Robot Path Navigation Method Based on Improved DWA Algorithm

Abstract

The present invention belongs to the technical field of path navigation, and particularly relates to a robot path navigation method based on an improved DWA algorithm, including: collecting environmental input information related to the current navigation task; generating a global optimal path composed of multiple discrete path points; introducing CMA-ES into the DWA algorithm; designing upper-layer optimization and lower-layer optimization of CMA-ES; introducing a feedback adjustment mechanism; in each planning cycle, constructing a trajectory candidate set; setting a trajectory evaluation function, and scoring each trajectory using the optimized weight parameters, regenerating the trajectory candidate set and scoring again after optimization, and selecting the trajectory with the highest score as the optimal path for the current cycle and sending it to the robot for navigation execution; and looping until the robot reaches the target point. The present invention realizes the real-time online adjustment of the key parameters of the DWA algorithm, and significantly improves the adaptability, smoothness, safety and planning accuracy of path planning in a dynamic environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of path navigation, and particularly relates to a robot path navigation method based on an improved DWA algorithm. Background Art

[0002] With the wide application of autonomous navigation systems such as service robots and intelligent vehicles, path planning algorithms have become their key supporting technologies. The core goal of path planning is to generate a safe, efficient, and smooth path for a robot or an intelligent vehicle in a complex dynamic environment, enabling it to reach the target point from the starting point while avoiding obstacles and adapting to environmental changes. In recent years, significant progress has been made in the research of path planning algorithms, but many challenges still remain, especially the real-time and adaptability issues in dynamic obstacle environments.

[0003] Path planning algorithms can be roughly divided into two categories: global path planning and local path planning. Global path planning generates a complete path from the starting point to the target point based on prior map information, while local path planning focuses on real-time obstacle avoidance and dynamic adjustment. As a classic local path planning method, the Dynamic Window Approach (DWA) is widely used in mobile robots and intelligent vehicles due to its strong real-time performance and dynamic obstacle avoidance ability.

[0004] The DWA algorithm searches for feasible speed combinations in the speed space, predicts the trajectory of the robot within a future time window, and selects the optimal trajectory according to a path scoring function. Its main advantages are real-time performance and dynamic obstacle avoidance ability. However, the DWA algorithm also has the following limitations:

[0005] Parameter dependence on manual experience: Multiple key parameters in the DWA algorithm (such as various weight parameters in the scoring function) usually rely on manual experience for setting, lacking environmental adaptability. Limited optimization scope: Most existing optimization methods only adjust some of the weights in the scoring function, without considering the joint adjustment of search parameters (such as linear velocity resolution and angular velocity resolution) and planning accuracy. Insufficient adaptability to dynamic scenarios: Existing methods mostly adopt an offline training method and only work in static environments, unable to meet the real-time optimization requirements in dynamic scenarios.

[0006] In order to overcome the limitations of the DWA algorithm, researchers have tried to introduce optimization methods such as genetic algorithms (GA) and particle swarm optimization (PSO) to adjust DWA parameters. However, these methods still have the following problems:

[0007] Limitations of offline optimization: Most optimization methods adopt offline training methods and cannot adapt to the changes in dynamic environments in real time. Single optimization objective: Existing methods only optimize some weights in the scoring function and do not comprehensively consider the multi-objective requirements of path planning (such as goal directivity, smoothness, safety, etc.). Low computational efficiency: Some optimization algorithms have low computational efficiency in complex dynamic environments and are difficult to meet the real-time requirements.

[0008] Therefore, there is an urgent need for an improved path planning algorithm with online learning ability, wide optimization scope, and applicable to dynamic obstacle environments. Summary of the Invention

[0009] In view of the above deficiencies in the prior art, the object of the present invention is to provide a robot path navigation method based on an improved DWA algorithm. Through a hierarchical optimization structure based on CMA-ES, real-time online adjustment of the key parameters of the DWA algorithm is achieved, and the adaptability, smoothness, safety, and planning accuracy of path planning are significantly improved in a dynamic environment.

[0010] To achieve the above object, the present invention provides a robot path navigation method based on an improved DWA algorithm, including the following steps:

[0011] S1. Collect environmental input information related to the current navigation task, including the current pose information and speed state of the robot, the position information of the target point, and obstacle information;

[0012] S2. Generate a globally optimal path from the current position of the robot to the target point through a path search algorithm. The globally optimal path consists of multiple discrete path points;

[0013] S3. Introduce the covariance matrix adaptation evolution strategy CMA-ES into the DWA algorithm, set the kinematic boundary parameters of the DWA, and initialize the control parameters of the CMA-ES optimizer;

[0014] S4. Design the upper-layer optimization of CMA-ES, combine the heading angle evaluation, smoothness evaluation, speed evaluation, and safety evaluation to form an upper-layer optimization function, and optimize the weight parameters of the trajectory evaluation function in the DWA algorithm;

[0015] S5. Design the lower-layer optimization of CMA-ES, combine the window calculation efficiency evaluation, smoothness evaluation, speed evaluation, and safety evaluation to form a lower-layer optimization function, and optimize the linear velocity resolution and angular velocity resolution in the DWA algorithm;

[0016] S6. Introduce a feedback adjustment mechanism to dynamically adjust the update frequency of the upper-layer optimization according to environmental changes, and the lower-layer optimization adjusts the trajectory accuracy in real time;

[0017] S7. In each planning cycle, based on the current speed state, use the optimized linear speed resolution and angular speed resolution to sample and generate speed combinations, predict their trajectories within a future time window, and construct a trajectory candidate set;

[0018] S8. Introduce a path guidance mechanism, set a trajectory evaluation function, score each trajectory using the optimized weight parameters, execute the CMA-ES optimizer to search for the optimal parameters, regenerate the trajectory candidate set after optimization, score again, and select the trajectory with the highest score as the optimal path for the current cycle and send it to the robot for navigation execution;

[0019] S9. S4 - S8 are executed in a fixed cycle, continuously generating trajectories, optimizing parameters, scoring paths, and controlling execution until the robot reaches the target point.

[0020] As a preferred solution of the present invention, in S1, the current pose information of the robot , , is the position coordinate of the robot in the global coordinate system, is the heading angle of the current pose of the robot; the speed state includes the current speed v and acceleration of the robot, that is, , the speed v includes linear speed and angular speed, and the acceleration includes linear acceleration and angular acceleration; the position information of the target point , , are the position coordinates of the target point in the global coordinate system; the obstacle information includes the positions and distributions of known and unknown obstacles in the environment. The number of known obstacles is n, and the number of unknown obstacles is m. The unknown obstacles include unknown dynamic and static obstacles, represents the position coordinate of the known obstacle u1 in the global coordinate system, represents the position coordinate of the unknown obstacle U1 measured by the robot in the global coordinate system, and the same applies to the remaining elements in;

[0021] The environmental input information is expressed as:

[0022] .

[0023] As a preferred solution of the present invention, in S2, the path search algorithm adopts one of the A* algorithm, Dijkstra algorithm, RRT algorithm, and PRM algorithm. The globally optimal path , Z is the number of discrete path points, Denote one of the discrete path points, where z = 1, 2, …, Z, and denote the position coordinates of the discrete path point in the global coordinate system.

[0024] As a preferred solution of the present invention, in the above-mentioned S3, the kinematic boundary parameters of DWA include the maximum linear velocity, the maximum angular velocity, the maximum acceleration, and the control period, and the initial linear velocity resolution and angular velocity resolution are set;

[0025] Initialize the control parameters of the CMA-ES optimizer, including the initial mean vector , the covariance matrix , the sampling step size , the population size , the number of parents and the maximum number of iterations , to form the initialization vector of the CMA-ES optimizer:

[0026] .

[0027] As a preferred solution of the present invention, in the above-mentioned S4, the optimization objectives of the upper-layer CMA-ES optimization include directivity, smoothness, and safety. By adjusting the heading angle weight parameter, safety weight parameter, and speed weight parameter of the trajectory evaluation function, the path planning performance is optimized in real time according to the environmental changes;

[0028] Design the heading angle evaluation , smoothness evaluation , speed evaluation and safety evaluation respectively, which are expressed as:

[0029] ;

[0030] ;

[0031] ;

[0032] ;

[0033] In the formula, is the heading angle of the current pose of the robot; denotes the heading angle of the target point relative to the position of the robot; denotes the angular change of the robot within the simulation time; denotes the simulation time of the robot; denotes the maximum angular velocity resolution of the robot; denotes the current linear velocity on the trajectory; represents the set maximum linear velocity; is a constant that controls the steepness of the logic curve; represents a point on the trajectory the minimum distance from the nearest obstacle; represents the set safety critical distance; C is the total number of trajectory points; t is the time integration variable;

[0034] the upper - layer optimization function, i.e., the upper - layer comprehensive evaluation function is expressed as:

[0035] ;

[0036] In the formula, , , , are respectively the weight parameters of the course - angle evaluation, smoothness evaluation, speed evaluation, and safety evaluation in the upper - layer optimization. The optimization objective of the upper - layer optimization , , , respectively represent the course - angle weight parameter, safety weight parameter, and speed weight parameter in the trajectory evaluation function.

[0037] As a preferred embodiment of the present invention, in the S5, the CMA - ES lower - layer optimization takes the linear velocity resolution and angular velocity resolution as the optimization objectives to improve the trajectory accuracy and evaluate the window calculation efficiency is expressed as:

[0038] ;

[0039] In the formula, represents the linear velocity resolution of the robot; represents the angular velocity resolution of the robot; represents the maximum linear velocity resolution of the robot;

[0040] the lower - layer optimization function, i.e., the lower - layer comprehensive evaluation function is expressed as:

[0041] ;

[0042] In the formula, , , , are respectively the weight parameters of the window calculation efficiency evaluation, smoothness evaluation, speed evaluation, and safety evaluation in the lower - layer optimization; the optimization objective of the lower - layer optimization .

[0043] As a preferred embodiment of the present invention, in step S6, through the coordination between the upper-layer optimization and the lower-layer optimization, the collaborative optimization of the weight parameters and the trajectory generation resolution is achieved. A feedback adjustment mechanism is introduced to dynamically adjust the update frequency of the upper-layer optimization according to environmental changes, while the lower-layer optimization adjusts the trajectory accuracy in real time. Define the index of the cycle as k, and the environmental obstacle change metric at the k-th cycle is expressed as:

[0044] ;

[0045] In the formula, , , respectively represent the obstacle areas in the neighborhoods with radii of , , at the k-th cycle; , , are three different set radii, and ; , , are the total areas of the neighborhoods with radii of , , at the k-th cycle; , , are the weight parameters corresponding to , , respectively;

[0046] Define as the parameter optimization adjustment coefficient of the upper-layer optimization at the k-th cycle, which controls the update frequency of the upper-layer optimization:

[0047] ;

[0048] In the formula, is the parameter optimization adjustment coefficient of the upper-layer optimization at the (k - 1)-th cycle; is the environmental change adjustment coefficient, which controls the degree of change of the update frequency of the upper-layer optimization; is the environmental obstacle change metric at the (k - 1)-th cycle.

[0049] As a preferred embodiment of the present invention, in step S7, in each planning cycle, based on the current speed state of the robot, the optimized linear velocity resolution and angular velocity resolution are used to sample in the allowed speed space to generate a number of speed combinations; for each group of speed combinations, predict their trajectories within a future time window to construct a trajectory candidate set :

[0050] ;

[0051] Among them, represents the i-th trajectory, where i = 1, 2, …, N, and N is the number of trajectories; each trajectory represents an expected motion path under a given control input.

[0052] As a preferred solution of the present invention, in S8, during the scoring process of the trajectory, a path guidance mechanism is introduced. By calculating the deviation degree between each trajectory and the global optimal path, that is, the average value of the shortest distance from the trajectory points to the global path, as an additional evaluation index; this deviation degree is used to correct the original fitness ranking, guiding the trajectory to approach the global path direction, so as to improve the continuity and navigation accuracy of the overall path while maintaining the local obstacle avoidance ability. The deviation degree function is expressed as:

[0053] ;

[0054] In the formula, represents the deviation degree function of; W is the total number of sampling points of the trajectory points; represents the position at the w-th sampling point.

[0055] As a preferred solution of the present invention, in S8, a trajectory evaluation function is set and each trajectory is scored using the optimized weight parameters, and the CMA-ES optimizer is executed to search for the optimal parameters; after the optimization is completed, based on the obtained linear velocity resolution and angular velocity resolution, a trajectory candidate set is regenerated, and each trajectory is scored and sorted again. The trajectory with the highest score is selected as the optimal path for the current cycle, and its trajectory velocity state is sent to the robot controller as a control instruction for execution, so as to achieve closed-loop control and continuous path tracking;

[0056] The set trajectory evaluation function is expressed as:

[0057] ;

[0058] In the formula, represents the trajectory evaluation function of; , , are respectively the heading angle evaluation, safety evaluation, and speed evaluation corresponding to ; is the adjustment parameter for path guidance.

[0059] The algorithm involved in the present invention can be executed by an electronic device provided on the robot. The electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. The above algorithm calculation is implemented by the processor executing the software.

[0060] The beneficial effects of the present invention are as follows:

[0061] The present invention proposes an improved DWA algorithm based on the covariance matrix adaptation evolution strategy (CMA-ES), which realizes the collaborative optimization of the weight parameters of the path scoring function and the trajectory generation parameters through a hierarchical optimization structure; this algorithm can adjust the parameters in real time to adapt to environmental changes, significantly improving the adaptability, smoothness and obstacle avoidance performance of path planning. The upper-layer optimization dynamically adjusts the key weight parameters of the path scoring function through CMA-ES to ensure the effectiveness and accuracy of the path under different environmental conditions; the lower-layer optimization adjusts the speed and angular velocity resolution in real time to ensure that the robot can flexibly respond to environmental changes.

[0062] The present invention introduces a non-linear safety penalty mechanism and a global path guidance strategy, further improving the safety and goal orientation of path planning. The simulation results show that compared with the traditional DWA algorithm, the improved algorithm has significant improvements in key indicators such as success rate, path length and average speed, has good real-time performance and robustness, and is suitable for local path planning tasks in a dynamic obstacle environment. Brief Description of the Drawings

[0063] Figure 1 is the process schematic diagram of the present invention;

[0064] Figure 2 is the experimental result diagram of the obstacle avoidance path test during the verification process of the present invention. Detailed Embodiment

[0065] The following further describes the embodiments of the present invention with reference to the drawings:

[0066] As Figure 1 shown, the robot path navigation method based on the improved DWA algorithm includes the following steps:

[0067] S1. Collect environmental input information related to the current navigation task, including the current pose information and speed state of the robot, the position information of the target point, and obstacle information;

[0068] S2. Generate a globally optimal path from the current position of the robot to the target point through a path search algorithm, and the globally optimal path consists of multiple discrete path points;

[0069] S3. Introduce the covariance matrix adaptation evolution strategy CMA-ES into the DWA algorithm, set the kinematic boundary parameters of the DWA, and initialize the control parameters of the CMA-ES optimizer;

[0070] S4. Design the upper-layer optimization of CMA-ES, combine the course angle evaluation, smoothness evaluation, speed evaluation, and safety evaluation to form an upper-layer optimization function, and optimize the weight parameters of the trajectory evaluation function in the DWA algorithm;

[0071] S5. Design the lower-layer optimization of CMA-ES, combine the window calculation efficiency evaluation, smoothness evaluation, speed evaluation, and safety evaluation to form a lower-layer optimization function, and optimize the linear velocity resolution and angular velocity resolution in the DWA algorithm;

[0072] S6. Introduce a feedback adjustment mechanism to dynamically adjust the update frequency of the upper-layer optimization according to environmental changes, and the lower-layer optimization adjusts the trajectory accuracy in real time;

[0073] S7. In each planning cycle (abbreviated as cycle, which is the frequency at which the DWA algorithm updates its speed and direction decisions), based on the current speed state, use the optimized linear velocity resolution and angular velocity resolution to sample and generate speed combinations, and predict their trajectories within a future time window to construct a trajectory candidate set;

[0074] S8. Introduce a path guidance mechanism, set a trajectory evaluation function, and use the optimized weight parameters to score each trajectory. Execute the CMA-ES optimizer to search for the optimal parameters. After the optimization is completed, regenerate the trajectory candidate set and score again. Select the trajectory with the highest score as the optimal path for the current cycle and send it to the robot for navigation;

[0075] S9. S4 - S8 are executed in a fixed cycle, continuously generating trajectories, optimizing parameters, scoring paths, and controlling executions until the robot reaches the target point.

[0076] In S1, the current pose information of the robot , , is the position coordinate of the robot in the global coordinate system, is the course angle of the current pose of the robot; the speed state includes the current speed v and acceleration of the robot, that is, , the speed v includes the linear velocity and angular velocity, and the acceleration includes the linear acceleration and angular acceleration; the position information of the target point , , is the position coordinate of the target point in the global coordinate system; the obstacle information includes the positions and distributions of known and unknown obstacles in the environment. The number of known obstacles is n, and the number of unknown obstacles is m. The unknown obstacles include unknown dynamic and static obstacles, represents the position coordinate of the known obstacle u1 in the global coordinate system, Indicates the position coordinates of the unknown obstacle U1 measured by the robot in the global coordinate system, and the same applies to the remaining elements in

[0077] Environmental input information is expressed as:

[0078] .

[0079] In S2, the path search algorithm adopts one of the A* algorithm, Dijkstra algorithm, RRT algorithm, and PRM algorithm. The globally optimal path , where Z is the number of discrete path points, represents one of the discrete path points, and z = 1, 2,..., Z, , represents the position coordinates of the discrete path point in the global coordinate system. The globally optimal path provides a guiding basis in the local path scoring stage, guiding the trajectory to approach this global path during the optimization process, thereby enhancing the coherence and goal - orientation of the path.

[0080] In S3, the kinematic boundary parameters of DWA include the maximum linear velocity, maximum angular velocity, maximum acceleration, and control period, which are used to define the boundaries of the velocity search space; and the initial linear velocity resolution and angular velocity resolution are set to control the accuracy and density of trajectory generation;

[0081] Initialize the control parameters of the CMA - ES optimizer, including the initial mean vector , covariance matrix , sampling step size , population size , number of parents and maximum number of iterations , which constitute the initialization vector of the CMA - ES optimizer:

[0082] .

[0083] In the CMA - ES optimizer, represents the mean of the parameters at the initial stage of the algorithm, usually set to the center of the parameter space or the prior estimate value; describes the shape and spread of the individual parameter distribution in the population, and is used to guide the mutation operation during the search process to explore the parameter space; determines the step size when sampling new individuals from the distribution defined by the covariance matrix. The larger the step size, the wider the search range, but it may miss fine - grained local searches; is the number of new individuals (candidate solutions) generated in each iteration; It represents the number of optimal individuals selected from the current population for updating the individuals in the next generation. These individuals are usually the optimal individuals evaluated according to the fitness function.

[0084] In S4, the optimization objectives of the upper-layer optimization of CMA-ES include directivity, smoothness, and safety. By adjusting the heading angle weight parameter, safety weight parameter, and speed weight parameter of the trajectory evaluation function, the path planning performance is optimized in real time according to the environmental changes;

[0085] Design the heading angle evaluation respectively , smoothness evaluation , speed evaluation and safety evaluation , which are expressed as:

[0086] ;

[0087] ;

[0088] ;

[0089] ;

[0090] In the formula, is the heading angle of the current posture of the robot; represents the heading angle of the target point relative to the position of the robot; represents the angular change of the robot within the simulation time (the time length considered when predicting the future motion trajectory of the robot, such as 0.1 second); represents the simulation time of the robot; represents the maximum angular velocity resolution of the robot; represents the current linear velocity on the trajectory; represents the set maximum linear velocity; is a constant that controls the steepness of the logic curve; represents the point on the trajectory the minimum distance from the nearest obstacle; represents the set safety critical distance; C is the total number of trajectory points; t is the time integration variable;

[0091] The upper-layer optimization function, that is, the upper-layer comprehensive evaluation function is expressed as:

[0092] ;

[0093] In the formula, , , , They are the weight parameters for the heading angle evaluation, smoothness evaluation, speed evaluation, and safety evaluation in the upper-layer optimization respectively, which are used to balance the importance of different factors in the comprehensive evaluation function in multi-objective optimization; the optimization objective of the upper-layer optimization , 、 、 respectively represent the heading angle weight parameter, safety weight parameter, and speed weight parameter in the trajectory evaluation function.

[0094] In S5, the CMA-ES lower-layer optimization takes the linear velocity resolution and angular velocity resolution as the optimization objectives to improve the trajectory accuracy and evaluate the window calculation efficiency It is expressed as:

[0095] ;

[0096] In the formula, represents the linear velocity resolution of the robot; represents the angular velocity resolution of the robot; represents the maximum linear velocity resolution of the robot;

[0097] The lower-layer optimization function, that is, the lower-layer comprehensive evaluation function It is expressed as:

[0098] ;

[0099] In the formula, 、 、 、 are the weight parameters for the window calculation efficiency evaluation, smoothness evaluation, speed evaluation, and safety evaluation in the lower-layer optimization respectively; the optimization objective of the lower-layer optimization .

[0100] Through the heterogeneous cycle scheduling between the upper-layer optimization (low-frequency update) and the lower-layer optimization (high-frequency update), the collaborative optimization of the scoring parameters and trajectory generation is achieved.

[0101] In S6, through the coordination between the upper-layer optimization and the lower-layer optimization, the collaborative optimization of the weight parameter optimization and trajectory generation resolution is achieved. A feedback adjustment mechanism is introduced to dynamically adjust the update frequency of the upper-layer optimization according to the environmental changes, while the lower-layer optimization adjusts the trajectory accuracy in real time to ensure a quick response to environmental changes. The index of the defined cycle is k, and the environmental obstacle change metric at the k-th cycle It is expressed as:

[0102] ;

[0103] In the formula, 、 、 respectively represent the obstacle areas in the neighborhoods with radii of , , at k cycles, , , are three different set radii, and ; , , represent the total neighborhood areas with radii of , , at k cycles; , , are the corresponding , , weight parameters;

[0104] The neighborhood is a circular area centered on the current position of the robot with radii of , , , and the radius size is set according to requirements.

[0105] Define as the parameter optimization adjustment coefficient for upper-layer optimization at k cycles, controlling the update frequency of upper-layer optimization:

[0106] ;

[0107] In the formula, is the parameter optimization adjustment coefficient for upper-layer optimization at k - 1 cycles; is the environmental change adjustment coefficient, controlling the degree of change in the update frequency of upper-layer optimization; is the measure of environmental obstacle change at k - 1 cycles.

[0108] If is larger, it means that the update step size of upper-layer optimization will change faster with the severity of environmental changes; if is smaller, it means that the adjustment of upper-layer optimization update is relatively slow.

[0109] In S7, in each planning cycle, based on the current speed state of the robot, using the optimized linear velocity resolution and angular velocity resolution, samples are taken in the allowed speed space to generate a number of speed combinations; for each group of speed combinations, their trajectories in a future time window are predicted to construct a trajectory candidate set :

[0110] ;

[0111] Among them, Denote the i-th trajectory, where i = 1, 2, …, N and N is the number of trajectories; each trajectory represents an expected motion path under a given control input.

[0112] In S8, during the scoring process of the trajectories, a path guidance mechanism is introduced. By calculating the deviation degree between each trajectory and the global optimal path, that is, the average value of the shortest distances from the trajectory points to the global path, as an additional evaluation index; this deviation degree is used to correct the original fitness ranking, guiding the trajectories to approach the global path direction, so as to improve the continuity and navigation accuracy of the overall path while maintaining the local obstacle avoidance ability. The deviation degree function is expressed as:

[0113] ;

[0114] In the formula, denotes the deviation degree function of ; W is the total number of sampling points of the trajectory points; denotes

[0115] the position of

[0116] at the w-th sampling point.

[0117] ;

[0118] In the formula, denotes the trajectory evaluation function of , , are respectively the heading angle evaluation, safety evaluation, and speed evaluation corresponding to ; is the adjustment parameter for path guidance.

[0119] The verification process is as follows:

[0120] Figure 2Shows the experimental results of the obstacle avoidance path test of the method in this embodiment in a complex dynamic environment. The experimental scenario is a two-dimensional planning area with a size of 50m × 50m. The starting point of the robot is at the lower left corner, and the target point is at the upper right corner. The environment contains various types of obstacles to simulate the multi-source obstacle situations that may be encountered in the actual scenario. Among them, the black squares represent known obstacles (static), which are set in advance during map construction; the black solid circles represent unknown static obstacles that appear during the experiment, and the robot needs to perceive them in real time (based on radar, etc.) and avoid them; the black hollow circles simulate unknown dynamic obstacles in the environment (obstacles that may move), and the path planning process needs to have a certain real-time response ability to avoid obstacles.

[0121] In Figure 2 it, the black-green dotted line trajectory represents the global optimal path, serving as a reference target for the robot to move forward; the red trajectory is the movement path formed by the robot during the actual execution process by combining the improved DWA algorithm and avoiding obstacles in real time; the purple line near the upper right corner represents the predicted trajectory; the red trajectory shows an offset near the obstacles compared with the black-green dotted line trajectory but still maintains a strong fit overall. This experiment verifies the adaptability and stability of the method proposed in this embodiment in a highly dynamic and mixed obstacle environment.

[0122] Table 1 Comparison between the improved DWA and the traditional DWA

[0123]

[0124] The performance comparison between the improved DWA and the traditional DWA in this embodiment is shown in Table 1. The experimental tests are divided into three rounds: A, B, and C. Each round of tests is executed 20 times, and the test environment is significantly adjusted in each round to simulate the actual complex environment changes. The results show that the improved DWA is superior to the traditional DWA in key indicators such as path success rate, average path length, and average speed: among them, the average success rate is increased from about 86.7% of the traditional method to about 98.3% of the improved method; the average path length is reduced by about 12.6%; the average path execution speed is increased by about 8.4%. These results verify the stability, efficiency, and practical adaptability of the improved method in different dynamic environments.

Claims

1. A robot path navigation method based on an improved DWA algorithm, characterized in that It includes the following steps: S1. Collect the environmental input information related to the current navigation task, including the current pose information and speed state of the robot, the position information of the target point, and the obstacle information; S2. Generate a globally optimal path from the current position of the robot to the target point through a path search algorithm. The globally optimal path consists of multiple discrete path points; S3. Introduce the covariance matrix adaptation evolution strategy CMA-ES into the DWA algorithm, set the kinematic boundary parameters of DWA, and initialize the control parameters of the CMA-ES optimizer; S4. Design the upper-layer optimization of CMA-ES, combine the heading angle evaluation, smoothness evaluation, speed evaluation, and safety evaluation to form an upper-layer optimization function, and optimize the weight parameters of the trajectory evaluation function in the DWA algorithm; S5. Design the lower-layer optimization of CMA-ES, combine the window calculation efficiency evaluation, smoothness evaluation, speed evaluation, and safety evaluation to form a lower-layer optimization function, and optimize the linear velocity resolution and angular velocity resolution in the DWA algorithm; S6. Introduce a feedback adjustment mechanism to dynamically adjust the update frequency of the upper-layer optimization according to environmental changes, and the lower-layer optimization adjusts the trajectory accuracy in real time; S7. In each planning cycle, based on the current speed state, use the optimized linear velocity resolution and angular velocity resolution to sample and generate speed combinations, and predict their trajectories within a future time window to construct a trajectory candidate set; S8. Introduce a path guidance mechanism, set a trajectory evaluation function, and use the optimized weight parameters to score each trajectory. Execute the CMA-ES optimizer to search for the optimal parameters. After the optimization is completed, regenerate the trajectory candidate set and score again. Select the trajectory with the highest score as the optimal path for the current cycle and send it to the robot for navigation execution; S9. S4 - S8 are executed in a fixed cycle, continuously generating trajectories, optimizing parameters, scoring paths, and executing controls until the robot reaches the target point.

2. The robot path navigation method based on the improved DWA algorithm according to claim 1, characterized in that: In the above-mentioned S1, the current pose information of the robot , 、 are the position coordinates of the robot in the global coordinate system, is the heading angle of the current pose of the robot; the speed state includes the current speed v and acceleration of the robot , that is , the speed v includes linear velocity and angular velocity, and the acceleration includes linear acceleration and angular acceleration; the position information of the target point , 、 are the position coordinates of the target point in the global coordinate system; the obstacle information includes the positions and distributions of known and unknown obstacles in the environment. The number of known obstacles is n, and the number of unknown obstacles is m. The unknown obstacles include unknown dynamic and static obstacles, represents the position coordinates of the known obstacle u1 in the global coordinate system, represents the position coordinates of the unknown obstacle U1 measured by the robot in the global coordinate system, The same applies to the remaining elements in Environmental input information Expressed as: 。 3. The robot path navigation method based on the improved DWA algorithm according to claim 1, characterized in that: In the above-mentioned S2, the path search algorithm adopts one of the A* algorithm, Dijkstra algorithm, RRT algorithm, and PRM algorithm, and the globally optimal path , where Z is the number of discrete path points, represents one of the discrete path points, and z = 1, 2,..., Z, , represents the position coordinates of the discrete path point in the global coordinate system.

4. The robot path navigation method based on the improved DWA algorithm according to claim 1, characterized in that: In the S3 described above, the kinematic boundary parameters of DWA include the maximum linear velocity, maximum angular velocity, maximum acceleration, and control period, and the initial linear velocity resolution and angular velocity resolution are set; Initialize the control parameters of the CMA-ES optimizer, including the initial mean vector , covariance matrix , sampling step size , population size , number of parents and maximum number of iterations , to form the initialization vector of the CMA-ES optimizer : 。 5. The robot path navigation method based on the improved DWA algorithm according to claim 3, characterized in that: In the S4 described above, the optimization objectives of the upper-layer optimization of CMA-ES include directivity, smoothness, and safety. By adjusting the heading angle weight parameter, safety weight parameter, and speed weight parameter of the trajectory evaluation function, the path planning performance is optimized in real time according to environmental changes; Separate design of course angle evaluation , smoothness evaluation , speed evaluation and safety evaluation , expressed as: ； ； ； ； Wherein, is the heading angle of the current posture of the robot; represents the heading angle of the target point relative to the position of the robot; represents the angular change of the robot within the simulation time; represents the simulation time of the robot; represents the maximum angular velocity resolution of the robot; represents the current linear velocity on the trajectory; represents the set maximum linear velocity; is a constant that controls the steepness of the logic curve; represents the point on the trajectory the minimum distance from the nearest obstacle; represents the set safety critical distance; C is the total number of trajectory points; t is the time integration variable; Upper layer optimization function, i.e., upper layer comprehensive evaluation function It is expressed as: ； In the formula, , , , are the weight parameters for the course angle evaluation, smoothness evaluation, speed evaluation, and safety evaluation in the upper-layer optimization respectively. The optimization objective of the upper-layer optimization is , , , represent the course angle weight parameter, safety weight parameter, and speed weight parameter in the trajectory evaluation function respectively.

6. The robot path navigation method based on the improved DWA algorithm according to claim 5, characterized in that: In the above-mentioned S5, the lower layer of CMA-ES is optimized with the linear velocity resolution and the angular velocity resolution as the optimization objectives to improve the trajectory accuracy and evaluate the window calculation efficiency It is expressed as: ； In the formula, represents the linear velocity resolution of the robot; represents the angular velocity resolution of the robot; represents the maximum linear velocity resolution of the robot; Lower layer optimization function, i.e., lower layer comprehensive evaluation function It is expressed as: ； Wherein, , , , are the weight parameters for window calculation efficiency evaluation, smoothness evaluation, speed evaluation, and security evaluation in the lower layer optimization respectively; Optimization objectives for lower-layer optimization 。 7. The robot path navigation method based on the improved DWA algorithm according to claim 6, characterized in that: In S6, through the coordination between upper-layer optimization and lower-layer optimization, the collaborative optimization of weight parameters and trajectory generation resolution is achieved. A feedback adjustment mechanism is introduced to dynamically adjust the update frequency of upper-layer optimization according to environmental changes, while the lower-layer optimization adjusts the trajectory accuracy in real time. Define the index of the period as k, and the change metric of environmental obstacles at the k-th period It is expressed as: ； Wherein, , , respectively represent the obstacle areas in the neighborhoods with radii of , , at the k-th period; , , are three different set radii, and ; , , are the total areas of the neighborhoods with radii of , , at the k-th period; , , are the weight parameters corresponding to , , respectively; Definition It is the parameter optimization adjustment coefficient for upper-layer optimization in the k cycle, which controls the update frequency of upper-layer optimization: ； In the formula, is the parameter optimization adjustment coefficient optimized by the upper layer in the (k - 1)th cycle; is the environmental change adjustment coefficient, which controls the degree of change in the upper layer optimization update frequency; is the environmental obstacle change metric in the (k - 1)th cycle.

8. The robot path navigation method based on the improved DWA algorithm according to claim 7, characterized in that: In the aforementioned S7, in each planning cycle, based on the current speed state of the robot, samples are taken in the allowable speed space using the optimized linear speed resolution and angular speed resolution to generate a number of speed combinations; for each group of speed combinations, its trajectory within a future time window is predicted to construct a trajectory candidate set : ； Among them, represents the i-th trajectory, where i = 1, 2, …, N and N is the number of trajectories; each trajectory represents an expected motion path under a given control input.

9. The robot path navigation method based on the improved DWA algorithm according to claim 8, characterized in that: In the S8 described above, during the scoring process of the trajectory, a path guidance mechanism is introduced. By calculating the deviation degree between each trajectory and the globally optimal path, that is, the average shortest distance from the trajectory point to the global path, as an additional evaluation index; this deviation degree is used to correct the original fitness ranking, guide the trajectory to approach the global path direction, so as to improve the continuity and navigation accuracy of the overall path while maintaining the local obstacle avoidance ability. The deviation degree function is expressed as: ； In the formula, represents the deviation function; W is the total number of sampling points of the trajectory points; represents the position at the w-th sampling point.

10. The robot path navigation method based on the improved DWA algorithm according to claim 9, characterized in that: In the above-mentioned S8, a trajectory evaluation function is set, and each trajectory is scored using the optimized weight parameters. The CMA-ES optimizer is executed to search for the optimal parameters. After the optimization is completed, based on the obtained linear velocity resolution and angular velocity resolution, a trajectory candidate set is regenerated, and each trajectory is scored and sorted again. The trajectory with the highest score is selected as the optimal path for the current cycle, and its trajectory velocity state is sent as a control command to the robot controller for execution, thereby realizing closed-loop control and continuous path tracking; The set trajectory evaluation function is expressed as: ； In the formula, represents the trajectory evaluation function; , , are respectively the heading angle evaluation, safety evaluation, and speed evaluation corresponding to ; is the adjustment parameter for path guidance.

Images