Robot path dynamic planning method and system

Abstract

The invention provides a robot path dynamic planning method and system in the technical field of robot path planning. The method comprises the following steps: S1, creating a robot path planning model comprising a first search strategy, a second search strategy, a third search strategy and a fourth search strategy based on a ripple diffusion algorithm; s2, creating a simulation environment, setting a plurality of dynamic obstacles and static obstacles in the simulation environment, and setting a motion track and a motion speed of each dynamic obstacle; s3, initializing a plurality of robots in the simulation environment, and training the robot path planning model based on the simulation environment and the initialized robots; s4, performing performance verification on the trained robot path planning model; and S5, deploying the robot path planning model passing the performance verification so as to perform path planning. The method has the advantage that the reliability and generalization ability of robot path planning are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of robot path planning, and in particular to a robot path dynamic planning method and system. Background Art

[0002] Robotics has been widely used in various fields. The versatility and adaptability of robots have driven continuous innovation, enabling them to handle increasingly complex tasks in dynamic and unstructured environments. Path planning refers to the robot's exploration of an optimal or feasible path from a starting point to a destination according to a certain metric (shortest distance, minimum time, etc.), and is an important part of the field of robot navigation. During the complete navigation process, the robot executes three parts in sequence: map construction and positioning, path planning, and motion control. First, a map is created based on sensor information. After the map is built, path planning can be performed, and finally the robot's motion control commands are output to enable it to drive safely along the planned path.

[0003] Traditional path planning methods, including Dijkstra's algorithm, A* algorithm, rapidly exploring random trees (RRT) algorithm, and artificial potential field method, have been proven to be effective in static or simple environments. However, the performance of traditional methods degrades in dynamic or complex environments. Balancing path optimality and computational efficiency becomes increasingly difficult in dynamic or complex environments. Dynamic environments also pose challenges such as real-time obstacle detection, non-stationary terrain, and the need for adaptive replanning, making traditional methods less suitable for practical applications. In autonomous driving, real-time path adjustment is crucial, especially when navigating in urban environments where road conditions, traffic signals, and pedestrian movements frequently change. With the development of science and technology, multi-agent interactions in urban traffic environments are becoming increasingly complex. In dynamic environments, traditional methods have certain limitations. They can only represent the position of dynamic obstacles, but cannot represent the speed and movement patterns of dynamic obstacles. This poses certain safety risks to robot navigation and is difficult to meet the needs.

[0004] Therefore, how to provide a robot path dynamic planning method and system to improve the reliability and generalization ability of robot path planning has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a robot path dynamic planning method and system to improve the reliability and generalization ability of robot path planning.

[0006] In a first aspect, the present invention provides a method for dynamic robot path planning, comprising the following steps:

[0007] Step S1: creating a robot path planning model including a first search strategy, a second search strategy, a third search strategy, and a fourth search strategy based on a ripple diffusion algorithm;

[0008] Step S2: creating a simulation environment, setting a plurality of dynamic obstacles and a plurality of static obstacles in the simulation environment, and setting the motion trajectory and motion speed of each of the dynamic obstacles;

[0009] Step S3: Initializing a plurality of virtual robots in the simulation environment, and training a robot path planning model based on the simulation environment and the virtual robots;

[0010] Step S4, performing performance verification on the trained robot path planning model based on the average path deviation error, the distance to the target not traveled, the total fitness, and the average execution time;

[0011] Step S5: deploy the robot path planning model that has passed the performance verification, and perform path planning using the deployed robot path planning model.

[0012] Furthermore, in step S1, both the first search strategy and the second search strategy introduce an intermediate operator, and the search is guided by the intermediate operator;

[0013] The formula of the first search strategy is:

[0014]

[0015] The formula for the second search strategy is:

[0016]

[0017] The formula of the third search strategy is:

[0018]

[0019] The formula of the fourth search strategy is:

[0020]

[0021] in, represents the t+1th iteration result of the i-th path; r1 represents the dynamic weight, represents a random vector consisting of 0 and 1, Z represents the iteration ratio, Z = t / T, t represents the current number of iterations, T represents the maximum number of iterations; Best represents the optimal path; represents the result of the tth iteration of the i-th path; rand represents a random number between [0,1]; represents the intermediate operator, represents the t-th iteration result of the c-th path, i and c are path numbers, both i and c belong to [1, N], N is the total number of paths, and i ≠ c; F represents the dynamic adaptive parameter, F = 0.5 + 0.25 × cos ((t × π) / T); represents the inertia parameter of the t-th iteration of the i-th path, represents the inertia parameter of the t+1th iteration of the i-th path; r2 represents the oscillation parameter, == represents logical judgment, ~ represents logical inversion; r3 represents the oscillation parameter, r3=4×r2-2.

[0022] Furthermore, the step S3 is specifically as follows:

[0023] Initialize several virtual robots in the simulation environment, set the starting coordinates and the ending coordinates of each virtual robot, and train the robot path planning model based on the simulation environment and the virtual robots. During the training process, continuously monitor whether there is a collision with dynamic obstacles, static obstacles or other robots. If a collision occurs, update the model parameters of the robot path planning model, and accumulate the number of iterations until the training reaches the maximum number of iterations.

[0024] Furthermore, in step S4, the formula for the average path deviation error is:

[0025]

[0026] Among them, APDE represents the average path deviation error; R represents the total number of repeated planning; k represents the number of planning times; PDE k represents the path deviation error of the k-th planning; NR represents the total number of robots; j represents the robot number; Indicates the actual travel distance of the jth robot in the kth planning; ID j represents the ideal shortest distance of the jth robot;

[0027] The formula for the untraveled target distance is:

[0028]

[0029] Among them, AUGD represents the distance to the target not traveled; UGD k represents the untraveled distance of the kth planning; MS represents the maximum number of steps in path planning; l represents the number of steps; G represents the position of the jth robot at step l in the kth planning; j represents the target position of the jth robot;

[0030] The formula for the total fitness is:

[0031]

[0032] Among them, Totalfitness represents the total fitness; f j,k represents the fitness of the jth robot in the kth planning;

[0033] The formula for the average execution time is:

[0034]

[0035] Among them, AET represents the average execution time; k represents the execution time of the kth path planning.

[0036] Furthermore, the step S5 is specifically as follows:

[0037] The robot path planning model that has passed the performance verification is deployed to the physical robot. The physical robot collects road condition data and motion data in real time through sensors, and inputs the road condition data and motion data into the deployed robot path planning model for path planning.

[0038] In a second aspect, the present invention provides a robot path dynamic planning system, comprising the following modules:

[0039] A robot path planning model creation module, configured to create a robot path planning model comprising a first search strategy, a second search strategy, a third search strategy, and a fourth search strategy based on a ripple diffusion algorithm;

[0040] A simulation environment creation module is used to create a simulation environment, set a plurality of dynamic obstacles and a plurality of static obstacles in the simulation environment, and set the movement trajectory and movement speed of each of the dynamic obstacles;

[0041] A robot path planning model training module is used to initialize a plurality of virtual robots in the simulation environment and train the robot path planning model based on the simulation environment and the virtual robots;

[0042] A performance verification module is used to verify the performance of the trained robot path planning model based on the average path deviation error, the distance to the target not traveled, the total fitness and the average execution time;

[0043] The dynamic path planning module is used to deploy the robot path planning model that has passed the performance verification and perform path planning using the deployed robot path planning model.

[0044] Furthermore, in the robot path planning model creation module, both the first search strategy and the second search strategy introduce an intermediate operator to guide the search;

[0045] The formula of the first search strategy is:

[0046]

[0047] The formula for the second search strategy is:

[0048]

[0049] The formula of the third search strategy is:

[0050]

[0051] The formula of the fourth search strategy is:

[0052]

[0053] in, represents the t+1th iteration result of the i-th path; r1 represents the dynamic weight, represents a random vector consisting of 0 and 1, Z represents the iteration ratio, Z = t / T, t represents the current number of iterations, T represents the maximum number of iterations; Best represents the optimal path; represents the result of the tth iteration of the i-th path; rand represents a random number between [0,1]; represents the intermediate operator, represents the t-th iteration result of the c-th path, i and c are path numbers, both i and c belong to [1, N], N is the total number of paths, and i ≠ c; F represents the dynamic adaptive parameter, F = 0.5 + 0.25 × cos ((t × π) / T); represents the inertia parameter of the t-th iteration of the i-th path, represents the inertia parameter of the t+1th iteration of the i-th path; r2 represents the oscillation parameter, == represents logical judgment, ~ represents logical inversion; r3 represents the oscillation parameter, r3=4×r2-2.

[0054] Furthermore, the robot path planning model training module is specifically used to:

[0055] Initialize several virtual robots in the simulation environment, set the starting coordinates and the ending coordinates of each virtual robot, and train the robot path planning model based on the simulation environment and the virtual robots. During the training process, continuously monitor whether there is a collision with dynamic obstacles, static obstacles or other robots. If a collision occurs, update the model parameters of the robot path planning model, and accumulate the number of iterations until the training reaches the maximum number of iterations.

[0056] Furthermore, in the performance verification module, the formula for the average path deviation error is:

[0057]

[0058] Among them, APDE represents the average path deviation error; R represents the total number of repeated planning; k represents the number of planning times; PDE k represents the path deviation error of the k-th planning; NR represents the total number of robots; j represents the robot number; Indicates the actual travel distance of the jth robot in the kth planning; ID j represents the ideal shortest distance of the jth robot;

[0059] The formula for the untraveled target distance is:

[0060]

[0061] Among them, AUGD represents the distance to the target not traveled; UGD k represents the untraveled distance of the kth planning; MS represents the maximum number of steps in path planning; l represents the number of steps; G represents the position of the jth robot at step l in the kth planning; j represents the target position of the jth robot;

[0062] The formula for the total fitness is:

[0063]

[0064] Among them, Totalfitness represents the total fitness; f j,k represents the fitness of the jth robot in the kth planning;

[0065] The formula for the average execution time is:

[0066]

[0067] Among them, AET represents the average execution time; k represents the execution time of the kth path planning.

[0068] Furthermore, the path dynamic planning module is specifically used to:

[0069] The robot path planning model that has passed the performance verification is deployed to the physical robot. The physical robot collects road condition data and motion data in real time through sensors, and inputs the road condition data and motion data into the deployed robot path planning model for path planning.

[0070] The advantages of the present invention are:

[0071] 1. Create a robot path planning model that includes the first search strategy, the second search strategy, the third search strategy, and the fourth search strategy using the ripple diffusion algorithm; then create a simulation environment, set several dynamic obstacles and several static obstacles in the simulation environment, set the motion trajectory and motion speed of each dynamic obstacle, initialize several virtual robots in the simulation environment, and train the robot path planning model based on the simulation environment and the virtual robots; then perform performance verification on the trained robot path planning model based on the average path deviation error, untraveled target distance, total fitness, and average execution time; finally, deploy the robot path planning model that passes the performance verification, and perform path planning using the deployed robot path planning model; That is, the ripple diffusion algorithm (RSA algorithm) is improved, and four different search strategies are set to enhance the diversity of planned paths. Based on the four search strategies, several paths are dynamically planned for the robot each time, and an intermediate operator (which integrates the position information of other operators (paths) in the population (path set)) is introduced to guide the path search. The optimal path output is then screened out based on the planned path. In the training process of the robot path planning model, the influence of dynamic obstacles, static obstacles and other robots is comprehensively considered. The four search strategies of the robot path planning model are based on the dynamic adaptive parameter F to dynamically adjust the proportion, ensuring that the robot path planning model converges in the later stage of iteration to jump out of the local optimal value, ultimately greatly improving the reliability and generalization ability of the robot path planning.

[0072] 2. By setting up a robot path planning model that integrates four search strategies (the first to fourth search strategies), the global search capability and local optimization capability are enhanced through the complementarity of different strategies, avoiding the limitations of a single strategy.

[0073] 3. By introducing intermediate operators to dynamically guide the search direction, the path diversity is improved through information interaction between paths, and the risk of local optimality is reduced.

[0074] 4. By setting the dynamic weight r1 and associating it with the iteration ratio Z, the weight is automatically adjusted as the number of iterations increases. The initial focus is on exploration, and the later focus is on development, which effectively optimizes the convergence speed.

[0075] 5. By setting the dynamic adaptive parameter F, the search strategy selection threshold is dynamically adjusted through the cosine function to effectively balance the exploration and development needs at different stages.

[0076] 6. By setting the inertia parameter to decrease with iteration, the global search capability is enhanced in the early stage, and gradually converges in the later stage, effectively improving the path smoothness.

[0077] 7. By setting up both dynamic and static obstacles in the simulation environment, the model can adapt to dynamically changing scenarios through training, effectively enhancing its robustness to complex environments. During the model training process, collisions (dynamic obstacles, static obstacles, and other robots) are monitored in real time, and the path is optimized by updating model parameters to ensure path safety and feasibility.

[0078] 8. By introducing oscillation parameters r2, r3 and random vectors in the third and fourth search strategies, the model's ability to respond to changes in the position of dynamic obstacles is enhanced, avoiding path planning lags.

[0079] 9. The robot path planning model is evaluated in multiple dimensions using average path deviation error, distance to untraveled targets, total fitness, and average execution time. The path length, accuracy, obstacle avoidance capability, and execution efficiency are comprehensively verified to ensure the comprehensive and reliable performance of the robot path planning model.

[0080] 10. Through collision monitoring and parameter update mechanisms, the risk of path failure caused by sudden environmental changes in actual deployment is reduced, and system stability is improved.

[0081] 11. Through multi-strategy fusion, dynamic parameter adaptation, complex environment simulation and comprehensive performance verification, efficient and robust dynamic path planning is achieved, which combines algorithm innovation and practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0083] Figure 1 It is a flow chart of a robot path dynamic planning method of the present invention.

[0084] Figure 2 It is a structural schematic diagram of a robot path dynamic planning system of the present invention.

[0085] Figure 3 It is a schematic diagram of the dynamic planning of the robot path of the present invention.

[0086] Figure 4 It is a schematic diagram of the iteration of the dynamic adaptive parameter F of the present invention. DETAILED DESCRIPTION

[0087] The technical solution in the embodiments of the present application has the following overall idea: improving the ripple diffusion algorithm, setting four different search strategies to enhance the diversity of planned paths, dynamically planning several paths for the robot each time based on the four search strategies, introducing intermediate operators to guide the path search, and then screening out the optimal path output based on the planned path. In addition, the influence of dynamic obstacles, static obstacles and other robots is comprehensively considered during the training process of the robot path planning model. The four search strategies of the robot path planning model dynamically adjust the proportion based on the dynamic adaptive parameter F to ensure convergence in the late iteration to jump out of the local optimal value, thereby improving the reliability and generalization ability of the robot path planning.

[0088] Please refer to Figures 1 to 4 As shown, a preferred embodiment of a robot path dynamic planning method of the present invention includes the following steps:

[0089] Step S1, creating a robot path planning model including a first search strategy, a second search strategy, a third search strategy and a fourth search strategy based on a ripple diffusion algorithm (RSA algorithm);

[0090] By setting up a robot path planning model that integrates four search strategies (the first to fourth search strategies), the global search capability and local optimization capability are enhanced through the complementarity of different strategies, avoiding the limitations of a single strategy.

[0091] Step S2: creating a simulation environment, setting a plurality of dynamic obstacles and a plurality of static obstacles in the simulation environment, and setting the motion trajectory and motion speed of each of the dynamic obstacles;

[0092] By setting up both dynamic and static obstacles in the simulation environment, the model can adapt to dynamically changing scenarios through training, effectively enhancing its robustness to complex environments. During the model training process, collisions (dynamic obstacles, static obstacles, and other robots) are monitored in real time, and the path is optimized by updating model parameters to ensure path safety and feasibility.

[0093] Step S3: Initializing a plurality of virtual robots in the simulation environment, and training a robot path planning model based on the simulation environment and the virtual robots;

[0094] For example, the simulation environment is set as a training scene of 300×300 cm, in which twelve virtual robots (marked as R1 to R12 and distinguished by different colors to enhance visual recognition), twenty-five static obstacles, and five dynamic obstacles (marked as D1 to D5) are deployed; the dynamic obstacles move according to preset trajectories, and their speeds are set to 0.5 cm / step (D1), 0.45 cm / step (D2), 0.6 cm / step (D3), 0.7 cm / step (D4), and 0.9 cm / step (D5).

[0095] Step S4, performing performance verification on the trained robot path planning model based on the average path deviation error, the distance to the target not traveled, the total fitness, and the average execution time;

[0096] In practical applications, robots often need to perform tasks in complex and dynamically changing environments, so they are required to have the ability to respond quickly and update in real time. Therefore, efficient search algorithms and optimization strategies are introduced to ensure that robots can generate safe and feasible paths in real time in dynamic environments. The average path deviation error and average execution time indicators in the simulation evaluation index system are used to quantitatively evaluate the path accuracy and computational efficiency. By introducing the untraveled target distance indicator, the impact of the collision avoidance strategy on the task completion is evaluated to ensure that the robot can efficiently approach the target position while avoiding collisions. The comprehensive indicator of total fitness is used to evaluate the performance of the algorithm in optimization to ensure a balance in path length, safety, energy consumption, etc., so as to adapt to the requirements of tasks of different scales and complexities.

[0097] Step S5: Deploy the robot path planning model that has passed the performance verification, and perform path planning using the deployed robot path planning model. The planned path includes the robot's next displacement speed, displacement angle, and displacement distance.

[0098] Through multi-strategy fusion, dynamic parameter adaptation, complex environment simulation and comprehensive performance verification, efficient and robust dynamic path planning is achieved, which combines algorithm innovation and practical application value.

[0099] In step S1, both the first search strategy and the second search strategy introduce an intermediate operator to guide the search, and the intermediate operator integrates the position information (solution vector) of other operators (paths) in the population (path set);

[0100] By introducing intermediate operators to dynamically guide the search direction, the path diversity is improved through information interaction between paths, and the risk of local optimality is reduced.

[0101] The formula of the first search strategy is:

[0102]

[0103] The formula for the second search strategy is:

[0104]

[0105] The formula of the third search strategy is:

[0106]

[0107] The formula of the fourth search strategy is:

[0108]

[0109] in, represents the t+1th iteration result of the i-th path; r1 represents the dynamic weight, represents a random vector consisting of 0 and 1, Z represents the iteration ratio, Z = t / T, t represents the current number of iterations, T represents the maximum number of iterations; Best represents the optimal path; represents the result of the tth iteration of the i-th path; rand represents a random number between [0,1]; represents the intermediate operator, represents the t-th iteration result of the c-th path, i and c are path numbers, both i and c belong to [1, N], N is the total number of paths, and i ≠ c; F represents the dynamic adaptive parameter, F = 0.5 + 0.25 × cos ((t × π) / T); represents the inertia parameter of the t-th iteration of the i-th path, represents the inertia parameter of the t+1th iteration of the i-th path; r2 represents the oscillation parameter, == represents logical judgment, ~ represents logical inversion; r3 represents the oscillation parameter, r3=4×r2-2.

[0110] By setting the dynamic weight r1 to be associated with the iteration ratio Z, the weight is automatically adjusted as the number of iterations increases. The initial stage focuses on exploration, and the later stage focuses on development, which effectively optimizes the convergence speed.

[0111] By setting the dynamic adaptive parameter F, the search strategy selection threshold is dynamically adjusted through the cosine function to effectively balance the exploration and development needs at different stages.

[0112] By setting the inertia parameter to decrease with iteration, the global search capability is enhanced in the early stage, and gradually converges in the later stage, effectively improving the path smoothness.

[0113] By introducing oscillation parameters r2, r3 and random vectors into the third and fourth search strategies, the model's ability to respond to changes in the position of dynamic obstacles is enhanced, avoiding path planning lags.

[0114] The step S3 is specifically as follows:

[0115] Initialize several virtual robots in the simulation environment, set the starting coordinates and the ending coordinates of each virtual robot, and train the robot path planning model based on the simulation environment and the virtual robots. During the training process, continuously monitor whether there is a collision with dynamic obstacles, static obstacles or other robots. If a collision occurs, update the model parameters of the robot path planning model, and accumulate the number of iterations until the training reaches the maximum number of iterations.

[0116] Through collision monitoring and parameter update mechanisms, the risk of path failure caused by sudden environmental changes in actual deployment is reduced, and system stability is improved.

[0117] In step S4, the formula for the average path deviation error is:

[0118]

[0119] Among them, APDE represents the average path deviation error; R represents the total number of repeated planning; k represents the number of planning times; PDE k represents the path deviation error of the k-th planning; NR represents the total number of robots; j represents the robot number; Indicates the actual travel distance of the jth robot in the kth planning; ID j represents the ideal shortest distance of the jth robot;

[0120] The formula for the untraveled target distance is:

[0121]

[0122] Among them, AUGD represents the distance to the target not traveled; UGD k represents the untraveled distance of the kth planning; MS represents the maximum number of steps in path planning; l represents the number of steps; G represents the position of the jth robot at step l in the kth planning; j represents the target position of the jth robot;

[0123] The formula for the total fitness is:

[0124]

[0125] Among them, Totalfitness represents the total fitness; f j,krepresents the fitness of the jth robot in the kth planning;

[0126] The formula for the average execution time is:

[0127]

[0128] Among them, AET represents the average execution time; k represents the execution time of the kth path planning.

[0129] The robot path planning model is evaluated in multiple dimensions using average path deviation error, distance to untraveled targets, total fitness, and average execution time. The path length, accuracy, obstacle avoidance capability, and execution efficiency are comprehensively verified to ensure the comprehensive and reliable performance of the robot path planning model.

[0130] The step S5 is specifically as follows:

[0131] The robot path planning model that has passed the performance verification is deployed to the physical robot. The physical robot collects road condition data and motion data in real time through sensors, and inputs the road condition data and motion data into the deployed robot path planning model for path planning.

[0132] A preferred embodiment of a robot path dynamic planning system of the present invention includes the following modules:

[0133] A robot path planning model creation module is used to create a robot path planning model including a first search strategy, a second search strategy, a third search strategy and a fourth search strategy based on a ripple diffusion algorithm (RSA algorithm);

[0134] By setting up a robot path planning model that integrates four search strategies (the first to fourth search strategies), the global search capability and local optimization capability are enhanced through the complementarity of different strategies, avoiding the limitations of a single strategy.

[0135] A simulation environment creation module is used to create a simulation environment, set a plurality of dynamic obstacles and a plurality of static obstacles in the simulation environment, and set the movement trajectory and movement speed of each of the dynamic obstacles;

[0136] By setting up both dynamic and static obstacles in the simulation environment, the model can adapt to dynamically changing scenarios through training, effectively enhancing its robustness to complex environments. During the model training process, collisions (dynamic obstacles, static obstacles, and other robots) are monitored in real time, and the path is optimized by updating model parameters to ensure path safety and feasibility.

[0137] A robot path planning model training module is used to initialize a plurality of virtual robots in the simulation environment and train the robot path planning model based on the simulation environment and the virtual robots;

[0138] For example, the simulation environment is set as a training scene of 300×300 cm, in which twelve virtual robots (marked as R1 to R12 and distinguished by different colors to enhance visual recognition), twenty-five static obstacles, and five dynamic obstacles (marked as D1 to D5) are deployed; the dynamic obstacles move according to preset trajectories, and their speeds are set to 0.5 cm / step (D1), 0.45 cm / step (D2), 0.6 cm / step (D3), 0.7 cm / step (D4), and 0.9 cm / step (D5).

[0139] A performance verification module is used to verify the performance of the trained robot path planning model based on the average path deviation error, the distance to the target not traveled, the total fitness and the average execution time;

[0140] In practical applications, robots often need to perform tasks in complex and dynamically changing environments, so they are required to have the ability to respond quickly and update in real time. Therefore, efficient search algorithms and optimization strategies are introduced to ensure that robots can generate safe and feasible paths in real time in dynamic environments. The average path deviation error and average execution time indicators in the simulation evaluation index system are used to quantitatively evaluate the path accuracy and computational efficiency. By introducing the untraveled target distance indicator, the impact of the collision avoidance strategy on the task completion is evaluated to ensure that the robot can efficiently approach the target position while avoiding collisions. The comprehensive indicator of total fitness is used to evaluate the performance of the algorithm in optimization to ensure a balance in path length, safety, energy consumption, etc., so as to adapt to the requirements of tasks of different scales and complexities.

[0141] The dynamic path planning module is used to deploy the robot path planning model that has passed performance verification and perform path planning using the deployed robot path planning model. The planned path includes the robot's next displacement speed, displacement angle, and displacement distance.

[0142] Through multi-strategy fusion, dynamic parameter adaptation, complex environment simulation and comprehensive performance verification, efficient and robust dynamic path planning is achieved, which combines algorithm innovation and practical application value.

[0143] In the robot path planning model creation module, both the first search strategy and the second search strategy introduce an intermediate operator to guide the search, and the intermediate operator integrates the position information (solution vector) of other operators (paths) in the population (path set);

[0144] By introducing intermediate operators to dynamically guide the search direction, the path diversity is improved through information interaction between paths, and the risk of local optimality is reduced.

[0145] The formula of the first search strategy is:

[0146]

[0147] The formula for the second search strategy is:

[0148]

[0149] The formula of the third search strategy is:

[0150]

[0151] The formula of the fourth search strategy is:

[0152]

[0153] in, represents the t+1th iteration result of the i-th path; r1 represents the dynamic weight, represents a random vector consisting of 0 and 1, Z represents the iteration ratio, Z = t / T, t represents the current number of iterations, T represents the maximum number of iterations; Best represents the optimal path; represents the result of the tth iteration of the i-th path; rand represents a random number between [0,1]; represents the intermediate operator, represents the t-th iteration result of the c-th path, i and c are path numbers, both i and c belong to [1, N], N is the total number of paths, and i ≠ c; F represents the dynamic adaptive parameter, F = 0.5 + 0.25 × cos ((t × π) / T); represents the inertia parameter of the t-th iteration of the i-th path, represents the inertia parameter of the t+1th iteration of the i-th path; r2 represents the oscillation parameter, == represents logical judgment, ~ represents logical inversion; r3 represents the oscillation parameter, r3=4×r2-2.

[0154] By setting the dynamic weight r1 to be associated with the iteration ratio Z, the weight is automatically adjusted as the number of iterations increases. The initial stage focuses on exploration, and the later stage focuses on development, which effectively optimizes the convergence speed.

[0155] By setting the dynamic adaptive parameter F, the search strategy selection threshold is dynamically adjusted through the cosine function to effectively balance the exploration and development needs at different stages.

[0156] By setting the inertia parameter to decrease with iteration, the global search capability is enhanced in the early stage, and gradually converges in the later stage, effectively improving the path smoothness.

[0157] By introducing oscillation parameters r2, r3 and random vectors into the third and fourth search strategies, the model's ability to respond to changes in the position of dynamic obstacles is enhanced, avoiding path planning lags.

[0158] The robot path planning model training module is specifically used for:

[0159] Initialize several virtual robots in the simulation environment, set the starting coordinates and the ending coordinates of each virtual robot, and train the robot path planning model based on the simulation environment and the virtual robots. During the training process, continuously monitor whether there is a collision with dynamic obstacles, static obstacles or other robots. If a collision occurs, update the model parameters of the robot path planning model, and accumulate the number of iterations until the training reaches the maximum number of iterations.

[0160] Through collision monitoring and parameter update mechanisms, the risk of path failure caused by sudden environmental changes in actual deployment is reduced, and system stability is improved.

[0161] In the performance verification module, the formula for the average path deviation error is:

[0162]

[0163] Among them, APDE represents the average path deviation error; R represents the total number of repeated planning; k represents the number of planning times; PDE k represents the path deviation error of the k-th planning; NR represents the total number of robots; j represents the robot number; Indicates the actual travel distance of the jth robot in the kth planning; ID j represents the ideal shortest distance of the jth robot;

[0164] The formula for the untraveled target distance is:

[0165]

[0166] Among them, AUGD represents the distance to the target not traveled; UGD k represents the untraveled distance of the kth planning; MS represents the maximum number of steps in path planning; l represents the number of steps; G represents the position of the jth robot at step l in the kth planning; j represents the target position of the jth robot;

[0167] The formula for the total fitness is:

[0168]

[0169] Among them, Totalfitness represents the total fitness; f j,k represents the fitness of the jth robot in the kth planning;

[0170] The formula for the average execution time is:

[0171]

[0172] Among them, AET represents the average execution time; k represents the execution time of the kth path planning.

[0173] The robot path planning model is evaluated in multiple dimensions using average path deviation error, distance to untraveled targets, total fitness, and average execution time. The path length, accuracy, obstacle avoidance capability, and execution efficiency are comprehensively verified to ensure the comprehensive and reliable performance of the robot path planning model.

[0174] The path dynamic planning module is specifically used for:

[0175] The robot path planning model that has passed the performance verification is deployed to the physical robot. The physical robot collects road condition data and motion data in real time through sensors, and inputs the road condition data and motion data into the deployed robot path planning model for path planning.

[0176] In summary, the advantages of the present invention are:

[0177] 1. Create a robot path planning model that includes the first search strategy, the second search strategy, the third search strategy, and the fourth search strategy using the ripple diffusion algorithm; then create a simulation environment, set several dynamic obstacles and several static obstacles in the simulation environment, set the motion trajectory and motion speed of each dynamic obstacle, initialize several virtual robots in the simulation environment, and train the robot path planning model based on the simulation environment and the virtual robots; then perform performance verification on the trained robot path planning model based on the average path deviation error, untraveled target distance, total fitness, and average execution time; finally, deploy the robot path planning model that passes the performance verification, and perform path planning using the deployed robot path planning model; That is, the ripple diffusion algorithm (RSA algorithm) is improved, and four different search strategies are set to enhance the diversity of planned paths. Based on the four search strategies, several paths are dynamically planned for the robot each time, and an intermediate operator (which integrates the position information of other operators (paths) in the population (path set)) is introduced to guide the path search. The optimal path output is then screened out based on the planned path. In the training process of the robot path planning model, the influence of dynamic obstacles, static obstacles and other robots is comprehensively considered. The four search strategies of the robot path planning model are based on the dynamic adaptive parameter F to dynamically adjust the proportion, ensuring that the robot path planning model converges in the later stage of iteration to jump out of the local optimal value, ultimately greatly improving the reliability and generalization ability of the robot path planning.

[0178] 2. By setting up a robot path planning model that integrates four search strategies (the first to fourth search strategies), the global search capability and local optimization capability are enhanced through the complementarity of different strategies, avoiding the limitations of a single strategy.

[0179] 3. By introducing intermediate operators to dynamically guide the search direction, the path diversity is improved through information interaction between paths, and the risk of local optimality is reduced.

[0180] 4. By setting the dynamic weight r1 and associating it with the iteration ratio Z, the weight is automatically adjusted as the number of iterations increases. The initial focus is on exploration, and the later focus is on development, which effectively optimizes the convergence speed.

[0181] 5. By setting the dynamic adaptive parameter F, the search strategy selection threshold is dynamically adjusted through the cosine function to effectively balance the exploration and development needs at different stages.

[0182] 6. By setting the inertia parameter to decrease with iteration, the global search capability is enhanced in the early stage, and gradually converges in the later stage, effectively improving the path smoothness.

[0183] 7. By setting up both dynamic and static obstacles in the simulation environment, the model can adapt to dynamically changing scenarios through training, effectively enhancing its robustness to complex environments. During the model training process, collisions (dynamic obstacles, static obstacles, and other robots) are monitored in real time, and the path is optimized by updating model parameters to ensure path safety and feasibility.

[0184] 8. By introducing oscillation parameters r2, r3 and random vectors in the third and fourth search strategies, the model's ability to respond to changes in the position of dynamic obstacles is enhanced, avoiding path planning lags.

[0185] 9. The robot path planning model is evaluated in multiple dimensions using average path deviation error, distance to untraveled targets, total fitness, and average execution time. The path length, accuracy, obstacle avoidance capability, and execution efficiency are comprehensively verified to ensure the comprehensive and reliable performance of the robot path planning model.

[0186] 10. Through collision monitoring and parameter update mechanisms, the risk of path failure caused by sudden environmental changes in actual deployment is reduced, and system stability is improved.

[0187] 11. Through multi-strategy fusion, dynamic parameter adaptation, complex environment simulation and comprehensive performance verification, efficient and robust dynamic path planning is achieved, which combines algorithm innovation and practical application value.

[0188] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and are not intended to limit the scope of the present invention. Equivalent modifications and changes made by those skilled in the art in accordance with the spirit of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A robot path dynamic planning method, characterized by: The steps include: Step S1: creating a robot path planning model including a first search strategy, a second search strategy, a third search strategy, and a fourth search strategy based on a ripple diffusion algorithm; Step S2: creating a simulation environment, setting a plurality of dynamic obstacles and a plurality of static obstacles in the simulation environment, and setting the motion trajectory and motion speed of each of the dynamic obstacles; Step S3: Initializing a plurality of virtual robots in the simulation environment, and training a robot path planning model based on the simulation environment and the virtual robots; Step S4, performing performance verification on the trained robot path planning model based on the average path deviation error, the distance to the target not traveled, the total fitness, and the average execution time; Step S5: deploy the robot path planning model that has passed the performance verification, and perform path planning using the deployed robot path planning model.

2. A robot path dynamic planning method according to claim 1, characterized in that: In step S1, both the first search strategy and the second search strategy introduce an intermediate operator to guide the search; The formula of the first search strategy is: The formula for the second search strategy is: The formula of the third search strategy is: The formula of the fourth search strategy is: in, represents the t+1th iteration result of the i-th path; r1 represents the dynamic weight, represents a random vector consisting of 0 and 1, Z represents the iteration ratio, Z = t / T, t represents the current number of iterations, T represents the maximum number of iterations; Best represents the optimal path; represents the result of the tth iteration of the i-th path; rand represents a random number between [0,1]; represents the intermediate operator, represents the t-th iteration result of the c-th path, i and c are path numbers, both i and c belong to [1, N], N is the total number of paths, and i ≠ c; F represents the dynamic adaptive parameter, F = 0.5 + 0.25 × cos ((t × π) / T); represents the inertia parameter of the t-th iteration of the i-th path, represents the inertia parameter of the t+1th iteration of the i-th path; r2 represents the oscillation parameter, == represents logical judgment, ~ represents logical inversion; r3 represents the oscillation parameter, r3=4×r2-2.

3. A robot path dynamic planning method according to claim 1, characterized in that: The step S3 is specifically as follows: Initialize several virtual robots in the simulation environment, set the starting coordinates and the ending coordinates of each virtual robot, and train the robot path planning model based on the simulation environment and the virtual robots. During the training process, continuously monitor whether there is a collision with dynamic obstacles, static obstacles or other robots. If a collision occurs, update the model parameters of the robot path planning model, and accumulate the number of iterations until the training reaches the maximum number of iterations.

4. A robot path dynamic planning method according to claim 1, characterized in that: In step S4, the formula for the average path deviation error is: Among them, APDE represents the average path deviation error; R represents the total number of repeated planning; k represents the number of planning times; PDE k represents the path deviation error of the k-th planning; NR represents the total number of robots; j represents the robot number; Indicates the actual travel distance of the jth robot in the kth planning; ID j represents the ideal shortest distance of the jth robot; The formula for the untraveled target distance is: Among them, AUGD represents the distance to the target not traveled; UGD k represents the untraveled distance of the kth planning; MS represents the maximum number of steps in path planning; l represents the number of steps; G represents the position of the jth robot at step l in the kth planning; j represents the target position of the jth robot; The formula for the total fitness is: Among them, Totalfitness represents the total fitness; f j,k represents the fitness of the jth robot in the kth planning; The formula for the average execution time is: Among them, AET represents the average execution time; k represents the execution time of the kth path planning.

5. A robot path dynamic planning method according to claim 1, characterized in that: The step S5 is specifically as follows: The robot path planning model that has passed the performance verification is deployed to the physical robot. The physical robot collects road condition data and motion data in real time through sensors, and inputs the road condition data and motion data into the deployed robot path planning model for path planning.

6. A robot path dynamic planning system, characterized by: Includes the following modules: A robot path planning model creation module, configured to create a robot path planning model comprising a first search strategy, a second search strategy, a third search strategy, and a fourth search strategy based on a ripple diffusion algorithm; A simulation environment creation module is used to create a simulation environment, set a plurality of dynamic obstacles and a plurality of static obstacles in the simulation environment, and set the movement trajectory and movement speed of each of the dynamic obstacles; A robot path planning model training module is used to initialize a plurality of virtual robots in the simulation environment and train the robot path planning model based on the simulation environment and the virtual robots; A performance verification module is used to verify the performance of the trained robot path planning model based on the average path deviation error, the distance to the target not traveled, the total fitness and the average execution time; The dynamic path planning module is used to deploy the robot path planning model that has passed the performance verification and perform path planning using the deployed robot path planning model.

7. A robot path dynamic planning system according to claim 6, characterized in that: In the robot path planning model creation module, both the first search strategy and the second search strategy introduce an intermediate operator to guide the search; The formula of the first search strategy is: The formula for the second search strategy is: The formula of the third search strategy is: The formula of the fourth search strategy is: in, represents the t+1th iteration result of the i-th path; r1 represents the dynamic weight, represents a random vector consisting of 0 and 1, Z represents the iteration ratio, Z = t / T, t represents the current number of iterations, T represents the maximum number of iterations; Best represents the optimal path; represents the result of the tth iteration of the i-th path; rand represents a random number between [0,1]; represents the intermediate operator, represents the t-th iteration result of the c-th path, i and c are path numbers, both i and c belong to [1, N], N is the total number of paths, and i ≠ c; F represents the dynamic adaptive parameter, F = 0.5 + 0.25 × cos ((t × π) / T); represents the inertia parameter of the t-th iteration of the i-th path, represents the inertia parameter of the t+1th iteration of the i-th path; r2 represents the oscillation parameter, == represents logical judgment, ~ represents logical inversion; r3 represents the oscillation parameter, r3=4×r2-2.

8. A robot path dynamic planning system according to claim 6, characterized in that: The robot path planning model training module is specifically used for: Initialize several virtual robots in the simulation environment, set the starting coordinates and the ending coordinates of each virtual robot, and train the robot path planning model based on the simulation environment and the virtual robots. During the training process, continuously monitor whether there is a collision with dynamic obstacles, static obstacles or other robots. If a collision occurs, update the model parameters of the robot path planning model, and accumulate the number of iterations until the training reaches the maximum number of iterations.

9. A robot path dynamic planning system according to claim 6, characterized in that: In the performance verification module, the formula for the average path deviation error is: Among them, APDE represents the average path deviation error; R represents the total number of repeated planning; k represents the number of planning times; PDE k represents the path deviation error of the k-th planning; NR represents the total number of robots; j represents the robot number; Indicates the actual travel distance of the jth robot in the kth planning; ID j represents the ideal shortest distance of the jth robot; The formula for the untraveled target distance is: Among them, AUGD represents the distance to the target not traveled; UGD k represents the untraveled distance of the kth planning; MS represents the maximum number of steps in path planning; l represents the number of steps; G represents the position of the jth robot at step l in the kth planning; j represents the target position of the jth robot; The formula for the total fitness is: Among them, Totalfitness represents the total fitness; f j,k represents the fitness of the jth robot in the kth planning; The formula for the average execution time is: Among them, AET represents the average execution time; k represents the execution time of the kth path planning.

10. A robot path dynamic planning system according to claim 6, characterized in that: The path dynamic planning module is specifically used for: The robot path planning model that has passed the performance verification is deployed to the physical robot. The physical robot collects road condition data and motion data in real time through sensors, and inputs the road condition data and motion data into the deployed robot path planning model for path planning.

Images