Robot path planning method, electronic equipment and medium

Abstract

The invention relates to a robot path planning method, electronic equipment and a medium, and belongs to the technical field of intelligent robos.The method comprises the steps that an obtained static cost map and an obtained dynamic cost map are combined to obtain a global cost map; performing global path planning on the global cost map to obtain a global optimal path; performing local path planning based on the global optimal path to obtain a local optimal track, and controlling the robot to execute the local optimal track until a global target point is reached; and determining the dynamic cost of the local optimal trajectory, updating the dynamic cost map, further updating the global cost map, and performing path planning again. According to the method, the dynamic cost of executing the local optimal track by the robot is updated into the dynamic cost map, so that the robot can actively avoid a historical difficult passing area in the path planning process, and the navigation stability and long-term operation efficiency of the robot in a complex and narrow environment are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent robots, and in particular to a robot path planning method, an electronic device and a medium. BACKGROUND

[0002] A mobile robot is a comprehensive system integrating environment perception, dynamic decision and planning, behavior control and execution, and is widely used in smart logistics, automatic sorting and many other scenarios. Among them, path planning is one of the core technologies of a mobile robot, and its quality directly affects the safety and efficiency of the mobile robot.

[0003] Traditional path planning generally uses (A-star) algorithm and dynamic window algorithm (Dynamic Window Approach, DWA) to plan cooperatively, The A-star algorithm is responsible for searching a globally optimal path from the starting point to the target point on the known map to avoid static obstacles and ensure path rationality, and the DWA algorithm is responsible for dynamically sampling feasible speed instructions according to real-time sensor data during the robot's travel to select the optimal trajectory to avoid local dynamic obstacles.

[0004] However, the above scheme stays at the level of "single planning-execution", and when the robot repeatedly travels in the same difficult area (such as a narrow passage or long congestion), the system lacks "memory" and "learning" ability and cannot benefit from historical travel experience, which may result in the robot continuously selecting a path that is statically optimal but difficult to travel in practice, resulting in low work efficiency. SUMMARY

[0005] Therefore, it is necessary to provide a robot path planning method, an electronic device and a medium to solve the technical problem of low work efficiency in the prior art.

[0006] To solve the above problems, in a first aspect, the present application provides a robot path planning method, comprising: S101, converting an acquired static grid map into a static cost map, initializing a dynamic cost map, merging the static cost map and the dynamic cost map, and obtaining a global cost map; S102, performing global path planning on the global cost map based on a preset global path planning algorithm to obtain a globally optimal path from a current position point to a global target point; S103, performing local path planning based on a preset local path planning algorithm and the globally optimal path to obtain a locally optimal trajectory and corresponding trajectory planning feature parameters, and controlling the robot to execute the locally optimal trajectory until the global target point is reached; S104, determine a dynamic cost of the local optimal trajectory based on the trajectory planning characteristic parameter, update the dynamic cost map based on the dynamic cost of the local optimal trajectory, and further update the global cost map, and return to step S102.

[0007] In a possible implementation, the global path planning algorithm adopts algorithm, the local path planning algorithm adopts a TEB algorithm, the local path planning algorithm performs local path planning through a multi-objective weighted optimization function, and the multi-objective weighted optimization function is: ; In the formula, denotes a trajectory to be optimized, denotes a cost of the trajectory to be optimized, denotes a time cost, denotes an obstacle cost, denotes a speed constraint cost, denotes an acceleration constraint cost, denotes a time weight coefficient, denotes an obstacle weight coefficient, denotes a speed constraint weight coefficient, denotes an acceleration constraint weight coefficient. In a possible implementation, the time weight coefficient, the obstacle weight coefficient, the speed constraint weight coefficient, and the acceleration constraint weight coefficient are calibrated through the following steps: In a preset first test scenario, the time weight coefficient and the acceleration constraint weight coefficient are fixed as corresponding initial weight coefficients, the obstacle weight coefficient and the speed constraint weight coefficient are adjusted to perform path planning and navigation, so that, in a case where a task completion rate parameter meets a preset requirement, a minimum distance parameter between a robot and an obstacle is maximized, and a speed tracking error parameter is minimized, calibration values of the obstacle weight coefficient and the speed constraint weight coefficient are obtained; In a preset first test scenario, the obstacle weight coefficient and the speed constraint weight coefficient are fixed as corresponding calibration values, the time weight coefficient is fixed as a corresponding initial weight coefficient, and the acceleration constraint weight coefficient is adjusted to perform path planning and navigation, so that, in a case where a task completion rate parameter, a minimum distance parameter between a robot and an obstacle, a speed tracking error parameter, and a response speed parameter all meet corresponding preset requirements, a path smoothness and mechanical jitter degree parameter is minimized, and a calibration value of the acceleration constraint weight coefficient is obtained. In a preset first test scene, the obstacle weight coefficient, the speed constraint weight coefficient and the acceleration constraint weight coefficient are fixed as corresponding calibration values, the time weight coefficient is adjusted for path planning and navigation, so that the task completion rate parameter, the minimum distance between the robot and the obstacle parameter, the speed tracking error parameter, the response speed parameter, the path smoothness and the mechanical jitter degree parameter all meet the corresponding preset requirements, and the total travel time parameter is minimized to obtain the calibration value of the time weight coefficient.

[0008] In a possible implementation, the determining, based on the trajectory planning feature parameters, of the dynamic cost of the local optimal trajectory, and the updating, based on the dynamic cost of the local optimal trajectory, of the dynamic cost map, comprises: calculating a difficulty score based on the trajectory planning feature parameters; calculating a dynamic cost of the local optimal trajectory based on the difficulty score, and updating the dynamic cost map based on the dynamic cost of the local optimal trajectory; wherein the trajectory planning feature parameters comprise a cost of the local optimal trajectory, a minimum distance between the local optimal trajectory and an obstacle, a number of optimization iterations of the local optimal trajectory, and a variance of all linear velocities on the local optimal trajectory, and a calculation formula of the difficulty score is: ; ; ; ; ; In the formula, difficulty score represents the difficulty score, cost represents the cost of the local optimal trajectory, iteration represents the number of optimization iterations of the local optimal trajectory, distance represents the minimum distance between the local optimal trajectory and the obstacle, variance represents the variance of all linear velocities on the local optimal trajectory, normalized iteration represents the normalized number of optimization iterations of the local optimal trajectory, normalized distance represents the normalized minimum distance between the local optimal trajectory and the obstacle, normalized cost represents the normalized cost of the local optimal trajectory, normalized variance represents the normalized variance of all linear velocities on the local optimal trajectory, trajectory cost weight coefficient represents the trajectory cost weight coefficient, trajectory iteration number weight coefficient represents the trajectory iteration number weight coefficient, trajectory obstacle distance weight coefficient represents the trajectory obstacle distance weight coefficient, trajectory linear velocity variance weight coefficient represents the trajectory linear velocity variance weight coefficient, ​represents a trajectory cost reference value, represents a maximum iteration number, represents a distance of interest, represents a minimum safety distance, represents a maximum acceptable speed variance threshold value.

[0009] In a possible implementation, the trajectory cost weight coefficient, the trajectory iteration number weight coefficient, the trajectory obstacle distance weight coefficient, and the trajectory linear speed variance weight coefficient are calibrated by the following steps: difficulty scores as a target layer, and the cost of the normalized local optimal trajectory, the optimization iteration number of the normalized local optimal trajectory, the minimum distance between the normalized local optimal trajectory and the obstacle, and the variance of all linear speeds on the normalized local optimal trajectory as a criterion layer, to construct a hierarchical structure model; The four indexes in the criterion layer are compared with each other, and the relative importance is determined by comparing the safety, the smoothness, the calculation overhead, and the path economy, to construct a judgment matrix, and the maximum eigenvalue of the judgment matrix and the corresponding normalized eigenvector are calculated, wherein the priority of the safety is higher than that of the smoothness, the priority of the smoothness is higher than that of the calculation overhead, and the priority of the calculation overhead is higher than that of the path economy; A consistency index is calculated based on the maximum eigenvalue, the judgment matrix is subjected to consistency check based on the consistency index, and the normalized eigenvector of the judgment matrix that passes the consistency check is taken as the trajectory cost weight coefficient, the trajectory iteration number weight coefficient, the trajectory obstacle distance weight coefficient, and the trajectory linear speed variance weight coefficient.

[0010] In a possible implementation, the dynamic cost map is updated according to the following formula: ; ; In the formula, represents a grid coordinate point on the dynamic cost map the updated dynamic cost, represents a grid coordinate point on the dynamic cost map the dynamic cost before updating, represents the dynamic cost of the local optimal trajectory, represents a difficulty score, represents an upper limit of single injection dynamic cost, represents a sensitivity factor, represents a decay factor, represents the cost of a grid coordinate point on the static cost map .

[0011] In a possible implementation, the single-injection dynamic cost upper limit, the sensitivity factor and the decay factor are calibrated by the following steps: In a preset fifth test scenario, the sensitivity factor and the decay factor are fixed as corresponding initial values, the single-injection dynamic cost upper limit is adjusted, robot path planning and navigation are performed, so that the robot globally plans a path for the first time to stably avoid a first preset area in the fifth test scenario within a preset iteration range, and a calibration value of the single-injection dynamic cost upper limit is obtained; In a preset sixth test scenario, the single-injection dynamic cost upper limit is fixed as the corresponding calibration value, the decay factor is fixed as the corresponding initial value, the sensitivity factor is adjusted, robot path planning and navigation are performed, so that the robot actively avoids a second preset area with a difficulty score higher than a threshold value in the sixth test scenario, and does not avoid a third preset area with a difficulty score lower than the threshold value, and a calibration value of the sensitivity factor is obtained; The single-injection dynamic cost upper limit and the sensitivity factor are fixed as the corresponding calibration values, the decay factor is adjusted, temporary obstacles are set on a robot passing path in a preset seventh test scenario, robot path planning and navigation are performed, a learning time of the robot to bypass the temporary obstacles is obtained, the temporary obstacles are removed in the preset seventh test scenario, robot path planning and navigation are performed, a forgetting time of the robot to reselect the robot passing path is obtained, so that the learning time and the forgetting time are within corresponding preset time ranges, and a calibration value of the decay factor is obtained.

[0012] In a possible implementation, returning to step S102 comprises: obtaining an initial total cost of the global optimal path on a global cost map before updating, and a cumulative cost of an actual driving path in the global optimal path; calculating a new cost of a remaining path in the global optimal path on the global cost map after updating; if a sum of the cumulative cost and the new cost is greater than a product of a preset sensitivity coefficient and the initial total cost, returning to step S102.

[0013] In a second aspect, the present application further provides an electronic device comprising a memory and a processor; The memory is configured to store a program. The processor is coupled to the memory and configured to execute the program stored in the memory to implement the steps of the robot path planning method described above.

[0014] In a third aspect, the present application further provides a computer readable storage medium, wherein the computer readable storage medium stores a program or instructions, and the program or instructions are executed by a processor to implement the steps of the robot path planning method.

[0015] The robot path planning method provided by the present application has the beneficial effects that: the robot path planning method provided by the present application introduces a dynamic cost backfill mechanism, updates the dynamic cost of the robot executing a local optimal trajectory to a dynamic cost map, deeply integrates global path planning and local path optimization, enables the robot to learn from experience and optimize itself, actively avoids historical difficult-to-pass areas in the path planning process, and improves the navigation stability and long-term operation efficiency of the robot in a complex and narrow environment. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 An embodiment flowchart of the robot path planning method provided by the present application is shown in the figure. Figure 2 An embodiment flowchart of the method for calibrating the weight coefficients of the multi-objective weighted optimization function provided by the present application is shown in the figure. Figure 3 An embodiment flowchart of the method for calibrating the weight coefficients of the difficult degree scoring calculation formula provided by the present application is shown in the figure. Figure 4 An embodiment flowchart of the method for calibrating the parameters of the dynamic cost map update formula provided by the present application is shown in the figure. Figure 5 An embodiment flowchart of the method for calibrating the parameters of the dynamic cost map update formula provided by the present application is shown in the figure. Figure 1 An embodiment flowchart of step S104 in the method for calibrating the parameters of the dynamic cost map update formula provided by the present application is shown in the figure. Figure 6 An embodiment structure diagram of the electronic device provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0018] The present application provides a robot path planning method, an electronic device and a medium, which are described below respectively.

[0019] Figure 1 An embodiment flowchart of the robot path planning method provided by the present application is shown in the figure. Figure 1 As shown in the figure, the robot path planning method comprises: S101, convert the acquired static grid map into a static cost map, and initialize a dynamic cost map, merge the static cost map and the dynamic cost map to obtain a global cost map; It should be noted that the static grid map is converted into a static cost map This process is mainly achieved through the obstacle inflation technology, and the specific steps include: first, the cells marked as "occupied" (such as walls) in the static grid map are set as "lethal obstacles" in the static cost map, and the cost value is set as a high value (for example, 254). Subsequently, a preset inflation radius is inflated outward from the "lethal obstacle" (the radius is usually set according to the physical size of the robot body), and the cost value of the cells within the inflation radius is attenuated according to the distance from the nearest "lethal obstacle": the closer to the obstacle, the higher the cost value (for example, when the robot center is located in this cell, it will be set as "inscribed obstacle" cost 253); the farther, the lower the cost value, until it is reduced to 0 (completely free). The static cost map generated by this contains a "potential field" with decreasing cost values around the physical obstacles.

[0020] It should also be noted that the dynamic cost map is used to store the "historical difficulty of passing" calculated in the subsequent steps, and when the system is first run, the dynamic cost map is created as a grid map with the same size as the static cost map, and the cost value of all cells is initialized to 0, so when the global path planning module runs the global path planning algorithm for the first time, the dynamic cost value is 0, and the global path planning algorithm is searched on the static cost map to obtain a global optimal path from the global starting point to the global target point.

[0021] It should also be noted that the global cost map is a combined cost function for global path search, and when searching for a path using the global path planning algorithm, the total passing cost of each grid coordinate point evaluated is obtained by adding the static cost and the dynamic cost of the point: ; Therefore, when planning, the global path planning algorithm will consider the static cost value of the physical obstacle and the dynamic cost value learned by the "dynamic cost value backfilling mechanism" of the application to find an optimal path with the lowest global total cost.

[0022] S102. Based on the preset global path planning algorithm, perform global path planning on the global cost map to obtain the global optimal path from the current location point to the global target point; In some embodiments of the present invention, the global path planning algorithm is as follows: (A-Star) algorithm. It should be noted that the choice of global path planning algorithm can be adjusted according to actual needs, and can also be... Algorithms such as Dijkstra's are not limited to... The algorithms will not be elaborated on here.

[0023] S103. Based on the preset local path planning algorithm and the global optimal path, perform local path planning to obtain the local optimal trajectory and its corresponding trajectory planning feature parameters, and control the robot to execute the local optimal trajectory until the global target point is reached. Considering that the commonly used local path planning algorithm in the existing technology is the dynamic window algorithm, which is based on the principle of instantaneous velocity sampling, its planning is short-sighted and lacks overall optimization for the spatiotemporal consistency of the trajectory. Especially in narrow corners or areas with dense obstacles, robots are prone to trajectory oscillation, repeated forward and backward movements, or even local jamming. In addition, the dynamic window algorithm also has insufficient control over the accuracy of the destination, which is difficult to meet some scenarios with high-precision docking requirements, such as the grasping and delivery of express packages. Small errors can lead to grasping or delivery failures, which seriously restricts the overall success rate of the task. To solve the limitations of the dynamic window algorithm, in some embodiments of the present invention, the Timed Elastic Band (TEB) algorithm is introduced as a local planner, that is, the local path planning algorithm is the Timed Elastic Band algorithm. The Timed Elastic Band algorithm can generate a smooth, continuous local path that conforms to robot dynamics by jointly optimizing the pose sequence and timestamp of a trajectory within a local window, which significantly improves the robot's passability in narrow spaces and the accuracy of destination control.

[0024] It should be noted that the time-elastic band algorithm involves the robot, during its journey along the globally optimal path, extracting a local reference path and a local target point within a preset range from the globally optimal path in each planning cycle. Based on a multi-objective weighted optimization function and a G2O solver, a locally optimal trajectory is obtained. The overall difficulty includes the locally optimal path and the linear and angular velocity motion control parameters, enabling the robot to continuously move towards the local target point until it reaches the final global target point. The multi-objective weighted optimization function is as follows: ; In the formula, This indicates the trajectory to be optimized. This represents the cost of the trajectory to be optimized. Indicates the cost of time. representing an obstacle cost, representing a speed constraint cost, representing an acceleration constraint cost, representing a time weight coefficient, representing an obstacle weight coefficient, representing a speed constraint weight coefficient, representing an acceleration constraint weight coefficient.

[0025] Further, in order to achieve the optimal balance of the navigation performance of the robot in different working conditions, in some embodiments of the present application, as shown in the following, Figure 2 The time weight coefficient, the obstacle weight coefficient, the speed constraint weight coefficient, and the acceleration constraint weight coefficient are calibrated by the following steps: S201, in a preset first test scene, fixing the time weight coefficient and the acceleration constraint weight coefficient as corresponding initial weight coefficients, adjusting the obstacle weight coefficient and the speed constraint weight coefficient for path planning and navigation, so that in the case that the task completion rate parameter meets the preset requirement, the minimum distance parameter between the robot and the obstacle is maximized, and the speed tracking error parameter is minimized, to obtain the calibration values of the obstacle weight coefficient and the speed constraint weight coefficient; It should be noted that before testing, a typical working condition library needs to be constructed, a series of simulation or physical test scenes are established, covering the typical application environment expected by the present application, for example: narrow channel, dynamic obstacle dense area, U-shaped turning, and high-precision parking point, i.e. the preset first test scene; setting performance evaluation indicators (KPIs): defining key indicators for quantifying navigation effects, including but not limited to: task success rate, total travel time, path smoothness (e.g. variance or root mean square value of speed / acceleration), minimum safety distance from obstacles; systematic iterative optimization is based on a set of balanced initial weight coefficients, and iterative adjustment and verification are performed according to priority and trade-off relationship.

[0026] It should be further noted that the calibration of the obstacle weight coefficient and the speed constraint weight coefficient has the highest priority, the obstacle weight coefficient is adjusted first to ensure that the robot can maintain a sufficient safety distance from the obstacle in all preset first test scenes (especially in narrow channels), and increasing this weight coefficient will make the robot more inclined to keep away from the obstacle, but too high may lead to the inability to find a path in narrow spaces; the speed constraint weight coefficient is adjusted simultaneously to ensure that the robot strictly obeys the preset global maximum speed, minimum speed, or speed limit requirements in specific areas (such as turning), and increasing this weight coefficient can enhance the tracking ability of the robot to the speed curve.

[0027] S202、in the preset first test scene, the obstacle weight coefficient and the speed constraint weight coefficient are fixed as the corresponding calibrated values, the time weight coefficient is fixed as the corresponding initial weight coefficient, and the acceleration constraint weight coefficient is adjusted for path planning and navigation, so that the task completion rate parameter, the minimum distance between the robot and the obstacle parameter, the speed tracking error parameter and the response speed parameter all meet the corresponding preset requirements, and the path smoothness and the mechanical jitter degree parameter are minimized, to obtain the calibrated value of the acceleration constraint weight coefficient; It should be noted that after meeting the safety and speed constraints, adjusting the acceleration constraint weight coefficient and increasing the acceleration constraint weight coefficient will inhibit the rapid acceleration and deceleration of the robot, so that the running trajectory is smoother and the mechanical jitter is smaller, which is particularly important for high-precision parking or transportation of sensitive goods, but too high may sacrifice the response speed.

[0028] S203、in the preset first test scene, the obstacle weight coefficient, the speed constraint weight coefficient and the acceleration constraint weight coefficient are fixed as the corresponding calibrated values, and the time weight coefficient is adjusted for path planning and navigation, so that the task completion rate parameter, the minimum distance between the robot and the obstacle parameter, the speed tracking error parameter, the response speed parameter, the path smoothness and the mechanical jitter degree parameter all meet the corresponding preset requirements, and the total travel time parameter is minimized to obtain the calibrated value of the time weight coefficient.

[0029] It should be noted that under the premise of ensuring safety, smoothness and complying with speed constraints, the time weight coefficient is finally adjusted, and the time weight coefficient is moderately increased to encourage the robot to find a path with shorter time, thereby shortening the task execution time, but too high time weight coefficient will lead to too aggressive trajectory, thereby sacrificing safety threshold or smoothness.

[0030] In the typical working condition library, the test is repeatedly run, the KPI performance under different weight coefficient combinations is compared, and a group of weight coefficient values that can achieve the best balance between safety, smoothness, efficiency and constraint compliance is selected as the final configuration parameters of the system.

[0031] S104、based on the trajectory planning feature parameters, determine the dynamic cost of the locally optimal trajectory, update the dynamic cost map based on the dynamic cost of the locally optimal trajectory, and then update the global cost map, and return to step S102; In order to calculate the dynamic cost of the locally optimal trajectory with higher quality, in some embodiments of the present application, in step S104, based on the trajectory planning feature parameters, the dynamic cost of the locally optimal trajectory is determined, and the dynamic cost map is updated based on the dynamic cost of the locally optimal trajectory, which comprises: calculating the difficulty score based on the trajectory planning feature parameters; based on the difficulty score, the dynamic cost of the locally optimal trajectory is calculated, and the dynamic cost map is updated based on the dynamic cost of the locally optimal trajectory.

[0032] wherein the trajectory planning characteristic parameters include a cost of the local optimal trajectory, a minimum distance between the local optimal trajectory and the obstacle, an optimization iteration number of the local optimal trajectory, and a variance of all linear velocities on the local optimal trajectory, and a calculation formula of the difficulty score is: ; ; ; ; ; wherein, represents the difficulty score, represents the cost of the local optimal trajectory, represents the optimization iteration number of the local optimal trajectory, represents the minimum distance between the local optimal trajectory and the obstacle, represents the variance of all linear velocities on the local optimal trajectory, represents the normalized optimization iteration number of the local optimal trajectory, represents the normalized minimum distance between the local optimal trajectory and the obstacle, represents the normalized cost of the local optimal trajectory, represents the normalized variance of all linear velocities on the local optimal trajectory, represents a trajectory cost weight coefficient, represents a trajectory iteration number weight coefficient, represents a trajectory obstacle distance weight coefficient, represents a trajectory linear velocity variance weight coefficient, represents a trajectory cost reference value, represents a maximum iteration number, represents a distance of interest, represents a minimum safety distance, represents a maximum acceptable velocity variance threshold value.

[0033] Further, in order to objectively reflect the contribution degree of different difficulty indicators to the difficulty score, in some embodiments of the present application, as shown in the following formula (2), the trajectory cost weight coefficient, the trajectory iteration number weight coefficient, the trajectory obstacle distance weight coefficient and the trajectory linear velocity variance weight coefficient are calibrated by the following steps: Figure 3 ​S301, taking the difficulty score as a target layer, taking the cost of the normalized local optimal trajectory, the optimization iteration number of the normalized local optimal trajectory, the minimum distance between the normalized local optimal trajectory and the obstacle, and the variance of all linear velocities on the normalized local optimal trajectory as a criterion layer, and constructing a hierarchical structure model; S302, comparing the four indexes in the criterion layer with each other, determining the relative importance by comparing the safety, stability, calculation overhead and path economy, constructing a judgment matrix, calculating the maximum eigenvalue of the judgment matrix and the corresponding normalized eigenvector, the priority of safety is greater than that of stability, the priority of stability is greater than that of calculation overhead, and the priority of calculation overhead is greater than that of path economy; It should be noted that according to the purpose of the application, that is, to improve the navigation stability, task endpoint accuracy and long-term operation efficiency of the robot in a complex narrow environment, the four indexes in the criterion layer are compared with each other by the skilled person to determine their relative importance, and the 1-9 scale method is used to construct the judgment matrix As an example, the skilled person can determine according to the priority principle of "safety">"stability">"calculation overhead">"path economy". For example: "obstacle distance" (corresponding to ) is "obviously important" than "speed variance" (corresponding to ), , .

[0034] "obstacle distance" (corresponding to ) is "strongly important" than "iteration number" (corresponding to ), , .

[0035] The rest is similar, and .

[0036] It should also be noted that the maximum eigenvalue and the corresponding normalized eigenvector of the judgment matrix are calculated by using mathematical methods (such as the sum product method or the square root method).

[0037] S303, calculating a consistency index based on the maximum eigenvalue, performing consistency test on the judgment matrix based on the consistency index, and taking the normalized eigenvector of the judgment matrix after passing the consistency test as the trajectory cost weight coefficient, the trajectory iteration number weight coefficient, the trajectory obstacle distance weight coefficient and the trajectory linear velocity variance weight coefficient.

[0038] It should be noted that the consistency check is to ensure the rationality of the judgment logic, and the consistency index is calculated (wherein ) and the consistency ratio (wherein is the preset average random consistency index), when , it is considered that the judgment matrix has satisfactory consistency, and after passing the consistency check, the normalized characteristic vector is the weight coefficient of each item of the difficulty score.

[0039] It can be understood that the weight coefficients of each item of the difficulty score can also be iteratively optimized according to a large number of simulation experiments or actual operation data through manual trial and error method or machine learning method to adapt to the needs of specific robot models or specific application scenarios (such as high-precision parking scenarios or narrow corner areas), which should be included in the protection scope of the present application.

[0040] Further, considering that the calculation of the difficulty score involves multiple normalization parameters, the calibration method thereof is crucial to the effect of the dynamic cost value back-feeding mechanism of the present application, and these normalization parameters include: trajectory cost reference value , maximum iteration number , attention distance , minimum safety distance , and maximum acceptable speed variance threshold .

[0041] Among them, the minimum safety distance defines the minimum safety distance threshold between the robot and the obstacle, when the minimum distance between the trajectory planned by the TEB algorithm and the nearest obstacle is less than or equal to , the minimum distance between the normalized local optimal trajectory and the obstacle is calibrated as the maximum value 1.0; when calibrating the minimum safety distance , the physical constraints need to be considered, i.e. the calibrated value of the minimum safety distance must be greater than the physical profile size of the robot, such as the envelope circle radius of the robot, and the error redundancy superposition needs to be considered, i.e. on the basis of the physical constraints, the system errors need to be considered, including the measurement error of the robot sensor (such as the 2D laser radar in the present embodiment) and the trajectory tracking error of the robot bottom controller (such as caused by servo delay or wheel slip), therefore, in some embodiments of the present application, the final calibrated value of the minimum safety distance is calculated and set by the following formula: ; In the formula, denotes the minimum safety distance, denotes the envelope radius of the robot, denotes the measurement error of the robot sensor, denotes the trajectory tracking error of the robot base controller.

[0042] It should be noted that the minimum safety distance After calibration, the robot is tested for passing through a real narrow passage (such as a door frame) to verify that the robot can stably pass through at this safety distance without triggering a collision due to slight control jitter, and to ensure The score can sensitively reflect the danger of a close-range obstacle.

[0043] The attention distance defines the upper limit of the distance at which the robot begins to pay attention to the obstacle and calculate the passing difficulty, when is greater than or equal to , the normalized obstacle distance score is calibrated to 0.0; when the attention distance is calibrated, the associated local planning horizon should be considered, that is, the value of the attention distance should match the local planning horizon of the TEB algorithm (i.e., the length of the local reference path to be cut off); if the obstacle is within the consideration range of the TEB algorithm, it should be included in the difficulty score calculation, and experience tuning should also be considered, that is, if the value of the attention distance is too small, the robot will be "short-sighted" and only accumulate difficulty when very close to the obstacle, which is not conducive to avoiding narrow areas in advance; if the value is too large, the robot will be "overly sensitive" and may accumulate unnecessary dynamic value due to obstacles far away in wide areas, therefore, in some embodiments of the present application, the attention distance is 1.0 to 1.5 times the local planning path length of the local path planning algorithm (TEB algorithm), or 30% to 50% of the effective detection distance of the robot's main sensor (such as a laser radar), and the best value is determined through simulation and field testing.

[0044] The maximum number of iterations is a normalized reference value for the number of optimization iterations of the TEB algorithm in a single planning period , used to quantify the calculation difficulty of the TEB solver (such as the g2o solver); in some embodiments of the present application, the maximum number of iterations is 80% to 90% of the reference maximum number of iterations that the solver of the preset local path planning algorithm can achieve in the preset planning period when the robot performs local path planning and navigation in the preset second test scenario.

[0045] It should be noted that the maximum number of iterations During calibration, the specific steps include: First, determining the planning period: First, determining the planning period of the local path planning algorithm, for example, 100ms (i.e., 10Hz); Next, conducting stress tests: Running robot navigation in multiple known complex and narrow test scenarios (e.g., sharp turns, U-turns, areas with dense obstacles, i.e., a preset second test scenario), and collecting data during the test: Recording the reference maximum number of iterations that the preset local path planning algorithm solver can achieve while ensuring no timeout (i.e., completing the calculation within the 100ms planning period); Finally, parameter tuning: Maximum number of iterations. Set to 80% to 90% of the maximum reference number of iterations observed in the stress tests described above. This calibration ensures that when Exceeding this threshold indicates that the solver is approaching its real-time limit, the planning computation complexity in this region is high, and the difficulty score is [not specified]. Consequently, it increased significantly.

[0046] Trajectory cost reference value It is the cost of the locally optimal trajectory output by the TEB algorithm. The normalized reference value, It is the TEB multi-objective weighted optimization function The calculation results comprehensively reflect the costs associated with time, obstacles, and velocity / acceleration constraints; in some embodiments of this invention, the trajectory cost reference value... This represents the average or 90th percentile cost of achieving the local optimal trajectory when performing local path planning and navigation for the robot in the preset third test scenario.

[0047] It should be noted that, regarding the reference value of trajectory cost During calibration, the specific steps include: designing baseline scenarios: designing standard test scenarios, such as "Scenario A" (straight-line driving in open terrain, a low-cost scenario) and "Scenario B" (low-speed driving in a narrow passage that is just passable, a high-cost scenario, i.e., a pre-set third test scenario); data acquisition: running the robot through "Scenario B" (high-cost scenario) several times, recording the output of the TEB algorithm during these runs. The average or 90th percentile value; final parameter tuning: set this statistic to This calibration ensures that the normalized score is only calculated when the overall cost of the trajectory reaches a preset "difficulty" level. Only then does it approach 1.0, thus accurately quantifying the difficulty of traversing the trajectory itself.

[0048] Maximum acceptable speed variance threshold It is the variance of all linear velocities on the local trajectory. a normalized reference value, which is intended to quantify the "smoothness" of the trajectory, for identifying the "trajectory oscillation, repeated advance and retreat" phenomenon mentioned in the background art; in some embodiments of the present application, the maximum acceptable speed variance threshold is the average or median of the linear speed variance on the locally optimal trajectory when the robot performs local path planning and navigation in a preset fourth test scenario, which is in front of a narrow door or facing a suddenly appearing dynamic obstacle.

[0049] It should be noted that the maximum acceptable speed variance threshold is calibrated as follows: reproduction of the problem scenario: a specific scenario that is prone to cause the robot to "oscillate" or "hesitate" is constructed (for example, in front of a narrow door or facing a suddenly appearing dynamic obstacle, i.e., the preset fourth test scenario); data collection: when the robot exhibits obvious "repeated advance and retreat" or "oscillation" behavior of dramatic speed fluctuations on the locally optimal path output by the TEB algorithm, the linear speed variance is collected; parameter setting: the average or median of the collected is set as . This calibration makes the score close to 0 when the robot moves smoothly (small ), and the score tends to 1.0 when the robot "oscillates" (dramatically increased), so that this "difficult experience" is injected into the dynamic cost map through a dynamic cost value feedback mechanism.

[0050] In some embodiments of the present application, the locally optimal trajectory includes a locally optimal path, the pose coordinate points on the locally optimal path are converted to grid coordinate points on the global cost map to form an influence area, and then the dynamic cost map is updated according to the following formula: ; ; wherein, represents the dynamic cost of the grid coordinate point on the dynamic cost map after updating, represents the dynamic cost of the grid coordinate point on the dynamic cost map before updating, represents the dynamic cost of the locally optimal trajectory, represents the difficulty score, represents the upper limit of the single injection dynamic cost, represents the sensitivity factor, represents the decay factor, represents the cost of the grid coordinate point on the static cost map.

[0051] In some embodiments of the present application, asFigure 4 As shown, the upper limit of dynamic cost per injection, sensitivity factor, and decay factor are calibrated using the following steps: S401. In the preset fifth test scenario, the sensitivity factor and attenuation factor are fixed to their corresponding initial values. The upper limit of the dynamic cost of a single injection is adjusted. Robot path planning and navigation are performed so that the number of iterations in which the robot first stably avoids the first preset area in the fifth test scenario during global path planning is stable within the preset iteration range, and the calibration value of the upper limit of the dynamic cost of a single injection is obtained. It should be noted that there is a maximum dynamic cost limit for a single injection. This parameter controls the maximum penalty value injected into the dynamic cost map when local path planning is difficult in a single attempt. This parameter directly affects the robot's "learning speed" and "memory strength" when avoiding difficult areas; it also controls the upper limit of the dynamic cost injected in a single attempt. The calibration employs the standard difficult scenario convergence analysis method. Specific steps include: constructing standard difficult scenarios: in a simulation environment or a controlled real environment, setting one or more standardized "static difficult" scenarios (e.g., a narrow S-shaped passage, a doorway requiring repeated posture adjustments to pass through, i.e., the fifth test scenario); setting evaluation metrics, such as learning efficiency (…). The robot repeatedly performs the same navigation task (e.g., traversing a difficult scenario 50 times). The algorithm begins to stably avoid the difficult region (i.e., the first preset region) during the Nth planning iteration. The smaller N is, the faster the learning speed and the better the path stability. This is achieved by observing whether the robot exhibits any issues during the learning process. Excessive risk leading to "over-avoidance" (e.g., choosing an extremely long detour to avoid minor difficulties); controlled variable experiment: fixed (Suggested initial value: 1.0) and (Suggested initial value: 0.95), select one group Candidate values ​​(e.g., 10, 30, 50, 80, 100), using each Run repeated navigation experiments, record evaluation metrics; analysis and decision-making: Too low: This results in an excessively large N value, leading to low learning efficiency and failing to demonstrate the advantage of "actively avoiding historically difficult areas." Too high: This results in an excessively small N value (e.g., N=1), leading to overly aggressive learning. The robot might permanently "blacklist" a usable path due to a single unsuccessful TEB planning attempt (e.g., sensor noise), resulting in poor path stability. Ideal value: Choose a value that keeps N stable within a reasonable range (e.g., N=3-5 times). The goal is to achieve a balance between "learning" and "stability".

[0052] S402. In the preset sixth test scenario, the upper limit of the dynamic cost of a single injection is fixed to the corresponding calibration value, the attenuation factor is fixed to the corresponding initial value, the sensitivity factor is adjusted, and robot path planning and navigation are performed so that the robot actively avoids the second preset area in the sixth test scenario where the difficulty score is higher than the threshold, and does not avoid the third preset area where the difficulty score is lower than the threshold, thereby obtaining the calibration value of the sensitivity factor. It should be noted that the sensitivity factor The dynamic cost calculation formula for a locally optimal trajectory The value of is used as an index to adjust the dynamic cost of local optimal trajectories. Difficulty rating The nonlinear response; the sensitivity factor calibration adopts the multi-difficulty gradient scenario comparison method, the specific steps of which include: constructing multi-difficulty scenarios: designing a test environment containing multiple different passage difficulties (i.e., the sixth test scenario), including area A (low difficulty): a slightly narrow straight passage (estimated difficulty score) Area B (High Difficulty): Extremely narrow corners with obstacles (estimated difficulty score below the threshold). (Difficulty score exceeds threshold); Set evaluation metric: Cost discrimination: After the robot passes through multiple times, the dynamic cost map In the study, the cost difference between region A and region B; path selection preference: observation. The algorithm's global planning result is either to avoid only B, or to avoid both A and B simultaneously; Controlled variable experiment: fixed (Using calibration values) and Select a group Candidate values ​​(e.g., 0.5, 1.0, 2.0); Analysis and decision-making: Improved the low The penalty weight of the value will cause the robot to be oversensitive to area A (low difficulty), which may lead to unnecessary avoidance; Linear relationship and Proportional; (Preferred) Increased the height The penalty weight of the value simultaneously suppresses low The penalty value allows the system to "tolerate" low-difficulty areas (Area A) but "severely punish" high-difficulty areas (Area B); Ideal value: To achieve the goal of "actively avoiding historically difficult areas" in the patent, especially for "complex and narrow environments", one should be selected; The value (e.g., 1.5 or 2.0) is used to enable intelligent differentiation of areas of different difficulty.

[0053] S403, fix the single injection dynamic cost upper limit and the sensitivity factor as the corresponding calibrated value, adjust the attenuation factor, set a temporary obstacle on the robot passage in a preset seventh test scene to perform robot path planning and navigation, obtain a learning time of the robot passing around the temporary obstacle, remove the temporary obstacle in the preset seventh test scene to perform robot path planning and navigation, obtain a forgetting time of the robot selecting the robot passage again, so that the learning time and the forgetting time are within a corresponding preset time range, and obtain a calibrated value of the attenuation factor.

[0054] It should be noted that the attenuation factor In the dynamic cost update rule, the control of the "forgetting speed" or "attenuation rate" of the dynamic cost over time determines the persistence of "historical experience"; it is calibrated by using a dynamic environment adaptability test method, and the specific steps include: constructing a dynamic change scene: stage 1 (learning) setting a temporary but difficult obstacle in the environment (for example, placing a box on the only way), the robot passes through multiple times and "learns" the high dynamic cost of the area, and starts to pass around, stage 2 (forgetting): removing the temporary obstacle; setting evaluation indexes: adaptation time (Tadapt): from stage 2 (obstacle removal) to the time or navigation times spent by the robot "forgetting" the historical difficulty and reselecting the path as the optimal path; control variable experiment: fixing and and (using the value calibrated in advance), a coefficient close to 1, a group of candidate values (for example: 0.90, 0.95, 0.99, 0.999) are selected; analysis and decision: Too low (for example 0.90): forgetting too fast, the experience learned by the robot in stage 1 is quickly attenuated, which may cause the robot to repeatedly attempt when the obstacle is still in place, reducing the avoidance efficiency; Too high (for example 0.999 or 1.0): memory is almost permanent, in stage 2, the robot cannot "forget" the removed obstacle, and will always avoid the path that has been restored to be smooth, which violates the adaptability of the system to the dynamic change of the environment; ideal value: The value of should match the update frequency of the cost map and the task dynamic, and a value (for example 0.995) that can ensure stable memory in stage 1 and "forget" old experience within a reasonable time (for example, several task cycles) in stage 2 should be selected.

[0055] In some embodiments of the present application, as shown in Figure 5 returning to step S102 in step S104 includes: ​S1041, obtain an initial total cost of the global optimal path on the global cost map before the update, and a cumulative cost of an actual driving path in the global optimal path; S1042, calculate a new cost of a remaining path in the global optimal path on the updated global cost map; S1043, if the sum of the cumulative cost and the new cost is greater than the product of a preset sensitivity coefficient and the initial total cost, return to step S102.

[0056] It should be noted that the calibration step of the sensitivity coefficient includes: constructing a standard test scenario: constructing one or more representative simulation or physical test environments, the environment should contain typical features targeted by the present application, such as narrow passages, U-shaped traps, dynamic human flow or obstacle simulation areas; set the candidate value interval: set a series of candidate values of the sensitivity coefficient according to experience. For example, test in an interval of [1.05, 1.5]. Wherein, the closer the sensitivity coefficient is to 1.0, the higher the sensitivity to cost changes, and the lower the threshold for triggering re-planning; the larger the sensitivity coefficient, the higher the tolerance; perform navigation test: in the standard test scenario, for each sensitivity coefficient candidate value, the robot performs multiple standard navigation tasks (e.g., from a fixed starting point to a fixed end point); statistical performance indicators: record and statistics the key performance indicators of the navigation task under different sensitivity coefficients, including but not limited to total task time, global re-planning times, total path cost or smoothness, and local planning failure times or stall time; trade-off to select the optimal value: analyze the performance indicator data, select a sensitivity coefficient candidate value that can achieve the best balance (Trade-off) between path optimality (can avoid high-cost paths in time) and planning overhead (avoid unnecessary frequent re-planning) as the final calibration value of the sensitivity coefficient.

[0057] It should also be noted that the "planning period" of the local path planning algorithm (TEB algorithm) is defined as the preset execution frequency of the TEB algorithm to update the local path, which is a configurable parameter of the robot system. In each planning period of the TEB algorithm, the cumulative cost of the actual driving path and the new cost of the remaining path are calculated, and the judgment is made.

[0058] In addition, if the local path planning algorithm (TEB algorithm) cannot solve the local optimal trajectory for 3 consecutive planning periods, return to step S102 global path planning algorithm (algorithm) to search on the global cost map to obtain a global optimal path from the current position to the target point.

[0059] ​Compared with the prior art, the application introduces a dynamic cost back-filling mechanism, updates the dynamic cost of the robot executing a local optimal trajectory to a dynamic cost map, deeply integrates global path planning and local path optimization, enables the robot to learn and optimize itself according to experience, and actively avoids historical difficult-to-pass areas in the path planning process, thereby improving the navigation stability and long-term operation efficiency of the robot in a complex and narrow environment.

[0060] As shown in Figure 6 The application also correspondingly provides an electronic device. The electronic device at least includes a processor 601 and a memory 602.

[0061] The processor 601 can include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 601 can be implemented in at least one of a hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), and a PLA (Programmable Logic Array). The processor 601 can also include a main processor and a coprocessor. The main processor is a processor for processing data in an awake state, also known as a CPU (Central Processing Unit). The coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 601 can be integrated with a GPU (Graphics Processing Unit) that is responsible for rendering and drawing the content required to be displayed by the display screen. In some embodiments, the processor 601 can also include an AI (Artificial Intelligence) processor for processing machine learning-related computing operations.

[0062] The memory 602 can include one or more computer-readable storage media, which can be non-transitory. The memory 602 can also include a high-speed random access memory, and a non-volatile memory such as one or more disk storage devices, flash storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 602 is used to store at least one instruction for being executed by the processor 601 to implement the robot path planning method provided by the method embodiment in the present application.

[0063] In some embodiments, the electronic device can further optionally include a peripheral device interface and at least one peripheral device. The processor 601, the memory 602 and the peripheral device interface can be connected through a bus or a signal line. Each peripheral device can be connected to the peripheral device interface through a bus, a signal line or a circuit board. Illustratively, the peripheral devices include, but are not limited to, a radio frequency circuit, a touch display screen, an audio circuit, a power supply and the like.

[0064] Of course, the electronic device can also include fewer or more components, and the present embodiments are not limited in this respect.

[0065] Correspondingly, the embodiments of the present application also provide a computer readable storage medium for storing computer readable programs or instructions, which are executed by a processor to implement the steps or functions of the robot path planning method provided by the above-mentioned method embodiments.

[0066] Those skilled in the art can understand that all or part of the processes of the above-mentioned embodiments can be completed by a computer program instructing related hardware. The above-mentioned program can be stored in a computer readable storage medium. The computer readable storage medium includes a magnetic disk, an optical disk, a read-only memory or a random access memory.

[0067] The above describes in detail a robot path planning method provided by the present application. The principles and implementation manners of the present application are described by using specific examples. The above-mentioned embodiments are only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manners and application ranges can be changed. In summary, the content of the present description should not be understood as a limitation of the present application.

[0068] The above describes only the preferred embodiments of the present application, but the protection scope of the present application is not limited to this. Any changes or replacements within the technical scope disclosed by the present application can be easily thought by those skilled in the art, which should be covered by the protection scope of the present application.

Claims

1. A robot path planning method, characterized by, The method comprises the following steps: S101, converting the obtained static grid map into a static cost map, initializing a dynamic cost map, merging the static cost map and the dynamic cost map, and obtaining a global cost map; S102, performing global path planning on the global cost map based on a preset global path planning algorithm to obtain a global optimal path from a current position point to a global target point; S103, performing local path planning based on a preset local path planning algorithm and the global optimal path to obtain a local optimal trajectory and corresponding trajectory planning characteristic parameters, and controlling the robot to execute the local optimal trajectory until the global target point is reached; S104, determining the dynamic cost of the local optimal trajectory based on the trajectory planning characteristic parameters, updating the dynamic cost map based on the dynamic cost of the local optimal trajectory, and further updating the global cost map, and returning to step S102.

2. The robot path planning method of claim 1, wherein, The global path planning algorithm adopts The local path planning algorithm adopts a TEB algorithm, and the local path planning algorithm performs local path planning through a multi-objective weighted optimization function, and the multi-objective weighted optimization function is: ； wherein denotes the trajectory to be optimized, denotes the cost of the trajectory to be optimized, denotes the time cost, denotes the obstacle cost, denotes the speed constraint cost, denotes the acceleration constraint cost, denotes the time weight coefficient, denotes the obstacle weight coefficient, denotes the speed constraint weight coefficient, denotes the acceleration constraint weight coefficient.

3. The robot path planning method of claim 2, wherein, The time weight coefficient, the obstacle weight coefficient, the speed constraint weight coefficient and the acceleration constraint weight coefficient are calibrated by the following steps: In a preset first test scene, the time weight coefficient and the acceleration constraint weight coefficient are fixed as the corresponding initial weight coefficients, the obstacle weight coefficient and the speed constraint weight coefficient are adjusted for path planning and navigation, so that in the case that the task completion rate parameter meets the preset requirement, the minimum distance parameter between the robot and the obstacle is maximized, and the speed tracking error parameter is minimized, to obtain the calibration values of the obstacle weight coefficient and the speed constraint weight coefficient; In a preset first test scene, the obstacle weight coefficient and the speed constraint weight coefficient are fixed as the corresponding calibration values, the time weight coefficient is fixed as the corresponding initial weight coefficient, and the acceleration constraint weight coefficient is adjusted for path planning and navigation, so that in the case that the task completion rate parameter, the minimum distance parameter between the robot and the obstacle, the speed tracking error parameter and the response speed parameter all meet the corresponding preset requirements, the path smoothness and mechanical jitter degree parameters are minimized, to obtain the calibration value of the acceleration constraint weight coefficient; In a preset first test scene, the obstacle weight coefficient, the speed constraint weight coefficient and the acceleration constraint weight coefficient are fixed as the corresponding calibration values, and the time weight coefficient is adjusted for path planning and navigation, so that in the case that the task completion rate parameter, the minimum distance parameter between the robot and the obstacle, the speed tracking error parameter, the response speed parameter, the path smoothness and the mechanical jitter degree parameter all meet the corresponding preset requirements, the total travel time parameter is minimized, to obtain the calibration value of the time weight coefficient.

4. The robot path planning method of claim 1, wherein, The method for determining the dynamic cost of the local optimal trajectory based on the trajectory planning characteristic parameters and updating the dynamic cost map based on the dynamic cost of the local optimal trajectory comprises the following steps: calculating a difficulty score based on the trajectory planning characteristic parameters; calculating the dynamic cost of the local optimal trajectory based on the difficulty score and updating the dynamic cost map based on the dynamic cost of the local optimal trajectory; The trajectory planning characteristic parameters include a cost of the local optimal trajectory, a minimum distance between the local optimal trajectory and an obstacle, an optimization iteration number of the local optimal trajectory, and a variance of all linear velocities on the local optimal trajectory, and a calculation formula of the difficulty score is: ； ； ； ； ； wherein, denotes the difficulty score, denotes the cost of the locally optimal trajectory, denotes the number of optimization iterations of the locally optimal trajectory, denotes the minimum distance of the locally optimal trajectory to the obstacles, denotes the variance of all linear velocities on the locally optimal trajectory, denotes the number of optimization iterations of the normalized locally optimal trajectory, denotes the minimum distance of the normalized locally optimal trajectory to the obstacles, denotes the cost of the normalized locally optimal trajectory, denotes the variance of all linear velocities on the normalized locally optimal trajectory, denotes the trajectory cost weight coefficient, denotes the trajectory iteration number weight coefficient, denotes the trajectory obstacle distance weight coefficient, denotes the trajectory linear velocity variance weight coefficient, denotes the trajectory cost reference value, denotes the maximum number of iterations, denotes the attention distance, denotes the minimum safety distance, denotes the maximum acceptable velocity variance threshold.

5. The robot path planning method of claim 4, wherein, The trajectory cost weight coefficient, the trajectory iteration number weight coefficient, the trajectory obstacle distance weight coefficient, and the trajectory linear velocity variance weight coefficient are calibrated through the following steps: The difficulty score is taken as a target layer, and the cost of the normalized local optimal trajectory, the optimization iteration number of the normalized local optimal trajectory, the minimum distance between the normalized local optimal trajectory and an obstacle, and the variance of all linear velocities on the normalized local optimal trajectory are taken as a criterion layer to construct a hierarchical structure model; The four indexes in the criterion layer are compared with each other, the relative importance is determined by comparing safety, stability, calculation overhead, and path economy, a judgment matrix is constructed, the maximum eigenvalue of the judgment matrix and a corresponding normalized eigenvector are calculated, the priority of the safety is higher than that of the stability, the priority of the stability is higher than that of the calculation overhead, and the priority of the calculation overhead is higher than that of the path economy; A consistency index is calculated based on the maximum eigenvalue, the judgment matrix is subjected to consistency test based on the consistency index, and the normalized eigenvector of the judgment matrix that passes the consistency test is taken as the trajectory cost weight coefficient, the trajectory iteration number weight coefficient, the trajectory obstacle distance weight coefficient, and the trajectory linear velocity variance weight coefficient.

6. The robot path planning method of claim 4, wherein, The dynamic cost map is updated according to the following formula: ； ； wherein, represents a dynamic cost of a local optimal trajectory, represents a difficulty score, represents an upper bound of a single injection dynamic cost, represents a sensitivity factor, represents a grid coordinate point on a dynamic cost map an updated dynamic cost, represents a grid coordinate point on a dynamic cost map a pre-updated dynamic cost, represents a decay factor, represents a grid coordinate point on a static cost map a cost.

7. The robotic path planning method of claim 6, wherein, The single-injection dynamic cost upper limit, the sensitivity factor, and the attenuation factor are calibrated through the following steps: In a preset fifth test scene, the sensitivity factor and the attenuation factor are fixed as corresponding initial values, the single-injection dynamic cost upper limit is adjusted, robot path planning and navigation are performed, so that the iteration number of the robot when starting to stably avoid a first preset area in the fifth test scene for the first time is stabilized within a preset iteration range, and a calibration value of the single-injection dynamic cost upper limit is obtained; In a preset sixth test scene, the single-injection dynamic cost upper limit is fixed as a corresponding calibration value, the attenuation factor is fixed as a corresponding initial value, the sensitivity factor is adjusted, robot path planning and navigation are performed, so that the robot actively avoids a second preset area in the sixth test scene with a difficulty score higher than a threshold value, and does not avoid a third preset area with a difficulty score lower than the threshold value, and a calibration value of the sensitivity factor is obtained; The single-injection dynamic cost upper limit and the sensitivity factor are fixed as corresponding calibration values, an attenuation factor is adjusted, a temporary obstacle is arranged on a robot passing path in a preset seventh test scene to perform robot path planning and navigation, a learning time of the robot to bypass the temporary obstacle is obtained, the temporary obstacle is removed in the preset seventh test scene to perform robot path planning and navigation, a forgetting time of the robot to reselect the robot passing path is obtained, the learning time and the forgetting time are within a corresponding preset time range, and a calibration value of the attenuation factor is obtained.

8. The robotic path planning method of claim 1, wherein, Returning to step S102 includes: obtaining an initial total cost of the global optimal path on a global cost map before updating, and a cumulative cost of an actual driving path in the global optimal path; calculating a new cost of a remaining path in the global optimal path on the global cost map after updating; if a sum of the cumulative cost and the new cost is greater than a product of a preset sensitivity coefficient and the initial total cost, returning to step S102.

9. An electronic device, comprising: comprising a memory and a processor; the memory is configured to store a program; the processor is coupled to the memory and is configured to execute the program stored in the memory to implement steps of the robot path planning method in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The readable storage medium has a program or instruction stored thereon, and the program or instruction is executed by the processor to implement steps of the robot path planning method in any one of claims 1 to 8.

Images