A terrain-adaptive navigation method and system for a robot

Abstract

The application relates to a terrain self-adaptive navigation method and system of a robot, and relates to the technical field of robot terrain navigation, which comprises the following steps: acquiring the moving speed of the robot in a past time window, collecting current moving road surface data, and identifying and obtaining a road surface terrain map; a first moving position is randomly selected, corresponding first moving speed and amplitude are obtained, a first moving stability coefficient is obtained in combination with first road surface terrain information; a moving error range is configured, a first moving position range is divided and obtained, and a first fusion stability coefficient is calculated and obtained according to first range road surface terrain information; a first navigation fitness is calculated and obtained, and an optimal moving position is obtained through moving position optimization selection. The application solves the problem that in traditional terrain self-adaptive navigation, it is difficult to balance the safety and smoothness in the moving process of the robot, the risk of tilting during moving is high, the moving rhythm is frequently interrupted, and the navigation effect is affected.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of robot terrain navigation, in particular to a robot terrain adaptive navigation method and system. BACKGROUND

[0002] With the expansion of robot application scenarios to complex terrains, the stability and smoothness of terrain adaptive navigation have become key technical problems affecting the efficiency of robot operation.

[0003] Currently, the traditional terrain adaptive navigation method lacks adaptability to the moving state of the robot and the terrain environment, and is difficult to dynamically adjust the navigation decision to cope with complex road conditions, which easily leads to the risk of robot tipping over or interruption of the marching rhythm during movement. SUMMARY

[0004] The present application provides a robot terrain adaptive navigation method and system, which improves the status quo of poor robot stability and disturbed movement continuity in traditional terrain adaptive navigation due to the difficulty in adapting to complex terrain conditions and the inability to dynamically balance movement safety and smoothness.

[0005] The embodiments of the present application disclose the following technical solutions:

[0006] In a first aspect, the embodiments of the present application provide a robot terrain adaptive navigation method, which comprises:

[0007] Obtaining the moving speed of the robot in the past time window, collecting the road surface data currently moving, and identifying and obtaining the road surface terrain map;

[0008] Randomly selecting a first moving position in the road surface terrain map, obtaining a corresponding first moving speed and a first moving amplitude, and analyzing and obtaining a first moving stability coefficient according to the first moving amplitude, the first moving speed and the first road surface terrain information of the first moving position;

[0009] According to the first moving stability coefficient, configuring a moving error range, dividing the range of the first moving position, obtaining a first moving position range, and analyzing and calculating a first fusion stability coefficient according to the first range road surface terrain information in the first moving position range;

[0010] According to the first moving speed, the moving speed and the first fusion stability coefficient, a first navigation fitness is calculated and obtained, and an optimal moving position is selected by optimizing the moving position.

[0011] In a second aspect, the embodiments of the present application provide a robot terrain adaptive navigation system, which comprises:

[0012] The road surface terrain map acquisition module is configured to acquire the moving speed of the robot within a past time window and collect road surface data of the current moving, and identify and obtain a road surface terrain map;

[0013] The moving stability coefficient analysis module is configured to randomly select a first moving position in the road surface terrain map, acquire a corresponding first moving speed and a first moving amplitude, and analyze and obtain a first moving stability coefficient according to the first moving amplitude, the first moving speed, and first road surface terrain information of the first moving position;

[0014] The fusion stability coefficient calculation module is configured to configure a moving error range according to the first moving stability coefficient, divide the first moving position into ranges to obtain a first moving position range, and analyze and calculate a first fusion stability coefficient according to first range road surface terrain information in the first moving position range.

[0015] The optimal moving position selection module is configured to calculate a first navigation fitness according to the first moving speed, a moving speed, and the first fusion stability coefficient, and select an optimal moving position through moving position optimization.

[0016] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0017] The present application provides a terrain adaptive navigation method and system for a robot, which realizes stable and accurate navigation of the robot in complex terrain through the cooperative operation of constructing a road surface terrain map by acquiring historical moving speed of the robot and current road surface data, selecting a moving position in the terrain map and calculating a moving stability coefficient thereof, dynamically configuring an error range and a fusion stability coefficient based on the moving stability coefficient, calculating a navigation fitness in combination with speed adaptability, and iteratively selecting an optimal moving position. First, the moving speed of the robot within a past time window is acquired, and road surface point cloud data in front is collected and a road surface terrain map is constructed. Then, a first moving position is randomly selected in the terrain map, a corresponding moving amplitude and speed are calculated, terrain information is input into a moving stability analyzer in combination, and a first moving stability coefficient is obtained. Based on the ratio of the first moving stability coefficient to an average moving stability coefficient, a moving error range is dynamically adjusted and a first moving position range is divided, a plurality of error positions in the range are selected to calculate a moving stability coefficient average, and a first fusion stability coefficient is obtained. Finally, a first speed stability coefficient is obtained by calculating the similarity of the first moving speed and a moving speed, a navigation fitness is obtained by fusing the two types of stability coefficients, a position with the largest navigation fitness is iteratively selected as an optimal moving position, and the robot is guided to navigate.

[0018] The technical scheme of the present application solves the problems of poor terrain adaptability, easy instability and motion jamming in traditional robot navigation caused by fixed error range, single position evaluation and neglect of speed continuity, realizes adaptive perception and safe and smooth navigation control of the robot on different complex terrains, and improves the work efficiency of the robot in complex terrain scenes. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0020] Figure 1 A flowchart of a terrain adaptive navigation method of a robot provided by the embodiment of the present application is shown in the figure.

[0021] Figure 2 A structural schematic diagram of a terrain adaptive navigation system of a robot provided by the embodiment of the present application is shown in the figure.

[0022] In the drawings, the components represented by the numbers are described as follows:

[0023] The road terrain map acquisition module 01, the movement stability coefficient analysis module 02, the fusion stability coefficient calculation module 03 and the optimal movement position selection module 04. DETAILED DESCRIPTION

[0024] The present application provides a terrain adaptive navigation method and system of a robot, which is used to solve the technical problem in the prior art that the traditional terrain adaptive navigation is difficult to balance the movement stability and continuity of the robot, and the stability is poor and the movement rhythm is disturbed in a complex road surface environment, resulting in unsatisfactory navigation effect.

[0025] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings of the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0026] In the description of the present application, the terms "first", "second" are only for descriptive purposes and cannot be understood to indicate or imply relative importance or implicitly indicate the number of the indicated technical features. Therefore, the features defined with "first", "second" can be explicitly or implicitly included one or more of the features. In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0027] In the description of the present application, the term "for example" is used to indicate "as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily interpreted as more preferred or more advantageous than other embodiments. The following description is given in order to enable any person skilled in the art to implement and use the present application. In the following description, details are listed for the purpose of explanation. It should be understood that those skilled in the art can recognize that the present application can be implemented without using these specific details. In other examples, well-known structures and processes will not be described in detail to avoid unnecessary details making the description of the present application obscure. Therefore, the present application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope consistent with the principles and characteristics disclosed in the present application.

[0028] Embodiment one, as shown in the accompanying Figure 1 The present application provides a terrain adaptive navigation method of a robot, the method comprising the following steps:

[0029] S110: Obtain the moving speed of the robot in the past time window, and collect the road surface data currently moving, and identify the road surface terrain map;

[0030] In the present application, in the scenario of robot terrain adaptive navigation, in order to accurately obtain the historical motion reference of the robot and real-time terrain information, speed calculation and point cloud acquisition modeling are required to realize comprehensive perception of the robot motion state and road surface environment, and provide data support for subsequent stability analysis and optimal moving position selection.

[0031] Specifically, first, for the motion of the robot in the past time window, the corresponding moving distance data in the time window is comprehensively collected, and the specific time length of the time window is recorded. The moving distance and the time length are ratio operated, that is, the moving speed of the robot in the past time window is accurately calculated and obtained. The moving speed will be used as a reference for subsequent evaluation of moving continuity.

[0032] At the same time, for the road surface environment in the current moving process of the robot, the road surface point cloud data in the preset area in front of the robot is collected. These road surface point cloud data need to cover the road surface features at different positions in the preset area, so as to completely reflect the terrain details such as ups and downs, pits and potholes of the road surface.

[0033] Further, based on the collected road point cloud data, a region coordinate system corresponding to a preset region is first constructed, and then the point cloud data is converted into a three-dimensional model and matched with the region coordinate system. In the matching process, the height information corresponding to each region coordinate needs to be clearly labeled, thereby forming a road terrain map containing coordinate height details.

[0034] Finally, the moving speed and the road terrain map obtained through the above steps can provide accurate data input for subsequent steps of selecting a moving position and analyzing a moving stability coefficient, thereby ensuring the scientificity and reliability of terrain adaptive navigation decision-making.

[0035] The step S110 in the method provided by the embodiment of the present application comprises:

[0036] The moving distance of the robot in a past time window is obtained, and the moving speed is calculated in combination with the time length of the time window;

[0037] The road point cloud data in a preset region in front of the current robot is collected;

[0038] A road terrain map is constructed according to the road point cloud data.

[0039] In the embodiment of the present application, in order to realize accurate perception of the current moving road surface by the robot and scientific decision-making of the subsequent navigation position, the historical moving speed calculation, the front road point cloud data collection and the three-dimensional road terrain map construction are required to provide accurate data support for subsequent moving stability analysis and optimal position selection.

[0040] Specifically, first, the moving distance of the robot in a past time window is obtained, and the moving speed is calculated in combination with the time length of the time window, so as to determine the current basic motion state of the robot through historical motion data, avoid disconnection between the subsequent navigation position selection and the existing moving speed, and ensure motion continuity.

[0041] The length of the time window can be adaptively adjusted according to the motion characteristics of the robot and the complexity of the environment, so as to balance the real-time and historical representativeness of the data.

[0042] For example, if the moving distance of the robot in the past 20 seconds is 5 meters, and the time window length is 20 seconds, the moving speed can be calculated by the formula "moving speed = moving distance / time window length" to be 0.25 meters / second.

[0043] Further, the road point cloud data in a preset region in front of the current robot is collected.

[0044] The front preset area needs to be set in combination with the robot moving step and the sensor detection range, for example, for a robot with a step of 0.3-0.5 meters, a rectangular area of 1-2 meters in front, 0.5 meters on the left and right can be set as a preset area to ensure that the collected terrain data can cover the potential range of the next foot landing.

[0045] Specifically, the collection of point cloud data can be realized by sensors such as laser radar and depth camera carried by the robot. These sensors can capture three-dimensional coordinate information of the road surface with millimeter-level precision, including terrain details such as protrusions, depressions, and potholes.

[0046] For example, the laser radar can collect 500-1000 points of cloud data per square meter in a preset area of 1.5 meters in front and 1 meter wide, each point containing “X (horizontal transverse), Y (horizontal longitudinal), Z (height)” three-dimensional coordinate information, which is the core data source for subsequent construction of accurate road surface terrain map.

[0047] On this basis, a road surface terrain map is constructed according to the collected road surface point cloud data to realize digital representation of the current moving road surface terrain characteristics.

[0048] The method provided in the embodiments of the present application includes:

[0049] Constructing a regional coordinate system of the preset area;

[0050] According to the road surface point cloud data, a three-dimensional model in the preset area is constructed, and the coordinate matching with the regional coordinate system is performed to obtain a road surface terrain map, wherein the road surface terrain map includes the height information of each regional coordinate.

[0051] In the embodiments of the present application, in order to realize accurate perception and safe navigation of the robot in a complex road surface environment, the discrete point cloud data is converted into a structured road surface terrain map through the cooperative processing of coordinate system anchoring, three-dimensional modeling and coordinate matching, so as to ensure that the robot foot landing position is highly adapted to the actual road surface conditions.

[0052] Specifically, a regional coordinate system of the preset area is first constructed. The regional coordinate system takes the current standing posture of the robot as a reference, selects the midpoint of the center line of the two feet as the coordinate origin, the X-axis extends along the forward direction (consistent with the moving speed direction), the Y-axis is the horizontal transverse (covering the left and right foot movement range), and the Z-axis is perpendicular to the ground (positive for protrusion, negative for depression), and the unit is unified as meters.

[0053] The design of the region coordinate system directly associates the terrain data with the robot's own position, for example, the coordinates (0.8m, 0.2m, 0.05m) can be intuitively interpreted as "0.8 meters in front, 0.2 meters to the right, and a 0.05-meter-high protrusion", providing a spatial reference for subsequent position determination.

[0054] Further, based on the collected road point cloud data, a three-dimensional model is constructed. The original road point cloud data needs to be filtered first. Statistical filtering is used to remove noise points such as dust and weeds, and radius filtering is used to remove isolated noise points to ensure data effectiveness.

[0055] Further, the existing greedy projection triangulation algorithm is used to convert the effective road point cloud data into a continuous three-dimensional grid model to completely restore the micro features of the road surface such as protrusions, depressions, and potholes.

[0056] For example, for a stone with a diameter of 0.3 meters and a height of 0.06 meters, the three-dimensional grid model can accurately present its edge profile and height changes to avoid missing key terrain that affects foot stability.

[0057] Further, the three-dimensional grid model is matched with the region coordinate system. That is, through the existing coordinate conversion algorithm, each vertex in the model corresponds to a unique "X, Y, Z" coordinate, and finally a road terrain map containing the height information of the entire region coordinate is generated.

[0058] For example, in a 1.5m x 1m preset area in front of a robot, the processed road terrain map contains 150 x 100 coordinate points, each corresponding to a specific height value. When (1.0m, 0m) is selected as the moving landing position, the height of this point can be directly obtained from the road terrain map as 0.01m (flat terrain).

[0059] The road terrain map constructed by the above steps not only ensures the spatial correlation of terrain data and robot movement scenarios, but also realizes complete quantization of terrain features, providing accurate data support for subsequent first moving position selection and stability coefficient calculation.

[0060] S120: Randomly select a first moving position in the road terrain map, obtain a corresponding first moving speed and a first moving amplitude, and analyze the first moving stability coefficient based on the first moving amplitude, the first moving speed, and the first road terrain information of the first moving position.

[0061] In the robot terrain adaptive navigation scenario, in the embodiments of the application, in order to accurately evaluate the safety of the moving landing position, the cooperative operation of the moving position selection, the amplitude and speed matching, the terrain information association and the stability analysis is required to provide a quantitative stability basis for subsequent moving decision, and to avoid the instability of the robot steps caused by single dimension judgment.

[0062] Specifically, first, a first moving position is randomly selected in the constructed road surface terrain map.

[0063] The first moving position needs to focus on the potential range of the next step landing of the robot, and is usually selected in the interval of 0.3-0.8 meters of the X-axis (forward direction) and -0.3 meters to 0.3 meters of the Y-axis (left-right transverse direction) of the terrain map along the forward direction starting from the current position, and needs to avoid the extreme protrusion or depression area marked in the terrain map to ensure the basic safety of the initial moving position.

[0064] Further, the first moving amplitude is calculated according to the spatial relationship between the first moving position and the current position of the robot. That is, the midpoint of the connecting line between the two feet of the current position is taken as the reference point, and the straight-line distance between the first moving position and the reference point in the X-Y plane is calculated, which is the first moving amplitude, reflecting the moving span of the robot foot from the current position to the first moving position.

[0065] Further, the first road surface terrain information of the first moving position coverage range is divided in the road surface terrain map. Considering that the actual landing of the robot foot has a contact area, a rectangular area covered by the foot is formed by extending a certain distance in the positive and negative directions of the X-axis and the Y-axis with the first moving position as the center, and the Z values of all coordinate points in the rectangular area are extracted to constitute the first road surface terrain information.

[0066] Finally, the obtained first moving amplitude, first moving speed and first road surface terrain information are input into the moving stability analyzer, and the first moving stability coefficient is output through the feature extraction and training analysis of the moving stability analyzer, so as to quantitatively evaluate the landing safety degree of the first moving position.

[0067] This step realizes the accurate judgment of the stability of a single moving position by fusing and analyzing the multidimensional data of the moving parameters and the terrain parameters, provides a quantitative basis for subsequent configuration of the moving error range according to the stability coefficient and further optimization of the moving position, and avoids the stability evaluation deviation caused by relying on a single parameter.

[0068] The step S120 in the method provided in the embodiments of the application includes:

[0069] randomly selecting a first moving position in the road surface topographic map, calculating a first moving amplitude according to the first moving position and a current position of the robot, and classifying a first moving speed according to the first moving amplitude, wherein the moving amplitude and the moving speed of the robot have a mapping relationship;

[0070] dividing first road surface topographic information in a range covered by the first moving position in the road surface topographic map;

[0071] inputting the first moving amplitude, the first moving speed and the first road surface topographic information into a moving stability analyzer, and outputting a first moving stability coefficient.

[0072] In the embodiments of the present application, in order to realize scientific evaluation and safe navigation decision of the robot on the moving position, the moving amplitude, the matching moving speed and the corresponding road surface topographic information of the moving position are inputted into the moving stability analyzer, and the moving stability coefficient is outputted, so as to provide quantifiable stability basis for subsequent moving error range configuration and optimal position selection.

[0073] Specifically, first, a first moving position is randomly selected in the road surface topographic map. The selection range needs to be strictly limited in the potential area of the next foot landing of the robot, and usually a rectangular area of 0.3-0.8 meters along the X axis (forward direction) and -0.3 to 0.3 meters along the Y axis (left-right transverse direction) is randomly sampled based on the current position, so as to ensure that the moving position meets the physiological movement limit of the robot.

[0074] At the same time, the extreme dangerous area marked in the road surface topographic map needs to be excluded in the selection process, for example, the convex with Z value > 0.1 meter or the concave with Z value < -0.08 meter, so as to initially screen out the candidate points with basic safety.

[0075] Exemplarily, in the effective range of 1.5 meters by 0.6 meters, the coordinate (0.55m, 0.08m) is selected as the first moving position by generating a random number, which corresponds to a gentle area with a Z value of 0.02 meters in the road surface topographic map, and meets the initial landing condition.

[0076] Further, the first moving amplitude is calculated according to the first moving position and the current position of the robot. That is, taking the midpoint of the connecting line between the two feet of the current position (coordinate origin) as the reference, the straight-line distance between the first moving position (X1, Y1) and the origin is calculated by using the Euclidean distance formula, that is, the first moving amplitude .

[0077] Exemplarily, the current position origin is (0m, 0m), and the first moving position is (0.55m, 0.08m). The first moving amplitude is calculated to be meters.

[0078] At the same time, the first moving speed is obtained based on the robot's pre-stored movement amplitude-speed mapping table.

[0079] The movement amplitude-velocity mapping table was constructed using a large amount of experimental data. It divides the velocity levels according to amplitude range, such as "0.3-0.5 meters corresponds to 0.2-0.25 m / s, 0.5-0.7 meters corresponds to 0.25-0.3 m / s, and 0.7-0.8 meters corresponds to 0.3-0.35 m / s". The velocity values ​​in each amplitude range have been smoothed to ensure the continuity of the motion.

[0080] Furthermore, the road surface terrain information covering the first moving position is divided within the road surface terrain map. Considering the actual contact area of ​​the robot's foot, a coverage area is formed by extending from the first moving position to ±0.06 meters along the X-axis and ±0.04 meters along the Y-axis. The Z-value (height) and the rate of change of the Z-value (the ratio of the height difference between adjacent points to the horizontal distance) of all coordinate points within this area are extracted.

[0081] For example, the coverage area contains 960 coordinate points ranging from (0.49m, 0.04m) to (0.61m, 0.12m), with Z values ​​ranging from 0.01 to 0.03 meters. The Z value variation rate is all <0.04, indicating that the terrain of the area is flat and without significant undulations.

[0082] At the same time, key parameters such as the average height, maximum height, minimum height, and height change rate of all coordinate points in the area are extracted as the first road surface terrain information to comprehensively reflect the terrain undulation characteristics and overall flatness within the coverage area of ​​the first moving position.

[0083] Furthermore, the obtained first movement amplitude, first movement speed, and first road surface terrain information are input into the motion stability analyzer. After multi-dimensional feature fusion and analysis calculation by the motion stability analyzer, the first motion stability coefficient is output to accurately quantify the landing safety level of the first movement position.

[0084] The method provided in this application includes the following training and construction process for the "moving stability analyzer":

[0085] A mobile stability analyzer is built based on machine learning.

[0086] Based on historical data of robot movement control, a set of sample movement amplitude, a set of sample movement speed, and a set of sample road surface terrain information are collected. The proportion of robot instability under different movement conditions is collected, the sample movement stability coefficient is calculated, and the set of sample movement stability coefficients is obtained by labeling.

[0087] The motion stability analyzer is trained and optimized in a supervised manner using the set of sample motion amplitude, sample motion speed, sample road surface terrain information, and sample motion stability coefficients until the training is verified to have converged.

[0088] In this embodiment of the application, in order to achieve accurate prediction of robot movement stability, a movement stability analyzer needs to be built and trained based on historical movement data. The complex mapping relationship between movement amplitude, movement speed and road surface terrain is learned through machine learning model, so as to achieve accurate output of movement stability coefficient under different movement conditions, and provide core decision basis for robot adaptive navigation.

[0089] First, a motion stability analyzer is built based on a machine learning framework. This motion stability analyzer uses a gradient boosting decision tree model (GBDT) as its basic architecture to effectively capture the nonlinear relationship between "motion amplitude, motion speed, and road surface terrain information," adapting to the stability assessment needs of robots in different road surface scenarios and ensuring that the output motion stability coefficients are accurate and reliable.

[0090] Specifically, this motion stability analyzer is trained on a sample dataset constructed from a large amount of historical robot motion control data and a corresponding set of sample motion stability coefficient annotations. The sample dataset includes a set of sample motion amplitudes, a set of sample motion velocities, and a set of sample road surface terrain information.

[0091] Among them, the sample movement amplitude set is extracted from the robot's historical movement records, covering its typical stride range to ensure that the data fits the actual movement range; the sample movement speed set is collected based on the mapping relationship between movement amplitude and movement speed to ensure the physical adaptability of the two; and the sample road surface terrain information set covers the average height, height fluctuation and other features of different road surfaces, which are derived from the processing results of historical road surface point cloud data.

[0092] In addition, the sample motion stability coefficient label set is generated by statistically analyzing the instability ratio of the robot under different combinations of "motion amplitude-motion speed-road terrain information". The motion stability coefficient of each sample is calculated and labeled according to "1-instability ratio" to ensure that the label is consistent with the actual stability situation.

[0093] Furthermore, the sample dataset and the sample annotation set are divided into training set, validation set and test set in a ratio of 8:1:1. When dividing, it is necessary to ensure that the samples of various scenarios are evenly distributed in the three sets, covering the common movement and road surface combination patterns of robots, so as to avoid data bias affecting the model training effect.

[0094] During the model training phase, the training set data is input into the gradient boosting decision tree framework. First, the input parameters such as movement amplitude, movement speed, and road terrain features are normalized, and the min-max normalization method is used to map all parameter values ​​to the [0, 1] interval to eliminate interference from differences in units.

[0095] Furthermore, decision trees are constructed through multiple iterations, with a total of 120 decision trees. The maximum depth of each decision tree is set to 8, and the minimum number of samples in the leaf nodes is set to 20. Each round aims to correct the residuals of the preceding model. The optimal feature splitting node is selected through the gradient descent direction to gradually optimize the prediction accuracy. At the same time, the above tree structure parameters are used to control and prevent overfitting.

[0096] Meanwhile, during training, mean squared error (MSE) is used as the loss function to calculate the average squared difference between the model's predicted moving stability coefficient and the sample's true coefficient. The network parameters are updated iteratively through the Adam optimizer. The initial learning rate is set to 0.001, and the learning rate is decayed to 0.8 times the previous rate every 20 iterations, forming a learning rate decay strategy.

[0097] Furthermore, the model performance is dynamically monitored using the validation set. The mean absolute error (MAE) of the validation set is calculated every 5 iterations. If the improvement in the validation set error is less than 0.0001 for 8 consecutive iterations, the model is considered to have converged and training is stopped.

[0098] During the model testing phase, the test set is input into the trained motion stability analyzer. If the model's predicted values ​​for motion stability coefficients deviate little from the actual values ​​and maintain high prediction accuracy in different scenarios, the motion stability analyzer is deemed qualified.

[0099] Ultimately, this motion stability analyzer can quickly combine robot motion parameters with road terrain information to output quantified motion stability coefficients, providing a reliable basis for subsequent configuration of motion error range and selection of optimal motion position, thus helping the robot achieve terrain-adaptive navigation.

[0100] Furthermore, the obtained first movement amplitude, first movement speed, and first road surface terrain information are input into the constructed motion stability analyzer. After processing and integration by the motion stability analyzer, the first motion stability coefficient is accurately calculated and output, providing a core quantitative basis for configuring the movement error range and dividing the first movement position range based on this coefficient.

[0101] For example, if the straight-line distance between the robot's current position and the first moving position is 0.5 meters (i.e., the first moving amplitude is 0.5 meters), the first moving speed matched according to the amplitude-velocity mapping relationship is 0.25 meters / second, and the first road surface terrain information covered by the first moving position is "average height 0.02 meters, maximum height 0.05 meters, minimum height -0.01 meters, and height change rate 0.08". After inputting these three sets of information into the motion stability analyzer, the analyzer performs parameter normalization and gradient boosting decision tree model calculations, and finally outputs a first motion stability coefficient of 0.93. Subsequently, an appropriate motion error range can be configured based on this coefficient, and the range of the first moving position can be divided accordingly to ensure the safety and rationality of the range division.

[0102] S130: Based on the first movement stability coefficient, configure the movement error range, divide the first movement position into ranges to obtain the first movement position range, and analyze and calculate the first fusion stability coefficient based on the road surface terrain information within the first range of the first movement position.

[0103] In this embodiment of the application, in order to avoid the problem that the robot's footsteps sway and deviate from the first moving position due to unstable movement, it is necessary to dynamically configure the range of movement error through the first movement stability coefficient, thereby expanding to form the range of the first moving position, and calculate the fusion stability coefficient based on the terrain information within the range, so as to solve the problem that the traditional fixed error range does not combine with the actual stable state and is prone to risk misjudgment or resource waste.

[0104] Specifically, the average movement stability coefficient and average movement error range accumulated during the robot's historical movement process are first obtained.

[0105] Among them, the average movement stability coefficient is the average value calculated based on the stability coefficients of past effective movement scenarios, and the average movement error range is the benchmark range determined after multiple verifications based on the maximum deviation between the actual landing position and the target position in past movements.

[0106] Furthermore, based on the ratio of the first motion stability coefficient to the average motion stability coefficient, the average motion error range is dynamically adjusted and calculated to obtain a motion error range that adapts to the current scenario.

[0107] Among them, the mobility stability coefficient is positively correlated with the mobility error range. That is, the higher the mobility stability coefficient, the lower the probability of the robot's feet wobbling or deviating from the target position during movement, and the mobility error range can be appropriately widened; the lower the mobility stability coefficient, the higher the risk of mobility instability, and the mobility error range needs to be narrowed to avoid covering high-risk areas.

[0108] Furthermore, the calculated range of movement error is used to divide the first movement position into range compensation areas. That is, taking the coordinates (X0, Y0) of the first movement position on the road surface topographic map as the center, the range of movement error is used as the radius and uniformly expanded in the positive and negative directions of the X and Y axes to form a circular range of the first movement position, so as to ensure full coverage of potential error landing points.

[0109] Furthermore, multiple first error movement positions are randomly selected within the first movement position range. The number of selections must take into account both computational efficiency and result representativeness, and each position must be evenly distributed within the range to avoid stability assessment bias caused by position concentration.

[0110] Meanwhile, for each first error movement position, the corresponding road topography information is extracted from the road topography map, and combined with the determined first movement amplitude and first movement speed, it is input into the movement stability analyzer, and the analyzer outputs the movement stability coefficient corresponding to each first error movement position.

[0111] Finally, the arithmetic mean of the movement stability coefficients for multiple first error movement positions is calculated, and this mean is the first fused stability coefficient. This coefficient comprehensively reflects the overall stability level of the first movement position range. Compared with the first movement stability coefficient of a single position, it can better avoid the influence of local error positions, and the evaluation results are more reliable.

[0112] Step S130 in the method provided in this application embodiment includes:

[0113] Obtain the average movement stability coefficient of the robot and the average movement error range of the robot;

[0114] The average movement error range is adjusted and calculated based on the ratio of the first movement stability coefficient to the average movement stability coefficient to obtain the movement error range.

[0115] The first moving position is divided into range compensation sections using the moving error range to obtain the range of the first moving position;

[0116] Within the range of the first moving position, multiple first error moving positions are randomly selected. Based on the road surface terrain information of the multiple first error moving positions, combined with the first moving amplitude and the first moving speed, the moving stability coefficient is analyzed and the mean is calculated to obtain the first fusion stability coefficient.

[0117] In this embodiment of the application, in order to solve the problem that the actual landing position of the robot deviates from the first moving position due to instability during the movement, it is necessary to dynamically configure the movement error range and conduct a stability assessment of the range. This expands the stability judgment of a single position to a comprehensive assessment of the potential error landing point area, thereby avoiding navigation decision safety hazards caused by the omission of local risks.

[0118] Specifically, the average motion stability coefficient and average motion error range of the robot are first obtained. Both of these benchmark data are derived from the robot's historical motion control database and need to cover diverse scenarios to ensure representativeness.

[0119] The extraction of the average motion stability coefficient requires filtering nearly 300 valid motion records (excluding invalid data such as collisions and sudden external force interference). Each record contains complete information on "motion amplitude - motion speed - road surface and terrain information - motion stability coefficient", which is then calculated using an arithmetic mean. For example, if the sum of the motion stability coefficients in the 300 valid motion records is 258, then the average motion stability coefficient = 258 / 300 = 0.86.

[0120] Furthermore, the average movement error range is determined by statistically analyzing the deviation between the actual landing position and the target movement position in historical movements. Specifically, for the aforementioned 300 valid movement records, the actual coordinates (X, Y, X) of the robot's torso center point on the road surface topography map are calculated for each landing. 实 Y 实 ) and the target's first moving position coordinates (X 目 Y 目 The Euclidean distance (i.e., deviation value) of all deviation values ​​is taken as the average moving error range.

[0121] For example, if statistics show that 95% of the deviation values ​​are ≤4cm, then the average movement error range is set to 4cm. This value covers most common movement error scenarios and avoids interference from extreme abnormal deviations on the benchmark data.

[0122] Furthermore, based on the ratio of the first motion stability coefficient to the average motion stability coefficient, the average motion error range is adjusted and calculated to obtain a motion error range suitable for the current scenario.

[0123] Among them, the first mobility stability coefficient directly reflects the stability level of the current movement position, and the two are positively correlated. That is, when the first mobility stability coefficient is higher than the average level, the robot is more stable in moving at that position, the probability of footstep error and wobbling is lower, and the range of mobility error can be appropriately widened to cover more potential landing points.

[0124] Conversely, when the first motion stability coefficient is below average, the risk of instability in the robot's movement at that position increases, and the range of motion error needs to be reduced to control the risk.

[0125] For example, if the first motion stability coefficient is 0.94 (higher than the average of 0.86), then the adjustment ratio = 0.94 / 0.86≈1.09, and the motion error range = 4cm×1.09≈4.36cm, taking an approximate value of 4.4cm; if the first motion stability coefficient is 0.75 (lower than the average of 0.86), the adjustment ratio = 0.75 / 0.86≈0.87, and the motion error range = 4cm×0.87≈3.48cm, taking an approximate value of 3.5cm. Through this dynamic adjustment method, the error range is precisely matched with the current motion stability state.

[0126] Based on this, the calculated range of movement error is used to divide the first movement position into range compensation sections to obtain the range of the first movement position.

[0127] Specifically, when dividing the area, the coordinates (X0, Y0) of the first moving position on the road surface topography map are taken as the center, and the range of moving error is used as the radius to expand evenly in the positive and negative directions of the X-axis (forward direction) and Y-axis (horizontal lateral direction) to form a circular range of the first moving position.

[0128] The first moving position range fully covers all areas where the robot may actually land due to unstable movement. For example, when the movement error range is 4.4cm, the first moving position range is a circular area with a radius of 4.4cm centered at (X0, Y0) to ensure that no potential error landing points are missed. At the same time, the range boundary is clearly defined to avoid excessive expansion of the evaluation area, which would lead to a waste of computing resources.

[0129] Furthermore, multiple first error movement positions are randomly selected within the first movement position range. The number of selections should take into account both evaluation accuracy and computational efficiency, and is usually set to 6-8. They should be selected according to the principle of uniform distribution to avoid the evaluation results being biased towards local terrain due to the concentration of positions. For example, 6 positions can be selected within a circular range in the directions of 0°, 60°, 120°, 180°, 240°, and 300°.

[0130] Meanwhile, for each first error movement position, the road topography information within its coverage area is extracted from the road topography map, and the three types of parameters, namely the first movement amplitude and the first movement speed, are input into the trained motion stability analyzer.

[0131] Furthermore, the parameters of each first error movement position are processed by the moving stability analyzer, and the moving stability coefficient corresponding to each first error movement position is output by calculating the parameters through a pre-trained gradient boosting decision tree model.

[0132] Furthermore, the arithmetic mean of the motion stability coefficients corresponding to the multiple first error movement positions is calculated, and the resulting average value is the first fused stability coefficient, which reflects the overall motion stability level of the first movement position range. Compared to a single first motion stability coefficient, this first fused stability coefficient is more resistant to the influence of local terrain fluctuations or random errors, making the assessment results more reliable.

[0133] For example, if six first error movement positions are randomly selected within the first movement position range, and the movement stability analyzer analyzes them, the resulting movement stability coefficients are 0.91, 0.89, 0.93, 0.90, 0.92, and 0.88, respectively. The arithmetic mean of these six movement stability coefficients is calculated as (0.91+0.89+0.93+0.90+0.92+0.88) / 6=0.905. This 0.905 is the first fusion stability coefficient, which comprehensively reflects the stability of different potential landing points within the first movement position range, effectively avoiding the limitations that may exist in single-position evaluation.

[0134] S140: Based on the first moving speed, the first fusion stability coefficient, calculate the first navigation fitness, and perform moving position optimization to obtain the optimal moving position.

[0135] In this embodiment of the application, in order to achieve the dual goals of stability and motion continuity of the robot in complex terrain, it is necessary to calculate the navigation fitness by fusing range stability level and speed adaptability, and select the optimal movement position through multiple rounds of iterative optimization, so as to solve the problem that traditional navigation only focuses on single stability and is prone to motion stuttering or instability.

[0136] Specifically, the similarity between the first moving speed and the first moving speed is first calculated to obtain the first speed stability coefficient. Here, the moving speed refers to the robot's average moving speed over the past time window, and the first moving speed is the target moving speed that matches the first moving amplitude.

[0137] Furthermore, based on the obtained first fusion stability coefficient and first velocity stability coefficient, the first navigation fitness is calculated.

[0138] Since the first fusion stability coefficient reflects the overall stability level of the first movement position range and the first velocity stability coefficient reflects the continuity of motion state switching, both of which are key indicators for navigation decision-making, a weighted summation formula is required to calculate the first navigation fitness in order to avoid decision-making bias caused by evaluation of a single indicator.

[0139] Based on this, we continue to randomly select moving locations within the road surface topography map, repeatedly calculate the moving stability coefficient, fusion stability coefficient, velocity stability coefficient, and navigation adaptability of that location, and perform iterative optimization.

[0140] Finally, after convergence, the movement position with the highest navigation fitness is obtained, which is the optimal movement position. The robot will use this optimal movement position as the target for the next navigation step, thereby achieving adaptive navigation in complex terrain.

[0141] This step quantifies the two objectives of stability and safety and motion coherence into navigation fitness, and selects the optimal movement position through iterative optimization. It effectively balances the safety and smoothness of the robot's terrain-adaptive navigation, significantly reduces the risk of instability or jamming caused by improper position selection, and provides reliable decision support for the robot's continuous movement in complex terrain.

[0142] Step S140 in the method provided in this application embodiment includes:

[0143] Calculate the similarity between the first moving speed and the moving speed to obtain the first speed stability coefficient;

[0144] The first navigation fitness is calculated based on the first fusion stability coefficient and the first velocity stability coefficient.

[0145] Continue to randomly select movement positions within the road surface topography map, calculate navigation fitness, perform iterative optimization, and obtain the movement position with the highest navigation fitness after convergence. Obtain the optimal movement position and perform navigation.

[0146] In this embodiment of the application, in order to solve the problems that focusing only on the level of positional stability during the terrain-adaptive navigation process of the robot can easily lead to stuttering during the transition of motion state, and pursuing only the smoothness of speed can easily ignore the terrain risks, it is necessary to evaluate the navigation adaptability by integrating stability and speed adaptability, and to select the optimal movement position through multiple rounds of iteration, so as to achieve the dual goals of safety and stability and smooth movement, and provide a scientific basis for robot movement in complex terrain.

[0147] First, the similarity between the first moving speed and the moving speed is calculated to obtain the first speed stability coefficient. Here, the moving speed is the robot's average moving speed over a past time window, extracted from a historical motion control database.

[0148] Specifically, the robot's movement speed is calculated by first obtaining the distance traveled within the past time window and then combining this distance with the length of the time window. For example, if the distance traveled in the past 15 seconds is 3.75 meters, then the movement speed = 3.75 meters / 15 seconds = 0.25 meters per second.

[0149] In addition, the first moving speed is the target moving speed that is adapted to the first moving amplitude, and is determined based on a preset moving amplitude-moving speed mapping relationship. For example, if the first moving amplitude is 0.55 meters, then the matched first moving speed is 0.27 meters per second.

[0150] Meanwhile, to quantify the compatibility between the first movement speed and the current movement speed, the similarity calculation uses the formula "1 - |first movement speed - current movement speed| / current movement speed". This formula calculates the absolute deviation between the two by "|first movement speed - current movement speed|", then divides it by the current movement speed to standardize the deviation, and finally subtracts the standardized deviation value from 1, so that the calculation result is always in the range of 0-1, which can intuitively reflect the compatibility level between the two.

[0151] Furthermore, the similarity value calculated using this formula is the first velocity stability coefficient, which will directly serve as a key quantitative indicator for subsequent evaluation of the continuity of the robot's motion state.

[0152] Specifically, the closer the first moving speed is to the moving speed, the smaller the absolute deviation between the two, and the closer the similarity is to 1, the higher the corresponding first speed stability coefficient. This means that when the robot switches from the current motion state to the target motion state, the adjustment range of joint driving force and gait cycle is smaller, and it is less likely to cause the torso to tilt or the feet to slip due to sudden speed changes.

[0153] Conversely, the greater the absolute deviation between the first moving speed and the moving speed, the lower the stability coefficient of the first speed, and the higher the risk of switching motion states. At this time, the robot joints need to adjust the driving force output significantly in a short period of time, and the gait cycle also needs to be shortened or lengthened accordingly, thereby increasing the probability of motion instability.

[0154] For example, if the robot's movement speed within the past time window is 0.25 m / s, and the first movement speed corresponding to a certain first movement position is 0.26 m / s, the similarity is calculated using the similarity formula as 1 - |0.26 - 0.25| / 0.25 = 0.96. This means that the first speed stability coefficient is 0.96, indicating excellent speed adaptability. The robot's joint adjustment range is small when switching movement states, and it can move smoothly.

[0155] Furthermore, based on the obtained first fusion stability coefficient and first velocity stability coefficient, the first navigation fitness is calculated by weighted summation to ensure that the navigation fitness can accurately match the dual goals of safety and smoothness. The specific calculation formula can be expressed as "first navigation fitness = first fusion stability coefficient × first fusion stability coefficient weight + first velocity stability coefficient × first velocity stability coefficient weight".

[0156] The weight allocation needs to be determined based on the priority of the robot's application scenario. In complex terrains where safety is the priority, such as gravel roads and potholes, the weight of the first fusion stability coefficient is set to 0.6 and the weight of the first speed stability coefficient is set to 0.4 to prioritize landing safety. In flat terrains where efficiency is the priority, such as indoor tiled floors, the weights of the two are adjusted to 0.5:0.5 to balance safety and smoothness.

[0157] For example, in a bumpy terrain (with the weight of the first fusion stability coefficient set to 0.6 and the weight of the first velocity stability coefficient set to 0.4), if the first fusion stability coefficient is 0.91 and the first velocity stability coefficient is 0.96, then the first navigation adaptability = 0.91 × 0.6 + 0.96 × 0.4 = 0.93, which means that the moving position has both high safety and high smoothness, and excellent adaptability.

[0158] Furthermore, the movement position is randomly selected within the road surface topography map, and the complete process of "calculating the first movement stability coefficient - calculating the first fusion stability coefficient - calculating the first velocity stability coefficient - calculating the navigation fitness" is repeated for iterative optimization.

[0159] The randomly selected moving locations must cover the effective area within the road topographic map, excluding extreme protrusions or depressions with a height difference greater than 0.15 meters. The number of these locations should be controlled between 12 and 15 to ensure sufficient coverage of potential landing points while avoiding wasting computational resources due to an excessive number of locations.

[0160] Meanwhile, the randomly selected movement positions should be evenly distributed within the robot's forward movement range, such as a rectangular area of ​​1-2 meters in front and 0.5 meters to the left and right, to prevent concentration in local terrain from causing evaluation deviations.

[0161] Furthermore, a convergence condition is set during the iteration process. That is, if the difference between the newly calculated highest navigation fitness and the highest value of the previous round is less than 0.01 in four consecutive iterations, and the movement position corresponding to the highest navigation fitness remains unchanged, the iteration is considered to have converged, and the random selection of movement position is stopped.

[0162] Finally, after iterative convergence, the movement position with the highest navigation fitness is obtained, which is the optimal movement position. The robot will use this movement position as the target for the next navigation step.

[0163] When performing navigation, the robot first determines the final movement range based on the distance between the optimal moving position and the current position, and then matches the final movement speed with the range. Finally, it adjusts the joint motion parameters based on the terrain information of the optimal position on the road surface map to ensure that the steps conform to the terrain undulations during movement, while maintaining smooth speed switching, so as to achieve stable and smooth navigation in complex terrain.

[0164] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects:

[0165] This application proposes a terrain-adaptive navigation method for robots. First, the robot's movement speed within a past time window is acquired. Point cloud data of the road surface in a preset area ahead is collected, and a road surface topography map containing height information is constructed. Next, a first movement position is randomly selected within the topography map, and the corresponding first movement amplitude and first movement speed are calculated. Road surface topography information at this position is extracted and input into a trained motion stability analyzer to obtain a first motion stability coefficient. Then, based on the ratio of the first motion stability coefficient to the average motion stability coefficient, the movement error range is dynamically adjusted, and the range of the first movement position is divided. Multiple error positions within this range are selected, and the average stability coefficient is calculated to obtain a first fused stability coefficient. Finally, the similarity between the first movement speed and the first movement speed is calculated to obtain a first velocity stability coefficient. The two types of stability coefficients are fused to obtain navigation fitness. The position with the highest navigation fitness is iteratively selected as the optimal movement position to guide the robot's navigation.

[0166] The method provided in this application, through the technical solution of "constructing a road topographic map - first motion stability assessment - fusion stability coefficient calculation - velocity stability analysis - iterative optimization of multiple movement positions", solves the problems of poor terrain adaptation, easy movement instability, and motion stuttering caused by fixed error range, single position assessment, and neglect of velocity continuity in traditional robot navigation. It realizes accurate perception and safe and smooth navigation of robots in complex terrain, and takes into account motion continuity while ensuring movement stability, providing reliable support for improving the terrain adaptive navigation performance of robots.

[0167] Example 2, as shown in the appendix Figure 2 As shown, based on the inventive concept of a terrain-adaptive navigation method for a robot provided in Embodiment 1, this application also provides a terrain-adaptive navigation system for a robot, specifically including:

[0168] The road surface topography acquisition module 01 is used to acquire the robot's moving speed within the past time window, collect the road surface data of the current movement, and identify and obtain the road surface topography map.

[0169] The motion stability coefficient analysis module 02 is used to randomly select a first moving position within the road surface topography map, obtain the corresponding first moving speed and first moving amplitude, and analyze and obtain a first motion stability coefficient based on the first moving amplitude, the first moving speed, and the first road surface topography information of the first moving position.

[0170] The fusion stability coefficient calculation module 03 is used to configure the movement error range according to the first movement stability coefficient, divide the first movement position into ranges to obtain the first movement position range, and analyze and calculate the first fusion stability coefficient according to the first range of road surface terrain information within the first movement position range.

[0171] The optimal moving position selection module 04 is used to calculate the first navigation fitness based on the first moving speed, the first fusion stability coefficient, and to perform moving position optimization selection to obtain the optimal moving position.

[0172] In one embodiment, the road surface topography map acquisition module 01 is further configured to:

[0173] The robot's movement speed is calculated by obtaining the distance traveled within the past time window and combining it with the length of the time window.

[0174] Collect point cloud data of the road surface within a preset area in front of the robot;

[0175] A road surface topographic map is constructed based on the road surface point cloud data.

[0176] Furthermore, the road surface topography map acquisition module 01 also includes:

[0177] Construct a regional coordinate system for the preset area;

[0178] Based on the road surface point cloud data, a three-dimensional model is constructed within a preset area, and coordinates are matched with the area coordinate system to obtain a road surface topographic map, wherein the road surface topographic map includes the height information of each area coordinate.

[0179] In one embodiment, the moving stability coefficient analysis module 02 is further used for:

[0180] A first moving position is randomly selected within the road surface topography map. Based on the first moving position and the robot's current position, a first moving amplitude is calculated. A first moving speed is obtained by classifying the first moving amplitude. The robot's moving amplitude and moving speed have a mapping relationship.

[0181] First road terrain information that divides the coverage area of ​​the first moving position within the road terrain map;

[0182] The first movement amplitude, the first movement speed, and the first road surface terrain information are input into the movement stability analyzer, and the first movement stability coefficient is output.

[0183] Furthermore, the moving stability coefficient analysis module 02 also includes:

[0184] A mobile stability analyzer is built based on machine learning.

[0185] Based on historical data of robot movement control, a set of sample movement amplitude, a set of sample movement speed, and a set of sample road surface terrain information are collected. The proportion of robot instability under different movement conditions is collected, the sample movement stability coefficient is calculated, and the set of sample movement stability coefficients is obtained by labeling.

[0186] The motion stability analyzer is trained and optimized in a supervised manner using the set of sample motion amplitude, sample motion speed, sample road surface terrain information, and sample motion stability coefficients until the training is verified to have converged.

[0187] In one embodiment, the fusion stability coefficient calculation module 03 is further used for:

[0188] Obtain the average movement stability coefficient of the robot and the average movement error range of the robot;

[0189] The average movement error range is adjusted and calculated based on the ratio of the first movement stability coefficient to the average movement stability coefficient to obtain the movement error range.

[0190] The first moving position is divided into range compensation sections using the moving error range to obtain the range of the first moving position;

[0191] Within the range of the first moving position, multiple first error moving positions are randomly selected. Based on the road surface terrain information of the multiple first error moving positions, combined with the first moving amplitude and the first moving speed, the moving stability coefficient is analyzed and the mean is calculated to obtain the first fusion stability coefficient.

[0192] In one embodiment, the optimal moving position selection module 04 is further configured to:

[0193] Calculate the similarity between the first moving speed and the moving speed to obtain the first speed stability coefficient;

[0194] The first navigation fitness is calculated based on the first fusion stability coefficient and the first velocity stability coefficient.

[0195] Continue to randomly select movement positions within the road surface topography map, calculate navigation fitness, perform iterative optimization, and obtain the movement position with the highest navigation fitness after convergence. Obtain the optimal movement position and perform navigation.

[0196] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0197] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0198] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A terrain-adaptive navigation method for a robot, characterized in that, The method includes: The robot's movement speed within a past time window is obtained, and road surface data of the current movement is collected to identify and obtain a road surface topography map; A first moving position is randomly selected within the road surface topography map, and the corresponding first moving speed and first moving amplitude are obtained. Based on the first moving amplitude, the first moving speed, and the first road surface topography information of the first moving position, the first moving stability coefficient is analyzed and obtained. Based on the first movement stability coefficient, a movement error range is configured, the first movement position is divided into ranges to obtain the first movement position range, and the first fusion stability coefficient is obtained by analyzing and calculating the road surface terrain information within the first range of the first movement position. Based on the first moving speed, the first fusion stability coefficient, the first navigation fitness is calculated, and the optimal moving position is obtained by optimizing the moving position selection.

2. The terrain-adaptive navigation method for a robot according to claim 1, characterized in that, The robot's movement speed within a past time window is obtained, and road surface data of the current movement is collected to identify and obtain a road surface topography map, including: The robot's movement speed is calculated by obtaining the distance traveled within the past time window and combining it with the length of the time window. Collect point cloud data of the road surface within a preset area in front of the robot; A road surface topographic map is constructed based on the road surface point cloud data.

3. The terrain-adaptive navigation method for a robot according to claim 2, characterized in that, Based on the road surface point cloud data, a road surface topographic map is constructed, including: Construct a regional coordinate system for the preset area; Based on the road surface point cloud data, a three-dimensional model is constructed within a preset area, and coordinates are matched with the area coordinate system to obtain a road surface topographic map, wherein the road surface topographic map includes the height information of each area coordinate.

4. The terrain-adaptive navigation method for a robot according to claim 3, characterized in that, Based on the road surface point cloud data, a first movement position is randomly selected within the road surface topography map, and the corresponding first movement speed and first movement amplitude are obtained. Based on the first movement amplitude, the first movement speed, and the first road surface topography information of the first movement position, a first movement stability coefficient is analyzed and obtained, including: A first moving position is randomly selected within the road surface topography map. Based on the first moving position and the robot's current position, a first moving amplitude is calculated. A first moving speed is obtained by classifying the first moving amplitude. The robot's moving amplitude and moving speed have a mapping relationship. First road terrain information that divides the coverage area of ​​the first moving position within the road terrain map; The first movement amplitude, the first movement speed, and the first road surface terrain information are input into the movement stability analyzer, and the first movement stability coefficient is output.

5. The terrain-adaptive navigation method for a robot according to claim 4, characterized in that, The process of building and training the motion stability analyzer includes the following steps: A mobile stability analyzer is built based on machine learning. Based on historical data of robot movement control, a set of sample movement amplitude, a set of sample movement speed, and a set of sample road surface terrain information are collected. The proportion of robot instability under different movement conditions is collected, the sample movement stability coefficient is calculated, and the set of sample movement stability coefficients is obtained by labeling. The motion stability analyzer is trained and optimized in a supervised manner using the set of sample motion amplitude, sample motion speed, sample road surface terrain information, and sample motion stability coefficients until the training is verified to have converged.

6. The terrain-adaptive navigation method for a robot according to claim 1, characterized in that, Based on the mobility stability coefficient, a mobility error range is configured, and the first mobility position is divided into ranges to obtain a first mobility position range. Based on the road surface terrain information within the first range of the first mobility position, a first fusion stability coefficient is analyzed and calculated, including: Obtain the average movement stability coefficient of the robot and the average movement error range of the robot; The average movement error range is adjusted and calculated based on the ratio of the first movement stability coefficient to the average movement stability coefficient to obtain the movement error range. The first moving position is divided into range compensation sections using the moving error range to obtain the range of the first moving position; Within the range of the first moving position, multiple first error moving positions are randomly selected. Based on the road surface terrain information of the multiple first error moving positions, combined with the first moving amplitude and the first moving speed, the moving stability coefficient is analyzed and the mean is calculated to obtain the first fusion stability coefficient.

7. The terrain-adaptive navigation method for a robot according to claim 1, characterized in that, Based on the first moving speed, the first fusion stability coefficient, a first navigation fitness is calculated, and the optimal moving position is obtained through moving position optimization, including: Calculate the similarity between the first moving speed and the moving speed to obtain the first speed stability coefficient; The first navigation fitness is calculated based on the first fusion stability coefficient and the first velocity stability coefficient. Continue to randomly select movement positions within the road surface topography map, calculate navigation fitness, perform iterative optimization, and obtain the movement position with the highest navigation fitness after convergence. Obtain the optimal movement position and perform navigation.

8. A terrain-adaptive navigation system for a robot, characterized in that, The system is used to execute the terrain-adaptive navigation method for a robot according to any one of claims 1-7, the system comprising: The road surface topography acquisition module is used to acquire the robot's moving speed within a past time window, collect road surface data of the current movement, and identify and obtain a road surface topography map. The moving stability coefficient analysis module is used to randomly select a first moving position within the road surface topography map, obtain the corresponding first moving speed and first moving amplitude, and analyze and obtain a first moving stability coefficient based on the first moving amplitude, the first moving speed, and the first road surface topography information of the first moving position. The fusion stability coefficient calculation module is used to configure the movement error range according to the first movement stability coefficient, divide the first movement position into ranges to obtain the first movement position range, and analyze and calculate the first fusion stability coefficient based on the first range of road surface terrain information within the first movement position range. The optimal moving position selection module is used to calculate the first navigation fitness based on the first moving speed, the first fusion stability coefficient, and to perform moving position optimization to obtain the optimal moving position.

Images