AGV dynamic path planning method and system in complex environment

Abstract

The application relates to an AGV dynamic path planning method and system in a complex environment, and belongs to the technical field of AGV autonomous navigation and intelligent path planning. The method comprises the following steps: generating an initial global feasible path of an AGV based on an improved RRT* algorithm; optimizing the initial global feasible path by using an MOPSO algorithm to obtain a final global planning path; acquiring real-time environment sensing data in the AGV driving process; when an obstacle is detected, completing dynamic adjustment of a local path based on an improved dynamic window method; if no obstacle is detected, maintaining the global planning path to continuously drive until the target endpoint is reached. Through static and dynamic simulation experiments, it is proved that the application can effectively make up for the performance short board of a single global or local path planning algorithm, significantly improve the path planning efficiency, trajectory smoothness and dynamic obstacle avoidance capability of the AGV in a complex environment, and meet the high-precision autonomous navigation requirements of the AGV in multiple scenes such as industry, warehousing and logistics.

Description

Technical Field

[0001] This invention relates to a dynamic path planning method and system for AGVs in complex environments, belonging to the field of AGV autonomous navigation and intelligent path planning technology. Background Technology

[0002] As a core execution device in intelligent manufacturing and automated logistics systems, AGVs' autonomous path planning capability is key to ensuring operational efficiency and safety. Currently, they are widely used in industrial scenarios such as material handling, warehousing and sorting, and production line support.

[0003] As application scenarios become increasingly complex, AGV operating environments not only contain fixed static obstacles, but also frequently encounter dynamic obstacles such as personnel and mobile devices. Furthermore, some environmental information remains unknown, which places higher technical demands on the real-time obstacle avoidance capability, trajectory smoothness, and global guidance of AGV path planning.

[0004] Current AGV path planning technologies are mainly divided into two types: global path planning and local path planning. Each type has inherent technical limitations that are difficult to overcome when used as a single algorithm. While some existing technologies attempt to combine global and local planning, these are mostly simple concatenation algorithms that fail to address the core flaws of the algorithms or consider comprehensive optimization across multiple metrics. This results in numerous problems with the fusion algorithms, leading to issues such as excessively long travel distances, operational jitter, and insufficient obstacle avoidance safety margins in practical applications. Consequently, these algorithms struggle to meet the high-precision operational requirements of AGVs in complex industrial scenarios.

[0005] In summary, existing AGV path planning technologies suffer from problems such as blind global planning sampling, lack of global guidance in local planning, absence of multi-objective path optimization mechanisms, and poor coordination between global and local planning. These issues result in low path planning efficiency, poor trajectory stability, and insufficient dynamic obstacle avoidance capabilities for AGVs in dynamic environments, failing to meet the high requirements of modern industrial production for autonomous navigation of AGVs. Summary of the Invention

[0006] To improve the efficiency of AGV path planning in dynamic environments, and to enhance trajectory stability and dynamic obstacle avoidance capabilities, this invention provides a dynamic path planning method and system for AGVs in complex environments. The technical solution is as follows: In a first aspect, the present invention provides a dynamic path planning method for AGVs in complex environments, characterized in that the method includes: Step 1: Construct an environmental map for AGV operations, set the starting point and target endpoint of AGV operations, and configure the kinematic parameters of AGVs; Step 2: Generate the initial globally feasible path for the AGV based on the improved RRT* algorithm; The improved RRT* algorithm includes: calculating the gravitational potential field of the target point to the AGV and the repulsive potential field of the obstacle to the AGV, constructing an artificial potential field composite force field model and performing node direction guidance; Step 3: Optimize the initial globally feasible path using the MOPSO algorithm to obtain the final globally planned path; Step 4: Acquire real-time environmental perception data during AGV's travel. When an obstacle is detected, dynamically adjust the local path based on the improved dynamic window method. After obstacle avoidance, guide the AGV back to the final global planned path. If no obstacle is detected, continue traveling along the global planned path until the target destination is reached. The dynamic adjustment process of the local path includes: generating a dynamic speed window based on AGV hardware performance boundary constraints, AGV motor acceleration and deceleration performance constraints, and AGV driving safety distance constraints; constructing a trajectory evaluation function; traversing all speed combinations within the dynamic speed window; and selecting the optimal speed combination of the trajectory evaluation function to generate a local obstacle avoidance trajectory.

[0007] Optionally, the processing procedure for the artificial potential field composite force field model includes: Define the gravitational potential field function of the target point with respect to the AGV, as follows:

[0008] in, This is the gravitational gain coefficient. The Euclidean distance between the current position of the AGV and the target point position; The gravitational force generated at the AGV by this gravitational potential field is the negative gradient of the potential field function, expressed as:

[0009] Define the repulsive potential field function of obstacles to AGV. Introduce a target point distance correlation term into the traditional repulsive potential field to obtain an improved repulsive potential field function, the expression of which is:

[0010] in The repulsive gain coefficient is... This represents the Euclidean distance between the current position of the AGV and the position of the obstacle. The maximum effective distance of the repulsive force of the obstacle; The repulsive force generated at the AGV by the improved repulsive potential field is the negative gradient of the potential field function, which can be decomposed into components in two directions, expressed as:

[0011]

[0012] The net force acting on an AGV in a combined potential field is the vector sum of attractive and repulsive forces, expressed as: .

[0013] Optionally, the objective function for path optimization using the MOPSO algorithm in step 3 is determined based on the path length evaluation function, the trajectory smoothness evaluation function, and the driving safety evaluation function; wherein, the path length evaluation function calculates the sum of the Euclidean distances of all adjacent nodes on the path, the trajectory smoothness evaluation function calculates the cumulative value of the steering angles of all adjacent nodes on the path, and the driving safety evaluation function calculates the minimum distance from all nodes on the path to the nearest obstacle.

[0014] Optionally, the process of determining the velocity sampling space using the improved dynamic window method includes: Based on the AGV hardware performance boundary constraints, the first constraint space for speed sampling is determined:

[0015] in v For linear velocity, oh Angular velocity; Based on the acceleration and deceleration performance constraints of the AGV motor, the second constraint space for speed sampling is determined:

[0016] in and Δ represents the current linear velocity and angular velocity of the AGV. t The sampling period; Based on the AGV's safe travel distance constraint, the third constraint space for speed sampling is determined:

[0017] in To predict the shortest distance from the AGV to the obstacle on the trajectory.

[0018] Considering the above three types of constraints, the final speed sampling space of the AGV is: .

[0019] Optionally, the dynamic adjustment process of the local path includes constructing a comprehensive trajectory evaluation function, expressed as:

[0020] Among them, the heading angle evaluation function The safe distance evaluation function represents the degree of matching between the heading deviation of the predicted trajectory endpoint and the target point. The velocity evaluation function represents the minimum distance between the predicted trajectory and the obstacle. This indicates the magnitude of the AGV's linear speed.α , β , c These are the weighting coefficients for the three indicators.

[0021] Optionally, the method further includes: normalizing the heading angle evaluation function, the safety distance evaluation function, and the speed evaluation function before constructing a comprehensive trajectory evaluation function.

[0022] Optionally, the kinematic parameters of the AGV include: maximum linear velocity, minimum linear velocity, maximum angular velocity, minimum angular velocity, maximum linear acceleration, and maximum angular acceleration.

[0023] Secondly, the present invention provides an AGV dynamic path planning system for complex environments, comprising: The environment modeling and parameter configuration module is used to build an environmental map for AGV operations, set the starting point and target endpoint of AGV operations, and configure the kinematic parameters of AGVs. The initial path generation module is used to generate an initial globally feasible path for the AGV based on the improved RRT* algorithm. The improved RRT* algorithm includes: calculating the gravitational potential field of the target point to the AGV and the repulsive potential field of the obstacle to the AGV, constructing an artificial potential field composite force field model and performing node direction guidance. The global path planning module is used to optimize the initial globally feasible path using the MOPSO algorithm to obtain the final globally planned path. The path execution and collaborative control module is used to acquire real-time environmental perception data during the AGV's travel. When an obstacle is detected, the local path is dynamically adjusted based on the improved dynamic window method. After obstacle avoidance, the AGV is guided back to the final global planned path. If no obstacle is detected, the AGV continues to travel along the global planned path until it reaches the target destination. The dynamic adjustment process of the local path includes: generating a dynamic speed window based on AGV hardware performance boundary constraints, AGV motor acceleration and deceleration performance constraints, and AGV driving safety distance constraints; constructing a trajectory evaluation function; traversing all speed combinations within the dynamic speed window; and selecting the optimal speed combination of the trajectory evaluation function to generate a local obstacle avoidance trajectory.

[0024] Thirdly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the AGV dynamic path planning method in complex environments as described in any of the preceding claims.

[0025] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that: when the computer program is executed by a processor, it implements the steps of the AGV dynamic path planning method under complex environments as described in any of the preceding claims.

[0026] The beneficial effects of this invention are: This invention integrates an improved artificial potential field composite force field model and an adaptive dynamic step size strategy into the RRT* algorithm to achieve directional generation of random tree nodes, significantly reducing sampling blindness, reducing redundant nodes, improving algorithm convergence efficiency, and generating an initial global path that is more closely aligned with the target direction. This effectively solves the problem of blind sampling in global planning, improving both path planning efficiency and initial path quality.

[0027] This invention improves the DWA algorithm by optimizing the dynamic speed window and constructing a trajectory evaluation system. This ensures that the AGV can avoid dynamic and unknown obstacles in real time while always traveling along the preset direction of the global path. This fundamentally solves the problem that the traditional DWA algorithm is prone to deviating from the target, realizes deep collaboration between global planning and local obstacle avoidance, and improves trajectory stability.

[0028] In one embodiment, the present invention introduces the MOPSO algorithm to construct a multi-objective evaluation system that takes into account three core indicators: path length, trajectory smoothness, and driving safety. By solving the non-dominated solution set through Pareto optimality theory, the globally optimal path selected can balance the AGV's operating efficiency, operational stability, and driving safety, thus breaking through the limitations of traditional single-objective optimization.

[0029] Experiments have shown that the method of this invention can effectively cope with complex working environments with irregular static obstacle distribution, random dynamic obstacle appearance, and partially unknown environmental information. It exhibits excellent path planning performance in simple environments, industrial workshop environments, and complex multi-obstacle environments, and has high real-time obstacle avoidance capability and operational stability. It can be widely applied to the autonomous navigation system of various industrial AGVs.

[0030] The path planning system of this invention adopts a modular architecture design, with each module having independent functions and working together. It can flexibly connect to hardware devices and sensor systems of different AGV models, supports real-time modification and optimization of parameters, and is easy to implement in engineering and practical application scenarios. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart of the AGV dynamic path planning method under complex environments according to the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0034] Example 1: This embodiment provides a dynamic path planning method for AGVs in complex environments, including: Step 1: Construct an environmental map for AGV operations, set the starting point and target endpoint of AGV operations, and configure the kinematic parameters of AGVs; Step 2: Generate the initial globally feasible path for the AGV based on the improved RRT* algorithm; The improved RRT* algorithm includes: calculating the gravitational potential field of the target point to the AGV and the repulsive potential field of the obstacle to the AGV, constructing an artificial potential field composite force field model and performing node orientation guidance; Step 3: Optimize the initial globally feasible path using the MOPSO algorithm to obtain the final globally planned path; Step 4: Acquire real-time environmental perception data during AGV's travel. When an obstacle is detected, dynamically adjust the local path and generate an obstacle avoidance trajectory based on the improved dynamic window method. After obstacle avoidance, guide the AGV back to the final global planning path. If no obstacle is detected, continue traveling on the final global planning path until the target destination is reached. The process of determining the speed sampling space using the improved dynamic window method includes: constructing three constraint spaces based on AGV hardware performance boundary constraints, AGV motor acceleration and deceleration performance constraints, and AGV driving safety distance constraints, and taking the intersection of the three constraint spaces as the final speed sampling space.

[0035] Example 2: This embodiment provides a dynamic path planning method for AGVs in complex environments, including the following steps: Step 1: Construct an AGV operation environment map, obtain the spatial location and geometric dimensions of static obstacles in the environment, set the starting point and target endpoint of the AGV operation, and configure the AGV kinematic parameters, including the maximum linear velocity. Maximum angular velocity Maximum linear acceleration Maximum angular acceleration .

[0036] Step 2: Based on the environmental map information and AGV kinematic parameters, construct an improved RRT* algorithm model for global path planning, calculate the gravitational potential field of the target point to the AGV and the repulsive potential field of the obstacle to the AGV, establish an improved artificial potential field composite force field model and execute node direction guidance, combine potential field guidance with RRT* random tree expansion to generate the initial global feasible path of the AGV.

[0037] The processing procedure for the improved artificial potential field composite force field model is as follows: Define the gravitational potential field function of the target point relative to the AGV. The gravitational potential field is only related to the relative distance between the current position of the AGV and the position of the target point. The expression is:

[0038] in This is the gravitational gain coefficient. This represents the Euclidean distance between the current position of the AGV and the target point.

[0039] The gravitational force generated at the AGV by this gravitational potential field is the negative gradient of the potential field function, expressed as:

[0040] Define an improved repulsive potential field function for obstacles on the AGV, introducing a target point distance correlation term into the traditional repulsive potential field, expressed as:

[0041] in The repulsive gain coefficient is... This represents the Euclidean distance between the current position of the AGV and the position of the obstacle. The maximum effective distance of the repulsive force of the obstacle is where n For any given value greater than or equal to 2, in this embodiment n The value is 2.

[0042] The repulsive force generated at the AGV by this improved repulsive potential field is the negative gradient of the potential field function, which can be decomposed into components in two directions, expressed as:

[0043]

[0044] The net force acting on an AGV in a combined potential field is the vector sum of attractive and repulsive forces, expressed as:

[0045] Step 3: Perform path integration optimization based on multi-objective particle swarm optimization (MOPSO) on the initial global feasible path: transform the initial path into the initial solution space of the particle swarm, calculate the objective function vector of each particle based on three preset evaluation functions: path length, trajectory smoothness, and driving safety, solve the non-dominated solution set based on Pareto optimality theory, and select the solution with the best comprehensive performance as the final global planning path.

[0046] The path length evaluation function calculates the sum of the Euclidean distances between all adjacent nodes on the path, and its expression is:

[0047] in n This represents the total number of path nodes.

[0048] The trajectory smoothness evaluation function calculates the cumulative turning angle of all adjacent nodes on the path, and its expression is:

[0049] The driving safety evaluation function calculates the minimum distance from all nodes on the path to the nearest obstacle, and its expression is:

[0050] in Indicates the first j The equivalent collision radius of each obstacle It represents a positive number that approaches 0.

[0051] Based on the above three evaluation functions, the objective function vector of the particle can be determined. The optimization objective is to minimize the values ​​of each dimension of the objective function vector.

[0052] The optimization process of the MOPSO algorithm is as follows: A set of initial particles is randomly generated, each particle corresponding to a complete AGV global path. Each particle is substituted into a multi-objective evaluation system to calculate the three-dimensional objective function vectors corresponding to path length, trajectory smoothness, and driving safety.

[0053] The current particle's three-dimensional objective function vector is compared with the particle's own historical best solution using Pareto dominance. If the current solution dominates the historical best solution, the individual best solution is updated; otherwise, it remains unchanged. An external archive of fixed size is constructed. The current particle's solution is compared with the existing non-dominated solutions in the external archive using dominance. The external archive is updated according to the rules to ensure that the archive always contains non-dominated solutions and that the size remains stable.

[0054] Calculate the crowding distance of all non-dominated solutions in the external archive, and select the solution with the larger crowding distance as the global optimal particle to ensure the uniformity of the spatial distribution of solutions.

[0055] By combining parameters such as inertia weight and acceleration coefficient, the velocity vector and position vector of the particle are updated, the three-dimensional objective function vector of the new particle is recalculated, and the above steps are repeated until the preset number of iterations is reached. Finally, the global path with the best overall performance is selected from the external archives.

[0056] Step 4: Perform local path planning processing based on the improved dynamic window method (DWA) on the final global planning path: initialize the speed sampling space and trajectory evaluation system, generate a dynamic speed window by combining AGV kinematic constraints and safety distance constraints, construct a trajectory evaluation function, traverse all speed combinations within the speed window, select the speed combination with the optimal evaluation function to generate a local obstacle avoidance trajectory, and realize real-time avoidance of dynamic obstacles by the AGV.

[0057] The process of determining the dynamic speed window is as follows: Based on the AGV hardware performance boundary constraints, the first constraint space for speed sampling is determined:

[0058] in v For linear velocity, oh Angular velocity, Minimum linear velocity, For the maximum linear velocity, Minimum angular velocity, This represents the maximum angular velocity.

[0059] Based on the acceleration and deceleration performance constraints of the AGV motor, the second constraint space for speed sampling is determined:

[0060] in and The current linear velocity and angular velocity of the AGV. For linear acceleration, Let Δ be the angular acceleration. t The sampling period.

[0061] Based on the AGV's safe travel distance constraint, the third constraint space for speed sampling is determined:

[0062] in To predict the shortest distance from the AGV to the obstacle on the trajectory.

[0063] Considering the above three types of constraints, the final speed sampling space of the AGV is: .

[0064] A trajectory evaluation strategy that comprehensively considers heading angle, safety distance, and travel speed is proposed during the trajectory evaluation process.

[0065]

[0066]

[0067] in, This indicates the deviation of the heading angle between the current predicted trajectory endpoint and the target point of the AGV. Indicates the safe distance between the AGV and obstacles. This represents the shortest distance between the end point of the AGV's current predicted trajectory and an adjacent obstacle. This indicates the maximum range that the robot can explore.

[0068] Construct the basic evaluation functions for the above three dimensions, and the heading angle evaluation function. The predicted trajectory endpoint and target point heading deviation matching degree, and the safe distance evaluation function. The velocity evaluation function represents the minimum distance between the predicted trajectory and the obstacle. This indicates the magnitude of the AGV's linear speed.

[0069] The three basic evaluation functions are subjected to extreme value normalization to eliminate dimensional differences. The normalization formula is as follows:

[0070]

[0071]

[0072] Based on the normalized index, a comprehensive trajectory evaluation function is constructed, with the following expression:

[0073] in α , β , c These are the weighting coefficients for the three indicators.

[0074] Step 6: By combining the global path planning with the local dynamic trajectory adjustment, the AGV is controlled to complete the driving task from the starting point to the target point, realizing the dynamic path planning and autonomous navigation of the AGV in complex environments.

[0075] Example 3: This embodiment provides a dynamic path planning method for AGVs in complex environments, and experiments were conducted to verify the technical effectiveness of the method.

[0076] The implementation steps of this embodiment are as follows: Step 1: Construct an AGV operation environment map, obtain the spatial location and geometric dimensions of static obstacles in the environment, set the starting point and target endpoint of the AGV operation, and configure the AGV kinematic parameters, including the maximum linear velocity. Minimum linear velocity Maximum angular velocity Minimum angular velocity Maximum linear acceleration Maximum angular acceleration .

[0077] Step 2: Perform global path planning based on the improved RRT algorithm.

[0078] An improved artificial potential field method is incorporated into the RRT algorithm to reconstruct the repulsive potential field function and introduce a target point distance correlation term to solve the target unreachability problem. The node expansion step size is dynamically adjusted based on the spatial proportion of obstacles within the AGV's detection range to achieve directional generation of random tree nodes, overcoming the blind sampling of traditional RRT and generating an initial globally feasible path. The adaptive dynamic step size adjustment strategy expression is as follows:

[0079]

[0080]

[0081] in, Let x represent the unit vector from y to x. This indicates the percentage of space occupied by obstacles within the detection range of the mobile robot. This represents the area of ​​the sampling space occupied by the obstacle. denoted by , where m represents the total area of ​​the sampling space and m represents the number of obstacles.

[0082] Step 3: Optimize the initial global path using the MOPSO algorithm.

[0083] A multi-objective evaluation system based on path length, trajectory smoothness, and driving safety is constructed. Fifty initial particles are randomly generated, each corresponding to a global path. The three-dimensional objective function vector of each particle is calculated. The individual optimal solution and the external archive are compared and updated through Pareto dominance relationship. The congestion distance of non-dominated solutions is calculated and the globally optimal particle is selected. After 100 iterations, the particle velocity and position are updated in combination with inertia weight. After the iteration is completed, the global planning path with the best comprehensive performance is selected from the external archive.

[0084] Step 4: Configure the improved DWA algorithm to obtain local path planning parameters.

[0085] Three types of sampling constraints are set: speed boundary, acceleration / deceleration, and safety distance. Weighting coefficients of 0.3, 0.3, and 0.2 are set for the three indicators: heading angle, safety distance, and driving speed, respectively, to construct a comprehensive trajectory evaluation function. The AGV uses LiDAR to detect dynamic obstacles and unknown obstacles in the workshop in real time.

[0086] Step 5: The AGV performs material handling tasks according to the optimized global planning path. During the journey, the LiDAR collects information about the surrounding environment at a frequency of 10Hz. When dynamic obstacles such as forklifts or workers are detected, the improved DWA algorithm is immediately triggered to select the optimal speed combination from the speed sampling space, generate a local obstacle avoidance trajectory, and guide the AGV back to the global planning path after obstacle avoidance is completed. If no obstacles are detected, the AGV continues to travel along the global path until it reaches the unloading point.

[0087] Step 6: Conduct 30 static and dynamic simulation experiments in the industrial workshop environment, and simultaneously carry out on-site actual tests. Record performance indicators such as path length, trajectory smoothness, minimum safe distance, and planning time. Compare with traditional algorithms such as RRT+DWA and A*+DWA to verify the superiority of the method of this invention. Optimize and fine-tune parameters such as the weight coefficient and number of iterations of the algorithm based on the experimental and test results.

[0088] The experimental analysis is as follows: To verify the effectiveness of the fusion algorithm of this invention for path planning of mobile robots in complex environments, simulation experiments were conducted in this embodiment. To ensure that the algorithm is applicable to different environments, this embodiment performed 30 static path planning simulations in simple environments, industrial workshop environments, and complex multi-obstacle environments, respectively, and compared the average path length, smoothness, and minimum safety distance of the paths planned by the fusion algorithm of this invention, RRT*+ACO+DWA, A*+DWA, and RRT*+APF+DWA algorithms.

[0089] The kinematic and dynamic parameters of the AGV in the experiment were consistent with those of the commonly used material handling AGVs in actual industrial scenarios. The specific parameter settings are shown in Table 1. Table 1 Mobile Robot Parameters

[0090] To comprehensively verify the algorithm's performance under different complexity environments, this experiment sets up three types of static test environments and two types of dynamic test environments, as follows: Simple static environment: The map size is 14m×14m, and multiple equally sized circular static obstacles are distributed in the environment. The obstacles are distributed regularly and there are no narrow passages. This is used to verify the basic path planning ability of the algorithm. Static environment of industrial workshop: The map size is 14m×14m. Multiple rectangular obstacles are distributed in the environment to simulate industrial equipment such as machine tools and shelves. There are multiple narrow passages and continuous obstacle areas, which are close to the AGV operation scenario of the actual industrial workshop. Complex multi-obstacle static environment: The map size is 13m×13m. The obstacles in the environment include various shapes such as circles and rectangles, with large size differences and irregular spatial distribution. There are dense obstacle areas and U-shaped trap areas. It is used to verify the algorithm's path planning ability in extremely complex environments. Dynamic environment of industrial workshop: Based on the static environment of industrial workshop, three static obstacles with unknown locations and one dynamic obstacle moving along a fixed straight line trajectory are set up to verify the algorithm's real-time obstacle avoidance capability for unknown obstacles and dynamic obstacles. Complex multi-obstacle dynamic environment: Based on a complex multi-obstacle static environment, a static obstacle with an unknown location and a dynamic obstacle moving along a fixed straight line trajectory are set up to verify the stability of the algorithm's dynamic path planning in complex environments.

[0091] To highlight the performance advantages of the algorithm of this invention, three currently mainstream global-local fusion path planning algorithms were selected as a comparison group, as follows: Algorithm a: RRT*+APF+DWA fusion algorithm, which combines traditional RRT* with the basic artificial potential field method to achieve global planning, and combines the original DWA algorithm to achieve local obstacle avoidance; Algorithm b: A*+DWA fusion algorithm, which uses the A* algorithm to achieve global path planning and combines it with the original DWA algorithm to achieve local obstacle avoidance; Algorithm c: A fusion algorithm of RRT*, ACO and DWA, which optimizes the initial path of RRT* through ant colony algorithm and combines it with the original DWA algorithm to achieve local obstacle avoidance; Algorithm d: The improved RRT*+MOPSO+improved DWA fusion algorithm proposed in this invention.

[0092] To achieve a quantitative comparison of algorithm performance, this experiment selected three core evaluation metrics. Each metric was tested through 30 repeated experiments, and the average value was taken to ensure the reliability of the experimental results. The specific definitions of the metrics are as follows: Path length: The total length of the AGV's travel trajectory from the starting point to the target point, in meters. The smaller this value, the higher the AGV's travel efficiency. Path smoothness: The cumulative value of the turning angles at all nodes of the AGV's driving trajectory. The smaller this index is, the fewer the trajectory turning points, the higher the smoothness, and the better the AGV's driving stability. Minimum safe distance: The minimum distance from all points on the AGV's travel trajectory to the nearest obstacle boundary, in meters. The larger this indicator is, the more sufficient the safety margin between the AGV and the obstacle, and the higher the obstacle avoidance safety.

[0093] For three types of static environments, 30 path planning simulation experiments were conducted for each of the four algorithms. After removing abnormal data that failed to plan, the average value of each performance evaluation index was taken to obtain the algorithm performance comparison results under different environments.

[0094] Table 2 below shows a comparison of the performance metrics of the four algorithms under a simple static environment: Table 2 Comparison of Simple Environmental Data Indicators

[0095] Table 3 below shows a comparison of the performance metrics of the four algorithms under static industrial workshop conditions: Table 3 Comparison of Environmental Data Indicators in Industrial Workshops

[0096] Table 4 below shows a comparison of the performance metrics of the four algorithms in a complex, multi-obstacle static environment: Table 4 Comparison of Data Indicators in Complex and Multi-Barrier Environments

[0097] To verify the real-time obstacle avoidance capability and global path tracking capability of the algorithm of this invention for unknown obstacles and dynamic obstacles, simulation experiments were carried out in dynamic environments of industrial workshops and complex multi-obstacle dynamic environments, focusing on testing the algorithm's dynamic path adjustment capability and obstacle avoidance success rate.

[0098] In this experiment, based on the static environment of an industrial workshop, three static obstacles with unknown locations were set up with coordinates (5,7), (7,7), and (9,7), and obstacles with radii of 0.2m, 0.3m, and 0.3m, respectively. At the same time, one moving dynamic obstacle was set up with a radius of 0.25m, which moved back and forth at a constant speed along a straight line from coordinates (9,8) to (9,11) at a speed of 0.5m / s.

[0099] The experimental procedure and results are as follows: After the AGV starts, it generates a globally optimal path based on the improved RRT*+MOPSO algorithm and travels along the global path. When no unknown or dynamic obstacles are detected, the travel trajectory matches the globally planned path by more than 98%.

[0100] When the AGV travels into the obstacle detection range, the lidar identifies three unknown static obstacles in real time, immediately triggering the improved DWA algorithm to adjust the local path. Without deviating from the main direction of the global path, it smoothly bypasses the unknown obstacles and immediately returns to the globally planned path after obstacle avoidance.

[0101] During operation, the AGV detects dynamic obstacles moving in a straight line in real time. By improving the DWA algorithm, it predicts the trajectory of the obstacles in advance and generates a collision-free local obstacle avoidance trajectory. It safely avoids the dynamic obstacles during their movement without any sudden stops or significant trajectory deviations.

[0102] The results of 30 repeated experiments show that the algorithm of this invention achieves a 100% obstacle avoidance success rate in the dynamic environment of an industrial workshop. After avoiding unknown and dynamic obstacles, the average deviation of the AGV returning to the global path is less than 0.08m, and the average smoothness of the entire driving trajectory is 3.02, which is only 5.2% higher than that in the static environment, maintaining excellent trajectory smoothness.

[0103] In this experiment, a static obstacle with an unknown location is set up in a complex, multi-obstacle static environment with coordinates (2,2) and a radius of 0.3m. At the same time, a moving dynamic obstacle with a radius of 0.4m is set up, which moves in a uniform reciprocating motion along a straight line from coordinates (10,10) to (11,10) at a speed of 0.8m / s.

[0104] The experimental procedure and results are as follows: Even in complex environments with dense obstacles, the AGV can still generate high-quality global paths based on the algorithm of this invention, maintain a safe distance from obstacles in narrow passages, and there are no cases of path planning failure.

[0105] During its operation, the AGV accurately identified unknown static obstacles and smoothly avoided them in narrow spaces by using an improved DWA algorithm. There were no collisions with surrounding obstacles, and the trajectory did not change drastically during the obstacle avoidance process.

[0106] For dynamic obstacles moving at high speeds, the algorithm of this invention can still predict the trajectory in advance, generate the optimal obstacle avoidance trajectory, safely avoid the obstacle during its movement, and quickly return to the globally planned path after obstacle avoidance.

[0107] The results of 30 repeated experiments show that the obstacle avoidance success rate of the algorithm of this invention reaches 100% in complex multi-obstacle dynamic environments, while the obstacle avoidance success rates of the comparative algorithms a, b, and c are 90%, 76.7%, and 83.3%, respectively. The dynamic obstacle avoidance reliability of the algorithm of this invention is significantly better than that of the comparative algorithms. At the same time, the average smoothness of the AGV's entire driving trajectory is 3.11, which is only 9.1% higher than that in the static environment, and the global path average fit reaches more than 95%, maintaining excellent path planning stability in complex dynamic environments.

[0108] Example 4: This embodiment provides an AGV dynamic path planning system for complex environments, including: The environment modeling and parameter configuration module is used to build an environmental map for AGV operations, set the starting point and target endpoint of AGV operations, and configure the kinematic parameters of AGVs. The initial path generation module is used to generate an initial globally feasible path for the AGV based on the improved RRT* algorithm. The improved RRT* algorithm includes: calculating the gravitational potential field of the target point to the AGV and the repulsive potential field of the obstacle to the AGV, constructing an artificial potential field composite force field model and performing node direction guidance. The global path planning module is used to optimize the initial globally feasible path using the MOPSO algorithm to obtain the final globally planned path. The path execution and collaborative control module is used to acquire real-time environmental perception data during the AGV's travel. When an obstacle is detected, the local path is dynamically adjusted based on the improved dynamic window method. After obstacle avoidance, the AGV is guided back to the final global planned path. If no obstacle is detected, the AGV continues to travel along the global planned path until it reaches the target destination. The dynamic adjustment process of the local path includes: generating a dynamic speed window based on AGV hardware performance boundary constraints, AGV motor acceleration and deceleration performance constraints, and AGV driving safety distance constraints; constructing a trajectory evaluation function; traversing all speed combinations within the dynamic speed window; and selecting the optimal speed combination of the trajectory evaluation function to generate a local obstacle avoidance trajectory.

[0109] Some steps in the embodiments of the present invention can be implemented using software, and the corresponding software program can be stored in a readable storage medium, such as an optical disc or a hard disk.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dynamic path planning method for AGVs in complex environments, characterized in that, The method includes: Step 1: Construct an environmental map for AGV operations, set the starting point and target endpoint of AGV operations, and configure the kinematic parameters of AGVs; Step 2: Generate the initial globally feasible path for the AGV based on the improved RRT* algorithm; The improved RRT* algorithm includes: calculating the gravitational potential field of the target point to the AGV and the repulsive potential field of the obstacle to the AGV, constructing an artificial potential field composite force field model and performing node direction guidance; Step 3: Optimize the initial globally feasible path using the MOPSO algorithm to obtain the final globally planned path; Step 4: Acquire real-time environmental perception data during AGV's travel. When an obstacle is detected, dynamically adjust the local path based on the improved dynamic window method. After obstacle avoidance, guide the AGV back to the final global planned path. If no obstacle is detected, continue traveling along the global planned path until the target destination is reached. The dynamic adjustment process of the local path includes: generating a dynamic speed window based on AGV hardware performance boundary constraints, AGV motor acceleration and deceleration performance constraints, and AGV driving safety distance constraints; constructing a trajectory evaluation function; traversing all speed combinations within the dynamic speed window; and selecting the optimal speed combination of the trajectory evaluation function to generate a local obstacle avoidance trajectory.

2. The AGV dynamic path planning method in complex environments according to claim 1, characterized in that, The processing procedure for the artificial potential field composite force field model includes: Define the gravitational potential field function of the target point with respect to the AGV, as follows: in, This is the gravitational gain coefficient. The Euclidean distance between the current position of the AGV and the target point position; The gravitational force generated at the AGV by this gravitational potential field is the negative gradient of the potential field function, expressed as: Define the repulsive potential field function of obstacles to AGV. Introduce a target point distance correlation term into the traditional repulsive potential field to obtain an improved repulsive potential field function, the expression of which is: in The repulsive gain coefficient is... This represents the Euclidean distance between the current position of the AGV and the position of the obstacle. The maximum effective distance of the repulsive force of the obstacle; The repulsive force generated at the AGV by the improved repulsive potential field is the negative gradient of the potential field function, which can be decomposed into components in two directions, expressed as: The net force acting on an AGV in a combined potential field is the vector sum of attractive and repulsive forces, expressed as: 。 3. The AGV dynamic path planning method in complex environments according to claim 1, characterized in that, The objective function for path optimization using the MOPSO algorithm in step 3 is determined based on the path length evaluation function, the trajectory smoothness evaluation function, and the driving safety evaluation function. The path length evaluation function calculates the sum of the Euclidean distances of all adjacent nodes on the path, the trajectory smoothness evaluation function calculates the cumulative value of the steering angles of all adjacent nodes on the path, and the driving safety evaluation function calculates the minimum distance from all nodes on the path to the nearest obstacle.

4. The AGV dynamic path planning method in complex environments according to claim 1, characterized in that, The process of determining the velocity sampling space using the improved dynamic window method includes: Based on the AGV hardware performance boundary constraints, the first constraint space for speed sampling is determined: in v For linear velocity, ω Angular velocity; Based on the acceleration and deceleration performance constraints of the AGV motor, the second constraint space for speed sampling is determined: in and Δ represents the current linear velocity and angular velocity of the AGV. t The sampling period; Based on the AGV's safe travel distance constraint, the third constraint space for speed sampling is determined: in To predict the shortest distance from the AGV to the obstacle on the trajectory. Considering the above three types of constraints, the final speed sampling space of the AGV is: .

5. The AGV dynamic path planning method in complex environments according to claim 1, characterized in that, The expression for the trajectory evaluation function is: Among them, the heading angle evaluation function The safe distance evaluation function represents the degree of matching between the heading deviation of the predicted trajectory endpoint and the target point. The velocity evaluation function represents the minimum distance between the predicted trajectory and the obstacle. This indicates the magnitude of the AGV's linear speed. α , β , γ These are the weighting coefficients for the three indicators.

6. The AGV dynamic path planning method in complex environments according to claim 5, characterized in that, The method further includes: normalizing the heading angle evaluation function, safety distance evaluation function, and speed evaluation function before constructing the trajectory evaluation function.

7. The AGV dynamic path planning method in complex environments according to claim 5, characterized in that, The kinematic parameters of the AGV include: maximum linear velocity, minimum linear velocity, maximum angular velocity, minimum angular velocity, maximum linear acceleration, and maximum angular acceleration.

8. A dynamic path planning system for AGVs in complex environments, characterized in that, The system includes: The environment modeling and parameter configuration module is used to build an environmental map for AGV operations, set the starting point and target endpoint of AGV operations, and configure the kinematic parameters of AGVs. The initial path generation module is used to generate an initial globally feasible path for the AGV based on the improved RRT* algorithm. The improved RRT* algorithm includes: calculating the gravitational potential field of the target point to the AGV and the repulsive potential field of the obstacle to the AGV, constructing an artificial potential field composite force field model and performing node direction guidance. The global path planning module is used to optimize the initial globally feasible path using the MOPSO algorithm to obtain the final globally planned path. The path execution and collaborative control module is used to acquire real-time environmental perception data during the AGV's travel. When an obstacle is detected, the local path is dynamically adjusted based on the improved dynamic window method. After obstacle avoidance, the AGV is guided back to the final global planned path. If no obstacle is detected, the AGV continues to travel along the global planned path until it reaches the target destination. The dynamic adjustment process of the local path includes: generating a dynamic speed window based on AGV hardware performance boundary constraints, AGV motor acceleration and deceleration performance constraints, and AGV driving safety distance constraints; constructing a trajectory evaluation function; traversing all speed combinations within the dynamic speed window; and selecting the optimal speed combination of the trajectory evaluation function to generate a local obstacle avoidance trajectory.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the AGV dynamic path planning method in complex environments as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the AGV dynamic path planning method in complex environments as described in any one of claims 1 to 7.

Images