A path planning method based on adaptive grid and improved A*-DWA fusion algorithm

Abstract

The application relates to a path planning method based on an adaptive grid and an improved A*-DWA fusion algorithm, and belongs to the field of robots, and comprises the following steps: an environment model based on adaptive grid size is constructed, the adaptive grid length is determined based on the size and quantity of obstacles, and the grid size and the optimal grid length are calculated; the A* algorithm is improved according to known obstacle information, global path planning is completed, and a path with few nodes and high smoothness is obtained; the global path planning nodes obtained by combining unknown obstacle information in the environment model and the improved A* algorithm are selected as temporary target points in the improved DWA algorithm in sequence after the starting point; the improved DWA algorithm continuously travels from the starting point to the temporary target points, the temporary target points are updated after reaching the temporary target points, and the end point is reached. The application can realize global optimal path planning, successfully avoid random obstacles and complete dynamic path planning in a complex environment.

Description

Technical Field

[0001] This invention relates to a path planning method based on an adaptive grid and an improved A*-DWA fusion algorithm, belonging to the field of robotics technology. Background Technology

[0002] Mobile robots, with their ability to automatically execute tasks, are widely used in numerous fields such as industrial production and daily services. The efficiency of a mobile robot's path planning directly affects its work efficiency; therefore, planning a safe and efficient path has become a problem of great research significance and application value. A good path planning algorithm enables the robot to avoid both known and unknown obstacles during its movement, and to plan a superior path that safely reaches the target point.

[0003] Based on the level of environmental information available, path planning can be categorized into global path planning and local path planning. Algorithms used for global path planning include Dijkstra's algorithm, A* algorithm, and RRT algorithm. These algorithms can plan a relatively optimal path from the starting point to the destination in a known environment, but they cannot avoid unknown obstacles and are prone to collisions during travel, preventing the robot from reaching the target point. Algorithms used for local path planning include TEB algorithm, artificial potential field method, and DWA algorithm. These algorithms can avoid obstacles in real time in unknown environments, ensuring the robot's safe movement. However, when there are many obstacles, the robot may get stuck in a local loop while trying to avoid them, failing to reach the target point.

[0004] Clearly, in complex environments, single global or local path planning algorithms are insufficient to meet the movement requirements of robots. Therefore, some researchers have proposed a path planning algorithm that integrates A* and DWA algorithms. The A* algorithm is a widely used global path planning algorithm with advantages such as shorter paths and faster computation speed. However, traditional A* algorithms suffer from numerous inflection points and uneven paths in practical applications, affecting robot operation. DWA (Dynamic Window Method) is a classic local path planning algorithm that infers the robot's trajectory based on its current state and motion model, and then evaluates and determines the robot's path. However, existing path planning algorithms that integrate A* and DWA generally suffer from low path planning efficiency, complex environmental model construction processes, and insufficient rationality. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a path planning method based on an adaptive grid and an improved A*-DWA fusion algorithm. The main innovations are as follows: (1) An adaptive grid size is introduced, improving the algorithm's execution efficiency; (2) The A* algorithm is improved by optimizing the algorithm's evaluation function and the selection of final path nodes, reducing the number of search nodes and final path nodes, thus improving the algorithm's computational speed and path smoothness; (3) The DWA algorithm is improved by subdividing obstacles into unknown and known obstacles, incorporating them into the algorithm's evaluation function to enhance obstacle avoidance capabilities and enable fusion with the A* algorithm. This invention enables path planning for robots in unknown environments and achieves good results in terms of algorithm execution efficiency and path selection. This invention is mainly applied in situations where the main part of the environment map is known but there are few unknown obstacles, such as path planning for video game characters.

[0006] The present invention adopts the following technical solution:

[0007] A path planning method based on adaptive grid and improved A*-DWA fusion algorithm includes the following steps:

[0008] (1) Construct an environment model based on adaptive grid size, determine the adaptive grid side length based on the size and number of obstacles, and calculate the grid size and optimal grid side length under the adaptive grid side length;

[0009] (2) Improve the A* algorithm based on the known obstacle information in the environment model to complete global path planning and obtain a path with fewer nodes and higher smoothness;

[0010] (3) Combining the information of unknown obstacles in the environment model and the global path planning nodes obtained by the improved A* algorithm, select the nodes after the starting point as temporary target points in the improved DWA algorithm. The improved DWA algorithm moves continuously from the starting point to the temporary target point, updates the temporary target point after reaching the temporary target point, until the destination is reached.

[0011] Preferably, in step (1), the minimum grid side length of the map is set to s. min The maximum grid side length is s max If the side length of the map is m, then the maximum grid size can be expressed as follows:

[0012]

[0013] The total area of ​​the known obstacles can be represented as follows:

[0014]

[0015] Among them, Size bar iThe size of the i-th obstacle is represented by n, and the number of obstacles is represented by n.

[0016] The adaptive grid side length can be represented as follows:

[0017]

[0018] The grid size under adaptive grid side length is as follows:

[0019]

[0020] The optimal grid side length is as follows:

[0021]

[0022] Preferably, the improved A* algorithm includes path integration and optimized heuristic function.

[0023] Preferably, the specific process of path integration includes:

[0024] (2.1) Starting from the starting point S, determine whether the current direction is the same as the direction of the next node. If they are the same, delete the next node; otherwise, keep it. Traverse all nodes until the end point G.

[0025] (2.2) Starting from the starting point S, determine whether the current node can be directly connected to a non-adjacent node to form a path. The distance from the integrated path to the obstacle is used to determine the path. If the distance is greater than the set safe distance, then the path is integrated; if it is less than the set safe distance, then the original path is maintained.

[0026] (2.3) Output the final path.

[0027] The preferred process for optimizing the heuristic function is as follows:

[0028] Let the robot's starting point be (x) i ,y i ), the endpoint is (x j ,y j N represents the number of obstacles within the rectangular grid from the start point to the end point. Environmental information K is introduced, where K represents the percentage of obstacles within the rectangular grid from the start point to the end point.

[0029]

[0030] By incorporating environmental information K into the evaluation function F(n) and adjusting the weights of the heuristic function G(n) based on the number of obstacles in the grid, the improved evaluation function can be expressed as:

[0031] F(n) = G(n) + exp(-K+1) * H(n).

[0032] It can be seen that when there are few obstacles in the environment, the coefficient of H(n) is large, which can reduce the search range and improve the search speed; when there are many obstacles in the environment, the coefficient of H(n) is small, which increases the search range and ensures the global optimal path.

[0033] Preferably, in step (3), the improved DWA distinguishes between known and unknown obstacles and incorporates them into the evaluation function. The improved evaluation function is as follows:

[0034] G(v,ω)=αHead(v,ω)+βVel(v,ω)+γDist_s(v,ω)+μDist_d(v,ω)

[0035] Where Head(v,ω) is the angular deviation between the endpoint direction and the target direction of the current simulated trajectory, Vel(v,ω) is the evaluation function for the magnitude of the currently simulated velocity value, Dist_s(v,ω) is the distance between the known obstacle and the current simulated trajectory, Dist_d(v,ω) is the distance between the unknown obstacle and the current simulated trajectory, and α, β, γ, and μ are the weighting coefficients for each term.

[0036] For any details not covered in this invention, please refer to the prior art.

[0037] The beneficial effects of this invention are as follows:

[0038] This invention constructs an environment model based on an adaptive grid map to improve the search efficiency and accuracy of the algorithm. In the improved A* algorithm, the evaluation function is optimized to enhance search efficiency; redundant nodes are removed to reduce turns and improve path smoothness. The key points of the optimized path are used as temporary targets for the DWA algorithm to achieve local planning. Based on global optimization, random obstacle avoidance is implemented, and a globally optimal smooth path is fused. Experimental results show that the fusion algorithm can achieve globally optimal path planning and successfully avoid random obstacles, completing dynamic path planning in complex environments. Attached Figure Description

[0039] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application.

[0040] Figure 1 This is a schematic diagram of the path planning method based on adaptive grid and improved A*-DWA fusion algorithm of the present invention;

[0041] Figure 2 This is a schematic diagram of the raster method coordinate system.

[0042] Figure 3 This is a schematic diagram of four-way expansion;

[0043] Figure 4 This is a schematic diagram of the eight-directional expansion.

[0044] Figure 5 A schematic diagram of the path before optimization;

[0045] Figure 6 This is a schematic diagram of the optimized path;

[0046] Figure 7 This is a flowchart of the fusion algorithm. Detailed implementation method:

[0047] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. However, this is not the only description; all aspects not described in detail herein are based on conventional techniques in the art.

[0048] Traditional path planning methods are mostly based on grid maps, but the grid size is fixed and cannot be dynamically adjusted according to obstacle distribution, affecting path planning efficiency. The A* and DWA fusion algorithm is one of the hybrid path planning methods for solving the shortest path. The traditional A* algorithm cannot handle unknown obstacles beyond known information, while the DWA algorithm can achieve real-time obstacle avoidance during operation, but it is prone to getting trapped in local optima, causing the robot to fail to reach the target point. To address these issues, we propose the Pagad method. Figure 1 As shown, the path planning method based on adaptive grid and improved A*-DWA fusion algorithm of the present invention includes the following steps:

[0049] (1) Construct an environment model based on adaptive grid size, determine the adaptive grid side length based on the size and number of obstacles, and calculate the grid size and optimal grid side length under the adaptive grid side length; this step uses the area of ​​obstacles to construct the adaptive grid size, which improves the execution efficiency of the algorithm.

[0050] (2) Improve the A* algorithm based on the known obstacle information in the environment model to complete global path planning and obtain a path with fewer nodes and higher smoothness;

[0051] (3) Combining the information of unknown obstacles in the environment model with the global path planning nodes obtained by the improved A* algorithm, the nodes after the starting point are selected sequentially as temporary target points in the improved DWA algorithm. The improved DWA algorithm continuously moves towards the temporary target points from the starting point, updates the temporary target points after reaching them, until the destination is reached. This invention optimizes the heuristic function of the A* algorithm, integrates the search path, introduces known and unknown obstacles into the evaluation function of the DWA algorithm, and realizes the fusion of the A* algorithm and the DWA algorithm. It reduces the number of search nodes, improves the path smoothness, and realizes real-time obstacle avoidance and global path planning for unknown obstacles by the robot.

[0052] Environment model construction based on adaptive grid size:

[0053] An environment model transforms the robot's real-world working environment into recognizable environmental information, which is crucial for path planning systems. This invention considers the experimental environment and the ease of model construction when choosing a grid-based method for constructing the environment model, and proposes an environment model construction method based on adaptive grid size.

[0054] Traditional raster mapping typically involves pre-setting the raster size and then using coordinate systems to number the raster map. This means placing the two-dimensional raster map in a Cartesian coordinate system, where any location on the map can be represented by coordinates (x, y). For example... Figure 2 As shown, the actual environment is constructed as a two-dimensional grid map of size 6×6. Obstacles are represented by black squares, the starting point is represented by square A, and the ending point is represented by square B. Therefore, the starting point coordinates are [1,6], the ending point coordinates are [6,1], the obstacle coordinates are [(2,1),(3,4),(4,4),(5,5)], and the remaining white squares represent the robot's activity area. The robot can move within the areas excluding obstacles.

[0055] When constructing a map using the grid method, the following points need to be considered: (1) the size of the environment map and the grid size; (2) the one-to-one correspondence between the positions of obstacles in the environment model and those in the experimental scene; and (3) the computing power of the computer. Among these, the choice of grid size is crucial. A grid size that is too large will hinder the discovery of the optimal path, while a grid size that is too small will affect the execution efficiency. To this end, this invention comprehensively considers three factors: the size of the real environment, the number of obstacles, and the size of the obstacles, and proposes an environment model construction method based on adaptive grid size.

[0056] In practice, the grid side length of this invention is related to the number and distribution of obstacles. The more numerous and complex the obstacles, the smaller the grid side length should theoretically be. Analysis shows that the total area ratio of obstacles can reflect their number and distribution to some extent. Therefore, an adaptive minimum grid side length construction method based on the total area ratio of obstacles is constructed.

[0057] Let the minimum grid side length of the map be s. min The maximum grid side length is s max If the side length of the map is m, then the maximum grid size can be expressed as follows:

[0058]

[0059] The total area of ​​the known obstacles can be represented as follows:

[0060]

[0061] Among them, Size bar_i The size of the i-th obstacle is represented by n, and the number of obstacles is represented by n.

[0062] The adaptive grid side length can be represented as follows:

[0063]

[0064] The grid size under adaptive grid side length is as follows:

[0065]

[0066] The optimal grid side length is as follows:

[0067]

[0068] Traditional A* algorithm:

[0069] The traditional A* algorithm is a heuristic algorithm capable of global path planning, primarily used in 2D raster maps. The A* algorithm combines the advantages of Dijkstra's search algorithm and Best-First Search (BFS). Dijkstra's algorithm searches for paths at the cost of the distance between the expanded point and the starting point, but it doesn't consider the distance relationship with the target point, resulting in a longer search time to find the optimal path. BFS, on the other hand, considers the relationship between the expanded point and the target point, but the resulting path is not necessarily optimal. The A* algorithm integrates the cost of the starting point in Dijkstra's algorithm and the cost of the target point in BFS, establishing a new evaluation function.

[0070] The A* algorithm constructs two lists, openlist and closelist. Starting from the starting point, the current node is set as the parent node, and the algorithm expands to the surrounding neighboring nodes. The expanded neighboring nodes are added to the openlist, and the evaluation function value is calculated. The node with the smallest evaluation function value in the openlist is removed from the openlist and added to the closelist. This node is then set as the parent node again. The search of the openlist and closelist is repeated until the endpoint is reached.

[0071] The evaluation function of the traditional A* algorithm is shown in the following formula:

[0072] F(n) = G(n) + H(n)

[0073] F(n) represents the total cost of the current node n in the algorithm; G(n) represents the actual cost from the starting point to node n; and H(n) represents the estimated cost from the current node to the destination, also known as the heuristic function. The choice of the heuristic function affects the calculation of the evaluation function, and thus the computation and path selection of the entire algorithm. Commonly used heuristic functions in the traditional A* algorithm include Manhattan distance, Chebyshev distance, and Euclidean distance. This paper selects Euclidean distance as the heuristic function for the traditional A* algorithm.

[0074] When the A* algorithm expands from the current node to neighboring nodes, it typically uses four-way or eight-way expansion, such as... Figure 3 , 4 As shown, the four-way expansion directions are up, down, left, and right, while the eight-way expansion adds four more directions: upper left, lower left, upper right, and lower right. Four-way expansion is often used in structured warehouse search methods; this invention employs eight-way expansion.

[0075] The evaluation function design of the traditional A* algorithm is unreasonable. The cost from the current node to the starting point and the cost to the ending point are given roughly equal weight in the evaluation function, resulting in a large number of search nodes, low search efficiency, and an inability to flexibly adjust to different environments. Traditional A* algorithms plan paths with many nodes, often resulting in multiple nodes lying on the same straight line. In such cases, the robot needs to track a large number of nodes for simple movements. Furthermore, there may be turning points between multiple path segments even without obstacles, making the path uneven and hindering robot movement. Finally, the A* algorithm performs global path planning in a known environment and cannot achieve real-time obstacle avoidance for unknown obstacles.

[0076] Improved A* algorithm:

[0077] To address the problems of numerous path nodes, many turning points, uneven paths, and low search efficiency in 2D raster maps, the traditional A* algorithm improves the A* algorithm in two aspects: path integration and optimization of the heuristic function.

[0078] A. Path Integration

[0079] By merging multiple paths into a single path and eliminating redundant nodes, the smoothness of the path can be improved. For example... Figure 5 As shown, the initial path consists of nodes S, G, and 1-8. It can be seen that the path has many turns and redundant nodes.

[0080] (2.1) Starting from the starting point S, determine whether the current direction is the same as the direction of the next node. If they are the same, delete the next node; otherwise, keep it. Traverse all nodes until the end point G. Delete nodes 1, 3, 5, 7, and 8 in the path. After this step, the path is S-2-4-6-G.

[0081] (2.2) Starting from the starting point S, determine whether the current node can be directly connected to a non-adjacent node to form a path. This is determined by the distance from the integrated path to the obstacle. If this distance is greater than the set safety distance, then integration is performed; if it is less than the set safety distance (the safety distance can be set according to the actual operating environment, for example, it can be set to the center-to-edge distance of the square representing the obstacle), then the original path is maintained. If the distance between road segment S-4 and the obstacle is greater than the set safety distance, node 2 is discarded; if the distance between road segment 2-6 and the obstacle is less than the safety distance, node 4 is retained. Figure 6 As shown, the final path obtained is S-4-6-G;

[0082] (2.3) Output the final path.

[0083] B. Optimize the heuristic function

[0084] The traditional A* algorithm selects paths based on an evaluation function. The actual cost H(n) from the current node to the starting point and the weight of the heuristic function G(n) in the evaluation function affect the algorithm's search efficiency. The computational speed of the algorithm can be optimized by adjusting the heuristic function.

[0085] In complex environments, the presence of obstacles often forces algorithms to expand their search range to ensure reaching the destination, leading to slower search speeds and more redundant nodes. When there are few obstacles, the search range can be reduced and the search speed increased by increasing the weight of the heuristic function H(n); conversely, when there are many obstacles, the weight of the heuristic function H(n) can be reduced to expand the search range and avoid getting trapped in local optima. To address this, environmental information K is introduced, where K represents the proportion of obstacles within the rectangular grid from the starting point to the destination.

[0086] Let the robot's starting point be (x)i ,y i ), the endpoint is (x j ,y j N is the number of obstacles within the rectangular grid from the start point to the end point.

[0087]

[0088] By incorporating environmental information K into the evaluation function F(n) and adjusting the weights of the heuristic function G(n) based on the number of obstacles in the grid, the improved evaluation function can be expressed as:

[0089] F(n) = G(n) + exp(-K+1) * H(n).

[0090] It can be seen that when there are few obstacles in the environment, the coefficient of H(n) is large, which can reduce the search range and improve the search speed; when there are many obstacles in the environment, the coefficient of H(n) is small, which increases the search range and ensures the global optimal path.

[0091] Traditional DWA algorithm:

[0092] DWA (Directed Path Avoidance) is a classic local path planning algorithm for robots. It samples linear and angular velocities in the velocity space and predicts the robot's trajectory for the next time interval based on its kinematic model. The algorithm then scores the evaluated trajectory to obtain a safer and smoother optimal local path. DWA is often used in scenarios where the coordinates of the robot and the destination are known, but the coordinates of obstacles are unknown; or where sensors can detect obstacles and provide the distance and angle between the robot and the obstacles, thus compensating for the A* algorithm's inability to perform real-time obstacle avoidance.

[0093] The dynamic window method simulates the robot's motion trajectory, scores the simulated paths, and selects the highest-scoring trajectory as the movement path. Let v(t) be the robot's linear velocity at time t, and ω(t) be the robot's angular velocity at time t. The kinematic model can be established as follows:

[0094]

[0095] The above formula constructs the robot as a model in a two-dimensional Cartesian coordinate system, where Δt represents a unit of time, x(t) represents the horizontal coordinate of the robot at time t, y(t) represents the vertical coordinate of the robot at time t, and θ(t) represents the angle of the robot at time t.

[0096] The dynamic window method samples the robot's velocity space and constrains the sampling range according to the actual situation. Obstacle avoidance is achieved through self-constraint and environmental constraints.

[0097] Robot speed constraints: The robot's moving speed and rotational angular velocity should be kept within a certain range;

[0098] v m ={(v,ω)v∈[v min ,v max ],ω∈[ω min ,ω max ]}

[0099] v m It is the combination of the robot's linear velocity and angular velocity;

[0100] Robot acceleration and deceleration constraints: Taking into account the actual performance of the motors used, the maximum acceleration and maximum deceleration of the robot are kept within a reasonable range;

[0101]

[0102] In the formula, v c ω represents the current linear velocity. c Indicates the current angular velocity. This represents the robot's maximum linear acceleration. This represents the robot's maximum angular acceleration. This represents the robot's minimum linear acceleration. This represents the robot's minimum angular acceleration.

[0103] Robot safety distance constraint: In order to avoid collisions with obstacles during the movement of the robot, a safe distance should be maintained from the obstacles under acceleration and deceleration constraints, so that the robot's speed is reduced to zero before colliding with the obstacles;

[0104]

[0105] In the above formula, dist(v,ω) represents the closest distance between the trajectory and the obstacle.

[0106] The final velocity space is the intersection of the three sets mentioned above, with the dynamic window V. r The definition is as follows:

[0107] V r =V m ∩V d ∩V s

[0108] Improved DWA algorithm:

[0109] Traditional dynamic window methods can successfully avoid obstacles during operation, but without the guidance of global path planning, they are prone to getting stuck in local optima, preventing them from reaching the target point. In the improved fusion algorithm, global path planning guides the dynamic window method, classifying obstacles into known and unknown categories, requiring improvements to the evaluation function of the traditional dynamic window method.

[0110] The evaluation function of the traditional dynamic window method includes three factors: deflection angle, linear velocity, and the distance between the end of the simulated trajectory and the obstacle. The formula is as follows:

[0111] G(v,ω)=αHead(v,ω)+βVel(v,ω)+γDist(v,ω)

[0112] Where Head(v,ω) is the angular deviation between the endpoint direction and the target direction of the current simulated trajectory, Vel(v,ω) is the evaluation function of the magnitude of the currently simulated velocity value, Dist(v,ω) is the distance between the obstacle and the current simulated trajectory, and α, β, and γ are the weighting coefficients of each term.

[0113] As can be seen from the evaluation function, the traditional DWA algorithm can only evaluate a single obstacle. When there are both unknown and known obstacles, the evaluation function cannot evaluate both types of obstacles at the same time, resulting in the final path overlapping with the obstacle and causing the robot to collide with it.

[0114] The improved DWA distinguishes between known and unknown obstacles and incorporates them into the evaluation function. The improved evaluation function is as follows:

[0115] G(v,ω)=αHead(v,ω)+βVel(v,ω)+γDist_s(v,ω)+μDist_d(v,ω)

[0116] Where Dist_s(v,ω) is the distance between the known obstacle and the current simulated trajectory, Dist_d(v,ω) is the distance between the unknown obstacle and the current simulated trajectory, and α, β, γ, and μ are the weighting coefficients of each term.

[0117] The weighting coefficients for each item can be selected based on the user's needs. For example, to enhance the robot's ability to reach the target point, α can be increased; to enhance the robot's ability to avoid unknown obstacles, μ can be increased. By adjusting the weighting coefficients in practical applications, the parameters most suitable for the user can be obtained.

[0118] Whether it's the traditional DWA algorithm or the improved DWA algorithm, the robot will move along the path with the minimum evaluation function value. In the application of the DWA algorithm, the evaluation function is prone to getting stuck in a local minimum, which makes it impossible for the robot to avoid obstacles and reach the target point.

[0119] Algorithm fusion:

[0120] The A* algorithm can perform global path planning using known environmental information, but it cannot perform real-time obstacle avoidance for unknown obstacles. In environments with unknown obstacles, the algorithm cannot plan a complete path. The DWA algorithm can perform local path planning with real-time obstacle avoidance, but it is prone to getting trapped in local optima, preventing it from reaching the target point and failing to plan a complete path. To address the respective advantages and disadvantages of the two algorithms, an optimized A* algorithm and the DWA algorithm are fused. The A* algorithm is used to guide the DWA algorithm globally, solving the obstacle avoidance problem for both known and unknown obstacles. Furthermore, while maintaining global path optimization, the smoothness of the path is improved, enabling the robot to perform global path planning with real-time obstacle avoidance.

[0121] Figure 7 A framework for the fusion algorithm is presented, which combines the optimized A* algorithm and the DWA algorithm for path planning. The main steps are as follows:

[0122] Step 1: Construct a grid map using known information. Obstacles in the map are classified as known obstacles. The optimized A* algorithm completes global path planning in the grid map, resulting in a path with fewer nodes and higher smoothness.

[0123] Step 2: Classify obstacles other than known obstacles on the map as unknown obstacles. Extract the nodes obtained by the A* algorithm. Nodes after the starting point are used as temporary target points in the optimized DWA algorithm. The optimized DWA algorithm continuously moves towards the temporary target points from the starting point. After reaching the temporary target point, the temporary target point is updated until the destination is reached, thus realizing real-time obstacle avoidance during the robot's movement.

[0124] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A path planning method based on an adaptive grid and an improved A*-DWA fusion algorithm, characterized in that, Includes the following steps: (1) Construct an environment model based on adaptive grid size, determine the adaptive grid side length based on the size and number of obstacles, and calculate the grid size and optimal grid side length under the adaptive grid side length; (2) Improve the A* algorithm based on the known obstacle information in the environment model to complete global path planning and obtain a path with fewer nodes and higher smoothness; (3) Combining the information of unknown obstacles in the environment model and the global path planning nodes obtained by the improved A* algorithm, select the nodes after the starting point as temporary target points in the improved DWA algorithm. The improved DWA algorithm moves continuously from the starting point to the temporary target point, updates the temporary target point after reaching the temporary target point, until the end point is reached. In step (1), let the minimum grid side length of the map be s. min The maximum grid side length is s max If the map side length is m, then the maximum grid size is expressed as follows: The total area of ​​the obstacles is represented as follows: Among them, Size bar_i The size of the i-th obstacle is represented by n, and the number of obstacles is represented by n. The adaptive grid side length is represented as follows: The grid size under adaptive grid side length is as follows: The optimal grid side length is as follows:

2. The path planning method based on adaptive grid and improved A*-DWA fusion algorithm according to claim 1, characterized in that, Improvements to the A* algorithm include path integration and optimization of the heuristic function.

3. The path planning method based on adaptive grid and improved A*-DWA fusion algorithm according to claim 2, characterized in that, The specific process of path integration includes: (2.1) Starting from the starting point S, determine whether the current direction is the same as the direction of the next node. If they are the same, delete the next node; otherwise, keep it. Traverse all nodes until the end point G. (2.2) Starting from the starting point S, determine whether the current node can be directly connected to a non-adjacent node to form a path. The distance from the integrated path to the obstacle is used to determine the path. If the distance is greater than the set safe distance, then the path is integrated; if it is less than the set safe distance, then the original path is maintained. (2.3) Output the final path.

4. The path planning method based on adaptive grid and improved A*-DWA fusion algorithm according to claim 3, characterized in that, The process of optimizing the heuristic function is as follows: Let the robot's starting point be (x) i ,y i ), the endpoint is (x j ,y j N represents the number of obstacles within the rectangular grid from the start point to the end point. Environmental information K is introduced, where K represents the percentage of obstacles within the rectangular grid from the start point to the end point. By incorporating environmental information K into the evaluation function F(n), and adjusting the weights of the heuristic function G(n) based on the number of obstacles in the grid, the improved evaluation function is expressed as: F(n) = G(n) + exp(-K+1) * H(n).

5. The path planning method based on adaptive grid and improved A*-DWA fusion algorithm according to claim 4, characterized in that, In step (3), the improved DWA distinguishes between known and unknown obstacles and incorporates them into the evaluation function. The improved evaluation function is as follows: G(v,ω)=αHead(v,ω)+βVel(v,ω)+γDist_s(v,ω)+μDist_d(v,ω) Where Head(v,ω) is the angular deviation between the endpoint direction and the target direction of the current simulated trajectory, Vel(v,ω) is the evaluation function for the magnitude of the currently simulated velocity value, Dist_s(v,ω) is the distance between the known obstacle and the current simulated trajectory, Dist_d(v,ω) is the distance between the unknown obstacle and the current simulated trajectory, and α, β, γ, and μ are the weighting coefficients for each term.

Images