms of energy consumption, air-ground robots prefer to use low-power ground driving mode in unnecessary scenarios, and have significantly better endurance performance than pure flight platforms. However, the multi-modal motion characteristics also lead to higher complexity in path planning for air-ground robots.

[0003] In terms of planning methods, the path planning system of air-ground robots generally includes both hybrid land-air global path planning and hybrid land-air local path planning modules. Air-ground robots often need to perform global planning based on known multi-modal environmental information, with the optimization goal of the shortest hybrid path, the minimum total time, or the lowest total energy consumption, to generate the overall optimal route from the starting point to the end point. However, in real-world applications, especially in mountainous regions with large terrain undulations and significant obstructions, it is often difficult to obtain complete global environmental information. At the same time, the presence of dynamic three-dimensional obstacles may also cause the pre-generated global path to be partially invalid. Therefore, the system needs to rely on multi-source sensors or artificial feedback to update local environmental information to ensure the reliability and accuracy of the planning results.

[0004] Currently, global path planning for air-ground robots mainly includes graph search-based, sampling-based, and bio-inspired planning methods. Among them, graph search and sampling methods often have large computational overhead and limited planning speed in multi-modal mountainous environments; in contrast, bio-inspired methods have the advantages of light computational burden and fast convergence speed while maintaining global optimization ability. However, traditional ant colony algorithms, sparrow search algorithms, and genetic algorithms generally do not fully consider factors such as kinematic constraints, multi-modal characteristics, path smoothness, safety boundaries, and task importance of air-ground robots, making it difficult to generate hybrid optimal paths that meet the needs of real-world scenarios. On the other hand, large language models have shown outstanding advantages in complex reasoning and strategic decision-making, and their combination with traditional bio-inspired algorithms can significantly improve the search efficiency of heuristic algorithms and the usability of path results, providing more space for the development of air-ground robot path planning methods. SUMMARY

[0005] 1. Purpose of the invention:

[0006] To address the problems mentioned above, the present invention aims to provide a mountain path planning method for air-to-ground robots based on a multi-strategy improved ant colony algorithm. This method incorporates the mode switching of air-to-ground robots into the planning framework based on the ant colony algorithm and utilizes a strategy that combines large language models with mechanisms to incorporate path smoothness, safety, and fuzzy importance into the evaluation system, thereby improving the feasibility, stability, and practical application value of the planned path.

[0007] 2. Technical Solution:

[0008] To achieve the above objectives, this invention provides a method for mountain path planning for air-to-ground robots based on a multi-strategy improved ant colony algorithm, comprising the following steps:

[0009] Step 1: Simplify all three-dimensional obstacles in the area to be planned into a single-peak exponential function, and establish a three-dimensional environment model of the mountainous area by combining layering and grid plane methods;

[0010] Step 2: Import the completed 3D environment model into the global planning module of the dual-modal air-ground robot, and set the starting point S and the ending point E of the planning;

[0011] Step 3: Connect the starting point S and the ending point E in the 3D environment, and generate a normally distributed non-uniform initial pheromone y based on the distance from each grid cell to the line segment SE. u , where u is the index number of the raster;

[0012] Step 4: Design a heuristic function η that considers the bimodal characteristics and motion direction of the air-to-ground robot. ij ;

[0013] Step 5: Considering the smoothness and safety of the aerial robot's movement process, as well as the importance of the path grid, the state transition probability of ant l in the ant colony is constructed as follows:

[0014]

[0015] Where l∈{1,2,…,m} represents the index number of an individual ant in the ant colony, and the ant colony size is m. τ represents the probability of ant l transitioning from grid i to grid j during the g-th iteration. ij (g) represents the current pheromone concentration on edge ij at the g-th iteration, S mij (g), S ej (g) and S j(g) represents the smoothness of the ant's movement from grid i to grid j in the g-th iteration, the environmental safety and importance of grid j, respectively; α, β, λ, and ω represent the pheromone factor, heuristic information factor, turning penalty factor, and safety penalty factor, respectively. l Let be the set of candidate neighbor nodes that ant l can access at its current position;

[0016] Step 6: Considering the bimodal characteristics of the air-to-ground robot, the pheromone update process is designed as follows:

[0017]

[0018] Where τ(g+1) represents the updated pheromone concentration, ρ∈(0,1) represents the pheromone evaporation factor, and Δτ ij (g) represents the basic pheromone released by all ants on edge ij in this round. and denoted as reinforcing pheromones contributed by high-quality paths and penalizing pheromones introduced by low-quality paths, respectively; κ represents the high-energy-consuming mode penalty factor.

[0019] Step 7: The air-ground robot generates the optimal path scheme in the area to be planned based on the multi-strategy improved ant colony algorithm, and travels from the starting point S to the destination E.

[0020] Furthermore, in step 1, the unimodal exponential function is:

[0021]

[0022] Where x and y represent the horizontal and vertical coordinates of the two-dimensional plane of the Earth's surface, respectively; z(x,y) represents the height of the three-dimensional obstacle at (x,y) on the Earth's surface; q is the index of the three-dimensional obstacle; n is the total number of three-dimensional obstacles; and h... q To control the terrain parameters of altitude, e() is a natural exponential function, x q and y q Let x and y represent the horizontal and vertical ground coordinates of the center of the three-dimensional obstacle q, respectively. sq and y sq These represent the attenuation of the three-dimensional obstacle q along the x-axis and y-axis, respectively.

[0023] Furthermore, in step 2, the dual-modal air-to-ground robot includes two modes: high-energy flight and relatively low-energy ground movement.

[0024] Furthermore, in step 3, the initial pheromone is:

[0025]

[0026] Among them, y uLet x be the initial pheromone of the u-th grid, a and b be the translation and scaling constants respectively, σ be the standard deviation of the normal distribution, and x be the pheromone of the u-th grid. u Let be the vertical distance from the u-th grid cell to line segment SE.

[0027] Furthermore, in step 4, the heuristic function η that considers the dual-modal characteristics and motion direction of the air-to-ground robot... ij for:

[0028]

[0029] Where, d ij +d j Ed i E is a length relationship derived from the triangle inequality in triangle ΔijE, used to determine the direction of motion. Let $\mathbf{i}$ represent the distances from grid $i$ to grid $j$, from grid $j$ to the endpoint $E$, and from grid $i$ to the endpoint $E$, respectively. $c = 0.01$ is a constant to prevent the denominator from being zero. $Energy$ ij This represents the energy consumption from grid i to grid j:

[0030]

[0031] Where, χ and These represent the unit energy consumption of the air-to-ground robot in flight mode and ground movement mode, respectively.

[0032] ψ represents the dynamic adjustment factor:

[0033]

[0034] Among them, G max This represents the maximum number of iterations.

[0035] Furthermore, in step 5, the smoothness S of the motion from grid i to grid j... mij (g) is:

[0036]

[0037] in, The angle between the path segment connecting the previous path grid (x1,y1,z1) and the current grid i (x2,y2,z2) and the path segment formed by the current grid i and the next potential path grid (x3,y3,z3);

[0038] The environmental safety S of the grid j ej (g) is:

[0039]

[0040] in, is the Euclidean distance from grid j to the nearest obstacle o, and maxSe is the maximum safe distance among all grids used for normalization;

[0041] The importance of the grid j is S j (g) is:

[0042] S j (g)=LLM(Prompt,Human_assessment,P)

[0043] Where LLM() represents the output of the large language model, Prompt is the preset prompt word, including the problem background, task requirements, output format, and S j (g) Range, Human_assessment represents the importance evaluation of grid j by experts, used to instruct the air-ground robot to collect key grid data synchronously during the planning process, P is the key parameter of the large language model, including top_p, top_k, Temperature, Max_Tokens, etc.

[0044] Furthermore, in step 5, the set of candidate neighbor nodes is allowed. l In flight mode, there are 6 grid cells in total: front, back, left, right, up, and down. In ground mode, there are 4 grid cells in total: front, back, left, and right.

[0045] Furthermore, in step 6, the aforementioned for:

[0046]

[0047] Where Q is the pheromone intensity constant, σ1 and σ2 are two weighting factors, and J min J is the multi-objective comprehensive evaluation value of the optimal path in the current iteration. max The bestJ value is the multi-objective comprehensive evaluation value of the worst path in the current iteration, while the bestJ value is the set of high-quality paths.

[0048] The aforementioned for:

[0049]

[0050] WorstJ is a set of low-quality paths;

[0051] Furthermore, in step 6, the pattern penalty factor κ is:

[0052]

[0053] 3. The present invention provides a mountain path planning method for air-to-ground robots based on a multi-strategy improved ant colony algorithm, which has the following advancements and advantages compared to existing technologies:

[0054] (1) Non-uniform initial pheromone distribution is performed based on the vertical distance of each grid to the starting and ending points SE, which accelerates the convergence speed of the algorithm in the early stage.

[0055] (2) The energy consumption characteristics and motion direction characteristics of the air-to-ground robot in different modes are simultaneously introduced into the heuristic function η. ij This effectively reduces the probability of the algorithm generating a path that deviates from the destination while ensuring the lowest possible energy consumption, thereby improving planning efficiency.

[0056] (3) Use large language models to evaluate the relative importance of grids and enhance the focus and rationality of the grid selection process.

[0057] (4) The construction of state transition probability takes into account motion directionality, path smoothness, environmental safety and grid importance, which improves the accuracy of state transition judgment and ensures the practicality of the final output path.

[0058] (5) The pheromone update process takes into account the energy consumption differences of the multimodal air-ground robots, avoids dependence on high-energy-consuming modes, and improves the endurance of air-ground robots.

[0059] In summary, this method is a highly efficient and practical global path planning method for air-to-ground robots. Attached Figure Description

[0060] Figure 1 This is a flowchart of the steps of the air-ground robot mountain path planning method based on multi-strategy improved ant colony algorithm provided by the present invention;

[0061] Figure 2 This is a 3D oblique view of the path of an air-ground robot based on a multi-strategy improved ant colony algorithm for mountainous terrain path planning, provided by the present invention.

[0062] Figure 3 This is a top-view view of the three-dimensional path of the air-ground robot, which is based on the multi-strategy improved ant colony algorithm for mountainous path planning of air-ground robots provided by the present invention.

[0063] Figure 4 This is a side view of the three-dimensional path of an air-ground robot in the positive Y direction, based on the air-ground robot mountain path planning method based on the multi-strategy improved ant colony algorithm provided by the present invention.

[0064] Figure 5 This is a side view of the three-dimensional path of the air-ground robot in the X direction, based on the multi-strategy improved ant colony algorithm for mountainous path planning of air-ground robots provided by the present invention. Detailed Implementation

[0065] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and specific examples:

[0066] Figure 1 This is a flowchart illustrating the steps of the mountain path planning method for air-to-ground robots based on a multi-strategy improved ant colony algorithm described in this invention. The specific implementation steps of this method include:

[0067] S1. Simplify all three-dimensional obstacles in the area to be planned into a single-peak exponential function, and establish a three-dimensional environment model of the mountainous area by combining layering and grid plane methods;

[0068] Furthermore, in step 1, the unimodal exponential function is:

[0069]

[0070] Where x and y represent the horizontal and vertical coordinates of the two-dimensional plane of the Earth's surface, respectively; z(x,y) represents the height of the three-dimensional obstacle at (x,y) on the Earth's surface; q is the index of the three-dimensional obstacle; n = 6 is the total number of three-dimensional obstacles; (h q＝1 ,h q＝2 ,h q＝3 ,h q＝4 ,h q＝5 ,h q＝6 )=(80,70,90,85,60,65) represents the terrain parameters, e() is the natural exponential function, (x q＝1 ,x q＝2 ,x q＝3 ,x q＝4 ,x q＝5 ,x q＝6 ) = (20, 35, 55, 80, 25, 75) and (y q＝1 ,y q＝2 ,y q＝3 ,y q＝4 ,y q＝5 ,y q＝6 )=(20,75,40,80,55,15) represent the horizontal and vertical ground coordinates of the center of different three-dimensional obstacles, respectively, and (8,8),(10,10),(9,9),(11,11),(7,7),(8,8) are the attenuation amounts of different three-dimensional obstacles along the x-axis and y-axis directions for q from 1 to 6.

[0071] Step 2: Import the completed 3D environment model into the global planning module of the dual-modal air-ground robot, and set the starting point S(100,5) and the ending point E(0,95) of the planning;

[0072] Furthermore, in step 2, the dual-modal air-to-ground robot includes two modes: high-energy flight and relatively low-energy ground movement.

[0073] Step 3: Connect the starting point S and the ending point E in the 3D environment, and generate a normally distributed non-uniform initial pheromone y based on the distance from each grid cell to the line segment SE. u , where u is the index number of the raster;

[0074] Furthermore, in step 3, the initial pheromone is:

[0075]

[0076] Among them, y u Let x be the initial pheromone of the u-th grid, a = 3 and b = 625, which are the translation and scaling constants, respectively, σ be the standard deviation of the normal distribution, and x be the pheromone of the u-th grid. u Let be the vertical distance from the u-th grid cell to line segment SE.

[0077] Step 4: Design a heuristic function η that considers the bimodal characteristics and motion direction of the air-to-ground robot. ij ;

[0078] Furthermore, in step 4, the heuristic function η that considers the dual-modal characteristics and motion direction of the air-to-ground robot... ij for:

[0079]

[0080] Where, d ij +d j Ed i E is a length relationship derived from the triangle inequality in triangle ΔijE, used to determine the direction of motion. Let $\mathbf{i}$ represent the distances from grid $i$ to grid $j$, from grid $j$ to the endpoint $E$, and from grid $i$ to the endpoint $E$, respectively. $c = 0.01$ is a constant to prevent the denominator from being zero. $Energy$ ij This represents the energy consumption from grid i to grid j:

[0081]

[0082] Where χ = 1 and ψ represents the unit energy consumption of the air-to-ground robot in flight mode and ground movement mode, respectively; ψ represents the dynamic adjustment factor.

[0083]

[0084] Among them, G max =300 is the maximum number of iterations.

[0085] Step 5: Considering the smoothness and safety of the aerial robot's movement process, as well as the importance of the path grid, the state transition probability of ant l in the ant colony is constructed as follows:

[0086]

[0087] Where l∈{1,2,…,m} represents the index number of an individual ant in the ant colony, and the ant colony size is m=30. τ represents the probability of ant l transitioning from grid i to grid j during the g-th iteration. ij (g) represents the current pheromone concentration on edge ij at the g-th iteration, S mij (g), S ej (g) and S j (g) represents the smoothness of the ant's movement from grid i to grid j in the g-th iteration, the environmental safety and importance of grid j, respectively; α = 1.05, β = 2.10, λ = 1.55 and ω = 1.60 represent the pheromone factor, heuristic information factor, turning penalty factor and safety penalty factor, respectively. l Let be the set of candidate neighbor nodes that ant l can access at its current position;

[0088] Furthermore, in step 5, the smoothness S of the motion from grid i to grid j... mij (g) is:

[0089]

[0090] in, The angle between the path segment connecting the previous path grid (x1,y1,z1) and the current grid i (x2,y2,z2) and the path segment formed by the current grid i and the next potential path grid (x3,y3,z3);

[0091] The environmental safety S of the grid j ej (g) is:

[0092]

[0093] in, Let be the Euclidean distance from grid j to the nearest obstacle o, and maxSe = 27.24 be the maximum safe distance among all grids used for normalization;

[0094] The importance of the grid j is S j (g) is:

[0095] S j (g)=LLM(Prompt,Human_assessment,P)

[0096] Where LLM() represents the output of the large language model, Prompt is the preset prompt word, including the problem background, task requirements, output format, and S j (g) The output range is 0 to 1. Human_assessment represents the importance evaluation of grid j by experts, which is used to instruct the air-ground robot to collect key grid data synchronously during the planning process. P is the key parameter of the large language model, including top_p, top_k, Temperature, Max_Tokens, etc.

[0097] Furthermore, in step 5, the set of candidate neighbor nodes is allowed. l In flight mode, there are 6 grid cells in total: front, back, left, right, up, and down. In ground mode, there are 4 grid cells in total: front, back, left, and right.

[0098] Step 6: Considering the bimodal characteristics of the air-to-ground robot, the pheromone update process is designed as follows:

[0099]

[0100] Where τ(g+1) represents the updated pheromone concentration, ρ=0.5 represents the pheromone evaporation factor, and Δτ ij (g) represents the basic pheromone released by all ants on edge ij in this round. and denoted as reinforcing pheromones contributed by high-quality paths and penalizing pheromones introduced by low-quality paths, respectively; κ represents the high-energy-consuming mode penalty factor.

[0101] Furthermore, in step 6, the aforementioned for:

[0102]

[0103] Where Q = 0.98 is the pheromone intensity constant, σ1 = 0.5 and σ2 = 0.5 are two weighting factors, and J min J is the multi-objective comprehensive evaluation value of the optimal path in the current iteration. max J represents the multi-objective comprehensive evaluation value of the worst path in the current iteration, while BestJ is the set of high-quality paths that includes all paths with a multi-objective evaluation value of J. min All edges along the path;

[0104] The aforementioned for:

[0105]

[0106] WorstJ is the set of low-quality paths, which includes all paths with a multi-objective evaluation value of J. max All edges along the path;

[0107] Furthermore, in step 6, the pattern penalty factor κ is:

[0108]

[0109] Step 7: As Figure 3 , Figure 4 and Figure 5 As shown, the air-to-ground robot generates the optimal path scheme in the area to be planned based on the multi-strategy improved ant colony algorithm, and travels from the starting point S to the destination E.

[0110] This invention has been described through embodiments based on a conceptual framework, but is not limited to the above embodiments. Equivalent modifications or substitutions made by those skilled in the art without departing from the scope of the claims are all within the scope of protection of this invention, and the scope of protection claimed in this invention should be determined by the appended claims.

Claims

1. A method for air-ground robot path planning in mountainous areas based on a multi-strategy improved ant colony algorithm, characterized in that, The method comprises the following steps: Step 1: all three-dimensional obstacles in the region to be planned are simplified into unimodal exponential functions, and a three-dimensional environment model of the mountainous area is established by combining the method of layering with the grid plane; Step 2: the three-dimensional environment model established is imported into the global planning module of the dual-mode air-ground robot, and the starting point S and the ending point E of planning are set; Step 3: connecting the start point S and the end point E in the three-dimensional environment, and generating the initial pheromone y according to the distance from each grid center to the line segment SE u where u is the index number of the grid; Step 4: Designing a heuristic function η that takes into account the dual-mode characteristics of the air-ground robot and the direction of motion ij ; Step 5: the state transition probability of the ant l in the ant colony is constructed by comprehensively considering the smoothness, safety and importance of the path grid in the movement process of the air-ground robot: where l ∈ {1, 2, …, m} denotes the index number of the ant individual in the ant colony, and the size of the ant colony is m, denotes the transition probability of the ant l from the grid i to the grid j in the gth iteration, τ ij (g) denotes the current pheromone concentration on the edge ij at the gth iteration, S mij (g), S ej (g) and S j (g) respectively denote the movement smoothness of the ant from the grid i to the grid j at the gth iteration, the environmental safety and importance of the grid j, and a, b, l and w respectively denote the pheromone factor, the heuristic information factor, the turning penalty factor and the safety penalty factor, allowed l is the set of candidate neighbor nodes accessible to the ant l at the current position; Step 6: the update process of the pheromone is designed by considering the dual-mode characteristics of the air-ground robot: where τ (g + 1) represents the updated pheromone concentration, p e (0, 1) represents the pheromone evaporation factor, Δτ ij (g) represents the basic pheromone released by all ants on the edge ij in this round, and respectively represent the reinforced pheromone contributed by the high-quality path and the penalty pheromone introduced by the low-quality path, and κ represents the high-energy mode penalty factor; Step 7: the air-ground robot generates an optimal path scheme in the region to be planned based on the multi-strategy improved ant colony algorithm, and travels from the starting point S to the ending point E.

2. The multi-strategy improved ant colony algorithm based air-ground robot mountain path planning method according to claim 1, characterized in that: In step 1, the unimodal exponential function is: where x and y represent the horizontal and vertical coordinates of the ground plane, respectively, z(x, y) represents the height of the three-dimensional obstacle at (x, y) on the ground, q is the index of the three-dimensional obstacle, n is the total number of three-dimensional obstacles, h q is the terrain parameter for controlling the height, e() is the natural exponential function, x q and y q respectively represent the horizontal and vertical ground coordinates of the center of the three-dimensional obstacle q, x sq and y sq are the attenuation amounts of the three-dimensional obstacle q along the x-axis and y-axis directions, respectively.

3. The multi-strategy improved ant colony algorithm based air-ground robot mountain path planning method according to claim 1, characterized in that: In step 2, the dual-mode air-ground robot comprises two modes of high-energy consumption flight and relatively low-energy consumption ground movement.

4. The multi-strategy improved ant colony algorithm based air-ground robot mountain path planning method according to claim 1, characterized in that: In step 3, the initial pheromone is: where y u is the initial pheromone of the u-th grid, a and b are the translation and scaling constants, respectively, σ is the standard deviation of the normal distribution, and x u is the vertical distance from the u-th grid to the line segment SE.

5. The multi-strategy improved ant colony algorithm based air-ground robot mountain path planning method according to claim 1, characterized in that: In step 4, the heuristic function η is considered which takes into account the dual modality of the air-ground robot and the direction of motion ij is: where d ij + d j E - d i E is a length relationship for judging the moving direction derived in triangle ΔijE based on the triangle inequality, respectively represent the distance from the grid i to the grid j, the grid j to the end point E and the grid i to the end point E, c=0.01 is a constant for preventing the denominator from being 0, Energy ij represents the energy consumption from the grid i to the grid j: wherein χ and respectively represent the unit energy consumption of the air-robot flight mode and the ground moving mode. ψ represents a dynamic adjustment factor: where G max is the maximum number of iterations.

6. The multi-strategy improved ant colony algorithm based air-ground robot mountain path planning method according to claim 1, characterized in that: In step 5, the motion smoothness S from grid i to grid j is defined as: mij (g) is: wherein, is the angle between the path segment connecting the previous way grid (x1, y1, z1) and the current grid i (x2, y2, z2) and the path segment between the current grid i and the next potential way grid (x3, y3, z3). The environmental safety S of the grid j ej (g) is: wherein, is the Euclidean distance from the grid cell j to the nearest obstacle o, and maxSe is the maximum safety distance used in all the grids normalized. The importance S of the grid j j (g) is: S j (g) = LLM(Prompt, Human_assessment, P) wherein LLM() represents a large language model output, Prompt is a preset prompt word, including question background, task requirement, output format, S j (g) Range, Human_assessment represents an expert evaluation of the importance of grid j, indicating that the unmanned ground vehicle synchronously collects key grid data during planning, and P is a key parameter of the large language model, including top_p, top_k, Temperature, Max_Tokens, etc.

7. The multi-strategy improved ant colony algorithm based air-ground robot mountain path planning method according to claim 1, characterized in that: In step 5, the candidate neighbor node set allowed l In the flight mode, it includes 6 grid units of front, back, left, right, up and down, and in the ground mode, it includes 4 grid units of front, back, left and right.

8. The multi-strategy improved ant colony algorithm based air-ground robot mountain path planning method according to claim 1, characterized in that: In step 6, the above-mentioned is: where Q is a pheromone intensity constant, σ1 and σ2 are two weight factors, J min is the multi-objective comprehensive evaluation value of the best path in the current iteration, J max is the multi-objective comprehensive evaluation value of the worst path in the current iteration, and BestJ is the high-quality path set. The Is: Wherein, WorstJ is a low-quality path set.

9. The multi-strategy improved ant colony algorithm based air-ground robot mountain path planning method according to claim 1, characterized in that: In step 6, the mode penalty factor κ is:

Images