Walking robot path planning method and system

Abstract

The invention relates to the technical field of robot path planning, and provides a walking robot path planning method and system, and the method comprises the steps: obtaining the environment parameters of a target region, constructing a multi-dimensional energy consumption map, carrying out the marching resistance analysis, obtaining a risk grading label, carrying out the path screening of the target region according to the risk grading label, and obtaining the path of the target region. After the path candidate set is obtained, map generation is carried out, and a path weight map is obtained; and acquiring electric energy stability distribution of the walking robot, performing path correction on the path weight map according to the electric energy stability distribution to obtain a path fitting map, and analyzing an optimal path output result according to the path fitting map and the path candidate set. Through the technical means of environment parameter acquisition, advancing resistance analysis, electric energy stability correction and the like, efficient optimization of walking robot path planning is successfully achieved, and the problem that in a complex environment, the walking robot path planning is prone to being affected by environment uncertainty, and consequently stability and reliability of walking robot path planning are poor is solved.

Description

Technical Field

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

[0002] With the rapid development of intelligent technology, robotics has been widely applied in various industries, such as industrial manufacturing, healthcare, and logistics. As a key branch of robotics, walking robots, with their autonomous mobility and path planning capabilities in complex environments, have become a research focus. Effective path planning, ensuring robots can complete tasks efficiently and safely in irregular and dynamic environments, is one of the major challenges currently facing the robotics field.

[0003] Among related technologies, path planning methods for walking robots often employ techniques based on environmental modeling and algorithm optimization. Common approaches include map-based path planning and dynamic path planning based on real-time perception. Map-based path planning typically relies on a pre-established map of the environment. The robot uses path planning algorithms (such as the A* algorithm and the Dijkstra algorithm) to calculate the shortest or optimal path from the starting point to the destination.

[0004] Regarding the above technical solution, although the path planning method based on environmental modeling and sensor input can achieve path calculation and dynamic adjustment, in complex environments, especially in variable terrain and robot load conditions, the path planning efficiency of traditional methods is low and is easily affected by environmental uncertainties (such as power consumption, travel resistance, etc.), resulting in poor stability and reliability of path planning. Summary of the Invention

[0005] In order to improve the problem that traditional path planning methods are inefficient in complex environments and are easily affected by environmental uncertainties, resulting in poor stability and reliability of walking robot path planning, the present application provides a walking robot path planning method and system.

[0006] The present invention provides a walking robot path planning method, comprising: acquiring environmental parameters of a target area, performing spatial label binding on the environmental parameters to obtain an environmental multi-factor sequence, and constructing a multidimensional energy consumption map based on the environmental multi-factor sequence; performing a travel resistance analysis on the target area using the multidimensional energy consumption map to generate a resistance vector field, performing a power consumption risk assessment on the resistance vector field to obtain a risk grading label, and performing path screening on the target area according to the risk grading label to obtain a path candidate set; performing a tensor transformation and map generation on the path candidate set and a preset power threshold mapping matrix to obtain a path weight map; acquiring the current voltage and current current of the walking robot, and performing trend fitting on the current voltage and the current current to obtain an electric energy stability distribution, performing path correction on the path weight map according to the electric energy stability distribution to obtain a path adaptability result; performing section disturbance injection and local multinomial fitting on the path adaptability result to generate a path fitting map, and analyzing an optimal path output result based on the path fitting map and the path candidate set.

[0007] As a preferred solution, the steps of obtaining environmental parameters of the target area, spatially tagging the environmental parameters to obtain an environmental multi-factor sequence, and constructing a multidimensional energy consumption map based on the environmental multi-factor sequence include: obtaining air pressure, humidity, temperature difference and surface material data of the target area to generate environmental parameters, spatially tagging the environmental parameters through geographic reference coordinates to obtain an environmental multi-factor sequence; wherein the geographic reference coordinates refer to current latitude, longitude and altitude data; performing sliding window processing on the environmental multi-factor sequence to extract response feature vectors and terrain switching marks, generating coupling nodes based on the response feature vectors, and constructing a disturbance amplitude map and a terrain continuity map through the coupling nodes and the terrain switching marks; extracting local stable area features based on the disturbance amplitude map, extracting surface structure change trends based on the terrain continuity map, fusing the local stable area features with the surface structure change trends to construct a response trend tensor and a material evolution sequence; performing frequency decomposition and multi-scale fusion on the response trend tensor to obtain a disturbance intensity map, and generating a multidimensional energy consumption map based on the disturbance intensity map and the material evolution sequence.

[0008] As a preferred solution, the step of using the multi-dimensional energy consumption map to analyze the travel resistance of the target area, generate a resistance vector field, perform power consumption risk assessment on the resistance vector field, obtain risk classification labels, and screen the paths of the target area according to the risk classification labels to obtain a path candidate set includes: performing directional gradient decomposition and regional density clustering on the multi-dimensional energy consumption map to obtain the boundaries of energy consumption abnormal areas, and using the boundaries of energy consumption abnormal areas to construct a resistance vector field; fusing the resistance vector field with historical energy consumption trajectory data to generate a path energy dissipation matrix and a power fluctuation map, extracting power critical points according to the path energy dissipation matrix, and identifying the path candidate set based on the power fluctuation map. The stable traveling area is distinguished; wherein the historical energy consumption trajectory data refers to historical energy consumption and path data; the power critical point and the stable traveling area are cross-analyzed, and a critical power spectrum and a safety margin distribution diagram are constructed based on the analysis results, and the risk index and energy consumption stability are calculated using the critical power spectrum and the safety margin distribution diagram; the risk index and the energy consumption stability are graded to generate a risk grading label, and a resistance balance area is screened out according to the risk grading label, and the resistance balance area is integrated with the preset path accessibility constraint to obtain a path scoring table and an obstacle avoidance level matrix; a path candidate set is extracted according to the path scoring table and the obstacle avoidance level matrix.

[0009] As a preferred solution, the step of performing tensor transformation and graph generation on the path candidate set and the preset power threshold mapping matrix to obtain a path weight graph includes: constructing a ternary path constraint vector according to the path length, slope, and surface material type of each path in the path candidate set, performing tensor transformation on the ternary path constraint vector and the preset power threshold mapping matrix to obtain a path energy consumption response value and a power distribution graph; performing principal component decomposition on the path energy consumption response value to extract the main characteristic axis of energy consumption, and generating a power offset factor and a characteristic compression sequence according to the main characteristic axis of energy consumption and the power distribution graph; and performing a principal component decomposition on the power offset value. The characteristic compression sequence is subjected to multi-scale interpolation to obtain a scale direction distribution sequence, and a direction accessibility matrix is constructed using the scale direction distribution sequence and the direction response vector. A dynamic adaptation scoring table is constructed based on the direction accessibility matrix and the ternary path constraint vector. The dynamic adaptation scoring table is fused with the energy consumption fluctuation map to obtain a path energy consumption distribution feature set and a load response vector group. A path weight map is constructed based on the path energy consumption distribution feature set and the load response vector group.

[0010] As a preferred solution, the step of performing multi-scale interpolation on the feature compression sequence to obtain a scale direction distribution sequence, using the scale direction distribution sequence and the direction response vector to construct a directional accessibility matrix, and constructing a dynamic adaptation score table based on the directional accessibility matrix and the ternary path constraint vector includes: dividing the feature compression sequence into different levels according to scale size, and interpolating each level using wavelet transform to obtain a scale direction distribution sequence at each scale; extracting the rate of change and fluctuation amplitude of the path direction according to the scale direction distribution sequence, and weightedly fusing the rate of change and the fluctuation amplitude with the direction response vector to obtain a directional accessibility matrix; combining the directional accessibility of the directional accessibility matrix with the path constraint condition of the ternary path constraint vector to obtain a dynamic adaptation score table.

[0011] As a preferred solution, the steps of obtaining the current voltage and current current of the walking robot, performing trend fitting on the current voltage and current current to obtain the power stability distribution, performing path correction on the path weight map according to the power stability distribution, and obtaining the path adaptability result include: using a sensor to obtain the current voltage and current current of the walking robot, calculating the power input and power output of the current voltage and current current, generating a current density vector and a power voltage drop value based on the power input and power output; performing short-term trend fitting analysis on the current density vector to obtain a voltage fluctuation factor and a load response delay, and performing path correction on the path weight map based on the power voltage. A multidimensional power stability matrix is constructed based on the drop value and the load response delay, and the multidimensional power stability matrix is integrated with the voltage fluctuation factor to generate an electric energy stability distribution; the path load entropy value and the power supply adaptation factor are calculated using the electric energy stability distribution and the path weight map, and a path constraint adjustment table and an alternative path index set are constructed based on the path load entropy value and the power supply adaptation factor; the priority of the path weight map is corrected based on the path constraint adjustment table to obtain an updated path weight map, the alternative path index set is screened for compatibility to obtain a screened index set, and the path adaptability result is extracted based on the updated path weight map and the screened index set.

[0012] As a preferred solution, the step of performing segment disturbance injection and local polynomial fitting on the path adaptability result to generate a path fitting map, and analyzing the optimal path output result based on the path fitting map and the path candidate set includes: performing segment disturbance injection on the weight distribution and historical path change trend in the path adaptability result to obtain a disturbance response sequence, performing local polynomial fitting on the disturbance response sequence, and extracting a segment fitting curvature map and a node offset scalar field; wherein the historical path change trend refers to the change pattern of the curvature change, slope fluctuation, and surface condition change data of the path when the walking robot moves on the path; performing segment disturbance injection on the weight distribution and historical path change trend in the path adaptability result to obtain a disturbance response sequence, performing local polynomial fitting on the disturbance response sequence, and extracting a segment fitting curvature map and a node offset scalar field; wherein the historical path change trend refers to the change pattern of the curvature change, slope fluctuation, and surface condition change data of the path when the walking robot moves on the path; performing local polynomial fitting on the segment fitting curvature map and the node offset scalar field. The field is segmented and compared to generate a path stability vector group and a direction fusion value, the path stability vector group is used to construct a dynamic correction path set, and the stability of the dynamic correction path set is screened by the direction fusion value to obtain a path fitting map; an energy consumption offset table and a travel efficiency distribution map are analyzed according to the path fitting map and the path candidate set, an efficiency loss factor table is constructed based on the energy consumption offset table, and a path hierarchical screening group is constructed based on the travel efficiency distribution map; a path scoring comparison matrix and a priority sequence index table are generated by the efficiency loss factor table and the path hierarchical screening group, and an optimal path output result is output according to the path scoring comparison matrix and the priority sequence index table.

[0013] The present application also provides a walking robot path planning system, comprising: an acquisition module for acquiring environmental parameters of a target area, performing spatial label binding on the environmental parameters to obtain an environmental multi-factor sequence, and constructing a multidimensional energy consumption map based on the environmental multi-factor sequence; an evaluation module for performing travel resistance analysis on the target area using the multidimensional energy consumption map, generating a resistance vector field, performing power consumption risk assessment on the resistance vector field to obtain a risk grading label, and performing path screening on the target area according to the risk grading label to obtain a path candidate set; a generation module for performing tensor transformation and map generation on the path candidate set and a preset power threshold mapping matrix to obtain a path weight map; a fitting module for acquiring the current voltage and current current of the walking robot, performing trend fitting on the current voltage and current current to obtain an electric energy stability distribution, performing path correction on the path weight map according to the electric energy stability distribution to obtain a path adaptability result; an analysis module for performing segment disturbance injection and local multinomial fitting on the path adaptability result to generate a path fitting map, and analyzing the optimal path output result based on the path fitting map and the path candidate set.

[0014] Compared with the existing technology, the present application has the following beneficial effects: high adaptability and strong stability. By using multi-dimensional energy consumption maps to analyze travel resistance and combining real-time power stability distribution to perform path correction, the problem of traditional path planning methods ignoring power consumption and path stability in complex environments is effectively avoided. The generation of path fitting maps and disturbance injection technology make path planning more flexible and adaptable, and can be dynamically adjusted according to actual power conditions and environmental changes, ensuring that the robot can always choose the optimal path in a changing environment. This improves the problem that traditional path planning methods are inefficient in complex environments and are easily affected by environmental uncertainties, resulting in poor stability and reliability of walking robot path planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] The structures, proportions, sizes, etc. depicted in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not intended to limit the conditions under which the present invention can be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportional relationships, or adjustments in size should still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and objectives that can be achieved by the present invention.

[0017] Figure 1 1 is a flow chart of a walking robot path planning method provided by an embodiment of the present invention; Figure 2 It is a schematic block diagram of the structure of a walking robot path planning system provided by an embodiment of the present invention.

[0018] Description of reference numerals: 10. Walking robot path planning system; 11. Acquisition module; 12. Evaluation module; 13. Generation module; 14. Fitting module; 15. Analysis module. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, combined, or partially merged, so the actual execution order may vary depending on the actual situation.

[0021] It should also be understood that the terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the present application. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] It should be further understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0023] The technical solution of the present invention will be further described below with reference to the accompanying drawings and through specific implementation methods.

[0024] Example 1: like Figure 1 As shown, the present application provides a walking robot path planning method, including steps S100 to S500.

[0025] Step S100: Acquire environmental parameters of the target area, perform spatial tag binding on the environmental parameters to obtain an environmental multi-factor sequence, and construct a multi-dimensional energy consumption map based on the environmental multi-factor sequence.

[0026] In this step, the walking robot uses sensors (such as temperature, humidity, and air pressure sensors) to acquire environmental parameters of the target area. These parameters include, but are not limited to, air pressure, humidity, temperature, and surface material. These environmental parameters are then integrated with a geographic information system (GIS) and spatially labeled to associate the environmental data with specific geographic locations (such as latitude, longitude, and altitude). This allows the robot to obtain accurate environmental parameters in real time and perform analysis at specific locations. Next, based on the acquired multi-factor sequence, a multidimensional energy consumption map is constructed. This map shows how energy consumption varies for each area under different environmental parameters. This energy consumption map takes into account the impact of different terrain, climate conditions, and surface materials on the robot's energy consumption. Specifically, during the environmental parameter acquisition process, sensors collect data at a specific frequency. After filtering and preprocessing, this data is combined with geographic reference information to form a multi-factor sequence. Using a data processing model, the influence weights of various environmental factors, such as air pressure, humidity, temperature, and surface material, are extracted. Based on these weights, the comprehensive energy consumption impact of each area is calculated, and an energy consumption map is then created.

[0027] For example, in a forest environment, changes in temperature and humidity affect a robot's battery consumption, while ground material (such as mud and sand) affects the robot's resistance to movement. The robot uses sensors to measure these environmental parameters in real time and calculates an energy consumption profile for the area based on a multi-factor sequence, thereby determining whether energy consumption in a particular area is high.

[0028] Step S200: Analyze the travel resistance of the target area using the multi-dimensional energy consumption map to generate a resistance vector field, perform power consumption risk assessment on the resistance vector field to obtain a risk classification label, and screen the paths of the target area based on the risk classification label to obtain a path candidate set.

[0029] In this step, the walking robot analyzes the travel resistance of the target area based on the multidimensional energy consumption map. The goal of this step is to identify the magnitude of travel resistance in different areas and generate a resistance vector field based on this. The travel resistance of each area is directly related to the energy consumption data in the multidimensional energy consumption map. The walking robot calculates the travel resistance at each location on the path, generating a resistance vector field with directionality and intensity, thereby revealing the travel difficulty of different areas. Next, based on this resistance vector field, the robot performs a power consumption risk assessment. This power consumption risk assessment process considers the resistance in different areas and the energy consumption required for travel, thereby assigning a risk classification label. Based on the risk classification label, the robot further selects suitable candidate paths. Specifically, the robot uses methods such as gradient descent or Gaussian process regression to analyze travel resistance and accurately identify areas with high travel resistance. This information is mapped onto a map of the target area to form a resistance vector field. Next, the robot uses a risk assessment model (such as a power consumption risk assessment model based on Monte Carlo simulation) to evaluate the paths, calculate the power consumption risk of each path, and ultimately generate a risk classification label.

[0030] For example, a desert area with high sand dunes and steep slopes requires more power for a walking robot to navigate, so it is marked as a high-risk area. The robot compares the power consumption risk classification labels of different paths to select the best path candidate set.

[0031] Step S300: Perform tensor transformation and graph generation on the path candidate set and the preset power threshold mapping matrix to obtain a path weight graph.

[0032] In this step, the set of candidate paths is combined with a preset power threshold mapping matrix and subjected to a tensor transformation. The purpose of the tensor transformation is to map each path in the candidate path set to a power threshold, forming a data structure containing multiple dimensions (such as power consumption, path length, slope, etc.). Through tensor operations, multidimensional data can be mapped into a unified graph. Ultimately, the generated path weight graph quantifies the comprehensive weight of each path, including factors such as power consumption, path resistance, and stability. Specifically, the tensor transformation is accomplished through matrix multiplication and high-dimensional tensor operations (such as Kronecker products). The preset power threshold mapping matrix represents the paths that the robot can traverse under different power conditions. By calculating each path, the comprehensive weight of the path is stored in the path weight graph.

[0033] For example, suppose a path has lower power consumption but a higher slope, so the comprehensive weight of this path will be higher in the path weight map; while another path has lower power consumption but a higher resistance, and will also be adjusted according to the weight value.

[0034] Step S400: Obtain the current voltage and current current of the walking robot, perform trend fitting on the current voltage and current current to obtain the power stability distribution, perform path correction on the path weight map according to the power stability distribution, and obtain the path adaptability result.

[0035] In this step, the current voltage and current of the walking robot are obtained and used for trend fitting. Trend fitting of voltage and current can use techniques such as linear regression and curve fitting to predict the trend of power consumption and further calculate the power stability distribution. The power stability distribution represents the stability of the path that the robot can support under the current power conditions. According to the power stability distribution, the priority of different paths in the path weight map is adjusted to obtain a path with stronger adaptability. Specifically, by collecting the voltage and current data of the robot battery in real time and performing trend analysis on this data, future power consumption is predicted. Based on this information, the model will adjust the path weight map so that the robot can choose a longer path when there is sufficient power, and a shorter or lower-energy-consuming path when there is insufficient power.

[0036] For example, if the robot's current voltage drops rapidly and the battery current is too large, the power stability will be poor. At this time, the system will give priority to paths that consume less power, thereby ensuring that the robot can complete the task under limited power conditions.

[0037] Step S500: perform segment disturbance injection and local multinomial fitting on the path adaptability result to generate a path fitting graph, and analyze the optimal path output result based on the path fitting graph and the path candidate set.

[0038] In this step, the path adaptability results are subjected to segment disturbance injection to simulate the impact of environmental changes on the path. Disturbance injection can simulate the impact of factors such as the appearance of obstacles and changes in road conditions, so that the stability and adaptability of the path can be more accurately evaluated. Afterwards, the perturbed path is corrected using local multinomial fitting technology to obtain a path fitting map. Through the path fitting map, the pros and cons of each path can be judged more accurately, and the optimal path can be output ultimately. Specifically, during the path fitting process, fitting techniques such as the least squares method are used to perform local fitting on the perturbed path data to ensure the accuracy and smoothness of the path. During the fitting process, comprehensive adjustments will be made based on factors such as path stability, energy consumption, and load to obtain a more reasonable path solution.

[0039] For example, if a path becomes more complex or infeasible after disturbance injection, it is corrected through local fitting technology, and the corrected path is compared with other candidate paths to select the optimal path.

[0040] In this embodiment, the environmental parameters of the target area are acquired and associated with specific geographic location data through spatial tag binding, thereby obtaining an environmental multi-factor sequence. Next, a multidimensional energy consumption map is constructed based on the environmental multi-factor sequence. Using this map, a travel resistance analysis is performed on the target area to generate a resistance vector field. This resistance vector field is then subjected to a power consumption risk assessment to obtain a risk classification label. Based on these risk classification labels, paths within the target area are screened to obtain a set of candidate paths. The candidate paths are then subjected to a tensor transformation and map generation with a preset power threshold mapping matrix to obtain a path weight map. Next, the current voltage and current of the walking robot are acquired and trend-fitted to obtain a power stability distribution. Based on the power stability distribution, the path weight map is further modified to obtain a path adaptability result. Finally, the path adaptability result is subjected to segment disturbance injection and local multinomial fitting to generate a path fitting map. The optimal path output result is analyzed based on the path candidate set. This approach takes into account the impact of environmental factors on the robot's path, such as resistance, power consumption risk, and energy consumption, in real time. It also dynamically corrects the path based on power stability, ensuring the robot's path adaptability under varying power states. By injecting segmented disturbances and performing local multinomial fitting, the accuracy of path planning can be further improved, ensuring a more stable and energy-efficient path optimization process. This significantly enhances the robot's navigation capabilities in complex environments and addresses the low efficiency and susceptibility of traditional path planning methods to environmental uncertainty, leading to poor stability and reliability in the robot's path planning.

[0041] Example 2: In step S100, the air pressure, humidity, temperature difference and surface material data of the target area are obtained to generate environmental parameters, and the environmental parameters are spatially labeled using geographic reference coordinates to obtain an environmental multi-factor sequence; wherein the geographic reference coordinates refer to the current latitude, longitude and altitude data.

[0042] The robot uses multiple sensors, including but not limited to temperature, humidity, air pressure, and surface material identification sensors, to collect real-time data on the target area's air pressure, humidity, temperature difference, and surface material to generate environmental parameters. The robot also uses the Global Positioning System (GPS) to collect the target area's georeferenced coordinates, including the current area's latitude, longitude, and altitude. A coordinate binding algorithm is used to correlate and fuse these environmental parameters with the georeferenced coordinates, generating a spatially labeled, multi-factor environmental sequence. This multi-factor environmental sequence reflects the surface environmental characteristics and spatial relationships of the target area across multiple dimensions. Specifically, to acquire air pressure, humidity, and temperature difference data, the robot uses sensors to measure the air pressure curve, humidity change rate, and temperature difference fluctuations within a specified range at preset intervals. For surface material data, a laser radar (LiDAR) or visual sensor classifies and identifies surface materials, such as soil, gravel, grass, or asphalt. The generated environmental parameters are then bound to the georeferenced coordinates using coordinate matching technology. Coordinate matching technology uses a spatial weighted algorithm to match and correct environmental data with location information and then store them in an environmental multi-factor sequence.

[0043] For example, in the target area, the robot measures humidity at 65% using a humidity sensor, air pressure at 985 hPa using a pressure sensor, and a temperature difference of 2°C using a temperature sensor. LiDAR identifies the surface material as gravel. This environmental data is bound to spatial coordinates, with latitude and longitude at 35.6586°N, longitude at 139.7454°E, and an altitude of 40 meters. Using a spatially weighted algorithm, this environmental parameter is stored as an item in a multi-factor sequence and labeled as a region with a combination of high humidity, low air pressure, and gravel.

[0044] Sliding window processing is performed on the environmental multi-factor sequence to extract the response eigenvector and terrain switching mark. Coupling nodes are generated according to the response eigenvector, and the disturbance amplitude map and terrain continuity map are constructed through the coupling nodes and terrain switching marks.

[0045] A sliding window technique is used to segment the multi-factor sequence, with each window size being a fixed spatial range (e.g., 100 m × 100 m). During this sliding window processing, response eigenvectors of environmental parameters are extracted, including the rate of change of air pressure, the frequency of humidity fluctuations, the temperature gradient, and the rate of surface material switching. Simultaneously, terrain transition markers are extracted to indicate the location of the transition (e.g., the transition point from sandy to grassland) and the intensity of the transition. Specifically, a weighted aggregation approach is used to generate coupling nodes based on the extracted response eigenvectors and terrain transition markers. Coupling nodes are the junctions where environmental factors respond to terrain changes, such as the combined effect of pressure and humidity changes and temperature gradients in sandy areas. Using the coupling nodes and terrain transition markers, a Gaussian kernel density estimation algorithm is used to construct a disturbance amplitude map, which reflects the intensity of local environmental disturbances. Furthermore, a terrain continuity map is constructed based on temporal continuity analysis, describing the surface transition patterns and temporal correlations of the target area.

[0046] For example, during its movement, the robot observed local changes in environmental parameters: a pressure drop rate of 0.3 hPa / s, a humidity fluctuation frequency of 0.15 Hz, and a temperature gradient of 2°C / km. Within the sliding window, the robot switched from gravel to grass with a switching rate of 0.8. Weighted aggregation generated coupling nodes, and a Gaussian kernel density estimation algorithm was used to generate a disturbance amplitude map, showing that the disturbance intensity was higher in gravel areas and lower in grass areas. A terrain continuity map was also constructed, identifying the transition pattern from gravel to grass, facilitating subsequent analysis of local stable regions.

[0047] The characteristics of local stable areas are extracted based on the disturbance amplitude map, and the surface structure change trend is extracted based on the terrain continuity map. The characteristics of local stable areas are integrated with the surface structure change trend to construct the response trend tensor and material evolution sequence.

[0048] By extracting features from the disturbance amplitude map, areas with low disturbance intensity are identified and used as features of local stable areas. Subsequently, the surface change trends are extracted based on the terrain continuity map, including slope change patterns, material transition patterns, and the temporal correlation of environmental parameters. This information forms the characteristics of surface structural changes. Specifically, the characteristics of local stable areas are integrated with the surface structural change trends, and a response trend tensor is constructed using two-dimensional tensor calculation technology to form the spatial distribution of environmental factors' responses to terrain. At the same time, sequence analysis technology is used to organize the temporal dynamic changes of materials into material evolution sequences, providing a reference basis for subsequent path planning.

[0049] For example, for the current target area, the disturbance amplitude map shows that the disturbance intensity at location A is 0.1 (low intensity) and at location B is 0.8 (high intensity). Therefore, location A is extracted as a local stable region feature. Simultaneously, the terrain continuity map shows that the slope slowly increases from 5° to 20°, accompanied by a gradual transition from grass to gravel. By integrating this information, a response trend tensor is constructed. This tensor contains the time series of slope gradient changes and continuous material changes, generating a material evolution sequence to record this dynamic process.

[0050] The response trend tensor is subjected to frequency decomposition and multi-scale fusion to obtain a disturbance intensity map, and a multidimensional energy consumption map is generated based on the disturbance intensity map and the material evolution sequence.

[0051] By performing frequency decomposition on the response trend tensor and using the Fast Fourier Transform (FFT) to decompose the frequency components of the dynamic response of environmental factors, the system combines this with multi-scale fusion technology to capture the impact of environmental response changes at different scales on path planning, thereby generating a disturbance intensity map. The disturbance intensity map is combined with the material evolution sequence to project multidimensional environmental parameters into spatial regions and correlate them with temporal changes. Specifically, the disturbance intensity map is combined with the material evolution sequence through multi-scale fusion to generate a multi-dimensional energy consumption map for the target area. This map can reflect the energy consumption patterns of the robot during path planning and the comprehensive impact of the environment on power consumption.

[0052] For example, applying a fast Fourier transform (FFT) to the response trend tensor for the current target area yields high-frequency components (humidity variation frequency of 0.8 Hz) and low-frequency components (pressure variation rate of 0.15 Hz). Through scale fusion, low-frequency humidity variations are combined with high-frequency pressure variations to form a disturbance intensity map. Combined with surface material switching information from the material evolution sequence, the resulting multidimensional energy consumption map clearly illustrates the power consumption distribution within the target area.

[0053] In step S200, directional gradient decomposition and regional density clustering are performed on the multi-dimensional energy consumption map to obtain the boundaries of the energy consumption abnormal area, and the resistance vector field is constructed using the boundaries of the energy consumption abnormal area.

[0054] By decomposing the multidimensional energy consumption map using directional gradients, significant trends in energy consumption distribution within the target area are extracted from the map. First, a gradient calculation algorithm (such as directional gradient decomposition based on partial differentials) is used to determine the rate of change of energy consumption along the path, as well as the gradient direction and magnitude at each node. Second, a regional density clustering algorithm (such as K-means clustering or DBSCAN) is used to analyze the distribution characteristics of high-energy consumption areas. Adjacent, high-energy-consumption areas are clustered into outlier regions, thereby defining the boundaries of these outlier regions. These boundaries clearly delineate the energy consumption characteristics of different regions along the path. Specifically, directional gradient decomposition is used to determine the direction and rate of energy consumption change for each segment within the target area. Regional density clustering is then used to cluster these high-energy-consumption segments, and regions with energy consumption exceeding a set threshold are identified as outlier regions, thereby forming the boundaries of the outlier regions. Finally, based on these boundaries, a resistance vector field is designed. By combining the direction of the energy gradient within the region with the regional density, a resistance vector with both direction and magnitude is formed.

[0055] For example, in the multidimensional energy consumption map, the surface of a certain target area has changed significantly, and the energy consumption gradient value is 5.2kWh / km. The distribution of high-density energy consumption areas is concentrated near the longitude and latitude coordinates (35.6586°N, 139.7454°E). Through the directional gradient decomposition algorithm, it is determined that the direction of energy consumption change in this area is 15° north-east, and the rate of change is 2.1kWh / km². The DBSCAN algorithm is used to cluster high-density areas, and the boundary definition threshold is 4.0kWh / km. The results show that there are two energy consumption anomaly areas in the area. These boundaries are used to construct a resistance vector field, where the vector direction is consistent with the energy consumption gradient direction, and the size reflects the degree of travel resistance.

[0056] The resistance vector field is fused with the historical energy consumption trajectory data to generate a path energy dissipation matrix and a power fluctuation diagram. The power critical point is extracted based on the path energy dissipation matrix, and the stable travel area is identified based on the power fluctuation diagram. Among them, the historical energy consumption trajectory data refers to the historical energy consumption and path data.

[0057] By fusing the resistance vector field with historical energy consumption trajectory data, the authors utilize the current environmental resistance information recorded in the resistance vector field and combine it with the path travel experience recorded in the historical energy consumption trajectory data (such as past energy consumption and path characteristics) to generate a path energy dissipation matrix and a power fluctuation map. The path energy dissipation matrix describes the energy consumption patterns within different path regions, while the power fluctuation map reflects the amplitude of power fluctuation along the path. Specifically, the energy dissipation matrix uses a constrained linear regression method to perform a weighted combination of the current vector field data and historical trajectory data to calculate the cumulative power consumption of each node along the path. The power fluctuation map analyzes the frequency and intensity of power changes along the path through a Fourier transform. Based on the energy dissipation matrix, power critical points are extracted: nodes in the matrix where the accumulated energy value exceeds a preset critical value. The power fluctuation map is used to identify stable travel areas: regions where the power fluctuation amplitude is below a specified threshold.

[0058] For example, the path power in a certain area of the resistance vector field is 120W, and historical energy consumption trajectory data shows a peak power of 250W and a cumulative energy dissipation of 1.4kWh. A constrained linear regression analysis is used to generate a path energy dissipation matrix. Nodes in the matrix where the cumulative power reaches the set critical value of 1.2kWh are marked as power critical points. Simultaneously, a Fourier transform is used to analyze the intensity of power fluctuations. When the fluctuation amplitude is less than 10%, the path is marked as a stable travel zone, covering the low-fluctuation power range in the matrix.

[0059] A cross-analysis is performed on the power critical point and the stable driving area, and based on the analysis results, a critical power spectrum and a safety margin distribution diagram are constructed. The risk index and energy consumption stability are calculated using the critical power spectrum and the safety margin distribution diagram.

[0060] By cross-analyzing power critical points and stable travel areas, the team correlated path power peaks with low-fluctuation segments, extracting information about critical and stable regions along the path. Based on this data, a critical power map was constructed, which displays the distribution of power consumption and travel stability along the path. Next, a safety margin calculation model was used to generate a safety margin distribution map. This map, combining environmental parameters with the current power state, reflects the safety of the robot's travel area. Specifically, the risk index was calculated by cross-calculating the two maps. The power critical point weights of the critical power map were combined with the safety threshold weights of the safety margin distribution map to estimate the risk value of each path. Energy consumption stability was normalized by combining the stable travel region weights of the critical power map with the regional stability of the safety margin distribution map.

[0061] For example, the critical power spectrum shows that the power consumption in a certain section of Path A is 1.8 kWh / km, at the critical power peak point, while Path B is 1.2 kWh / km, within the stable travel zone. A cross-analysis generates a safety margin distribution diagram, showing a safety margin of 35% for Path A and 80% for Path B. According to the risk index calculation formula, the risk index for Path A is 0.75, and the risk index for Path B is 0.25. The normalized energy consumption stability results show that the stability of Path A is 40%, and the stability of Path B is 90%.

[0062] The risk index and energy consumption stability are graded to generate risk classification labels. The resistance balance area is screened out according to the risk classification labels. The resistance balance area is integrated with the preset path accessibility constraints to obtain the path scoring table and obstacle avoidance level matrix.

[0063] By grading the risk index and energy consumption stability data, risk classification labels are generated according to preset classification thresholds (such as low, medium, and high risk). Subsequently, the resistance balance area is screened based on the classification label, that is, the area where the comprehensive risk index and stability are within the target range. The resistance balance area can exclude areas with high risk or low adaptability. Through path accessibility constraints (such as passability requirements, maximum slope restrictions, etc.), the screened balance area is integrated with the accessibility constraints to generate a path scoring table and an obstacle avoidance level matrix, respectively. Specifically, the path scoring table evaluates the priority of all candidate paths based on a weighted calculation of the accessibility constraints and risk classification labels; the obstacle avoidance level matrix determines the level of obstacle areas on each path to guide path planning.

[0064] For example, among the three paths, Path A has a high risk rating and a stability of 35%; Path B has a medium risk rating and a stability of 72%; and Path C has a low risk rating and a stability of 88%. Based on the scoring rules, Path C receives a score of 95, while Path B receives a score of 85. The obstacle avoidance matrix shows that Path C has no obstacle areas, while Path B has areas with minor obstacles. Both the final path scoring table and the obstacle avoidance matrix recommend Path C as the optimal path.

[0065] The path candidate set is extracted based on the path scoring table and obstacle avoidance level matrix.

[0066] By combining the data from the path score table and the obstacle avoidance level matrix, all paths are screened to extract a set of candidate paths. The paths in the candidate set have both high scores and reasonable obstacle avoidance levels, ensuring that they meet the requirements of robot path planning.

[0067] For example, after calculating the path score table, the top two paths are path B and path C. The obstacle avoidance level matrix shows the obstacle-free areas of the two paths. Ultimately, paths B and C are extracted as the path candidate set.

[0068] In step S300, a ternary path constraint vector is constructed based on the path length, slope, and surface material type of each path in the path candidate set, and a tensor transformation is performed on the ternary path constraint vector and the preset power threshold mapping matrix to obtain the path energy consumption response value and power distribution diagram.

[0069] By analyzing each path in the candidate set, the path length, slope, and surface material type are obtained. This information is then constructed into a three-dimensional path constraint vector. Path length represents the distance the robot must travel, slope describes the impact of path inclination on energy consumption, and surface material type (e.g., dirt, gravel, grass, etc.) is determined by a material recognition sensor. This is used to assess the impact of the path surface on the robot's motion resistance and energy consumption. Subsequently, a tensor transformation is performed between the three-dimensional path constraint vector and a preset power threshold mapping matrix (representing the robot's energy consumption limit region under different power conditions). The path energy response value is calculated using high-dimensional tensor operations (such as Kronecker products or tensor multiplications) to reflect the path's overall energy consumption. The result is visualized as a power distribution graph, depicting the distribution of power consumption along different paths. Specifically, path length, slope, and surface material type are used as the three dimensions of the path constraint vector to construct a three-dimensional path constraint representation. Weights are then defined in the power threshold mapping matrix, with path length accounting for 40%, slope for 35%, and surface material type for 25%. Through tensor transformation, the path energy consumption response value is used to calculate the power distribution diagram of each path to reflect the power consumption trend.

[0070] For example, path A is 200 meters long, has a slope of 15°, and is made of gravel. The path constraint vector is (200, 15, "sandstone"). The power threshold mapping matrix corresponds to a travel power range of [100W-300W] for the candidate path area. Through tensor transformation, the path energy consumption response value for path A is 120W. A power distribution map is also generated, showing that the power consumption of path A fluctuates significantly in the slope area, with consumption concentrated in the middle section.

[0071] The principal component decomposition of the path energy consumption response value is performed to extract the main characteristic axis of energy consumption. The power offset factor and characteristic compression sequence are generated according to the main characteristic axis of energy consumption and the power distribution diagram.

[0072] By performing principal component analysis (PCA) on the path energy response values, the principal characteristic axes of path energy consumption are extracted. These principal characteristic axes define the main factors influencing path energy consumption (such as the contribution of length, slope, or surface material ratio to power consumption). This step reduces the data dimensionality while retaining the core information of the path energy response values. Based on the correlation between the principal characteristic axes and the power distribution graph, the power offset factor is further calculated, which reflects the deviation in path energy consumption. At the same time, the path-related data is processed into a feature compression sequence for efficient subsequent analysis. Specifically, the PCA algorithm is used to reduce the dimensionality of the energy response values, selecting the characteristic axis with the highest contribution as the principal characteristic axis, such as paths with a high proportion of path length or significant slope changes. Based on the principal characteristic axes, the power offset factor is calculated to reflect the degree to which the path consumption deviates from the overall average level. Feature compression techniques are also used to generate a feature compression sequence, converting redundant feature data into an efficient sequence representation.

[0073] For example, the energy consumption response value for path B is decomposed into three characteristic dimensions: path length accounts for 60%, slope accounts for 30%, and material type accounts for 10%. The primary characteristic axis is path length. The power distribution diagram shows that the power of path B is concentrated in the slope variation area. Using the power offset calculation with a 15% offset factor, the resulting characteristic compression sequence is [60%, 15°].

[0074] The power offset factor is directionally reconstructed to obtain a directional response vector, and the directional response vector and the power offset factor are used to construct an energy consumption fluctuation graph.

[0075] Directional reconstruction is performed using the power offset factor to generate a directional response vector. Directional reconstruction combines the power change value in the offset factor with the path direction (such as the linear direction or slope direction) to reconstruct the directional vector of the path power consumption. The directional response vector represents the specific impact of the path power offset on the robot's travel direction. Subsequently, the directional response vector and the power offset factor are combined to construct an energy consumption fluctuation graph. This energy consumption fluctuation graph displays the spatial and temporal variations of power along the path and is used to assess path consumption stability. Specifically, the directional reconstruction equation is calculated using the numerical value of the offset factor, and the directional response vector is generated by combining the power changes in different path directions. Time series data is combined with the reconstructed vector to plot the energy consumption fluctuation graph.

[0076] For example, the power offset factor of path C is 10%, and the path direction is 30° east of north. After calculating the directivity reconstruction equation, the directional response vector [5%, 30°] is obtained. Combining this vector with the time series, an energy consumption fluctuation graph is generated, where the power peak is 230 W and the fluctuation amplitude is 12%.

[0077] Multi-scale interpolation is performed on the feature compression sequence to obtain the scale-direction distribution sequence. The scale-direction distribution sequence and the directional response vector are used to construct the directional accessibility matrix. Based on the directional accessibility matrix and the ternary path constraint vector, a dynamic adaptation score table is constructed.

[0078] Multiscale interpolation is performed on the feature compression sequence to generate a scale-direction distribution sequence. This interpolation technique employs wavelet transforms to smooth and compensate the path data at multiple scales, resulting in a multiscale variation pattern in the path direction data. Subsequently, the scale-direction distribution sequence is combined with the directional response vector to construct a directional accessibility matrix. The directional accessibility matrix describes the adaptability and stability of the path direction and reflects the robot's flexible adaptability in different path directions. Finally, the directional accessibility matrix is combined with a three-dimensional path constraint vector (including length, slope, and material) to calculate a dynamic adaptability score, which quantifies the overall adaptability score of each path. Specifically, the feature compression sequence is fused using wavelet decomposition technology, and the trajectory directions are smoothed and stratified into multiscale directional data. The directional response vector is then combined with the directional data to form a directional accessibility matrix. The interaction between the directional and path constraints is weighted and then combined with gradient grading to calculate the dynamic adaptability score.

[0079] For example, after applying a wavelet transform to the feature compression sequence of path D, a scaled directional distribution sequence {10°, 20°, …} is generated. This is combined with the directional response vector [7%, 15°] to form a directional accessibility matrix, resulting in a directional adaptability score of 85%. Combined with path length, slope, and material type, path D's adaptability score in the dynamic adaptability score table is 90.

[0080] The dynamic adaptation score table is integrated with the energy consumption fluctuation map to obtain the path energy consumption distribution feature set and the load response vector group. The path weight map is constructed based on the path energy consumption distribution feature set and the load response vector group.

[0081] By integrating the dynamic adaptability score table with energy consumption fluctuation data, and combining the path adaptability score with power fluctuation patterns, we generate a path energy consumption distribution feature set. This feature set reflects the regional patterns of path energy consumption variations and incorporates the load characteristics of each path segment to form a load response vector set. Based on this data, we generate a path weight map, which quantifies the overall path performance and is used to optimize path selection.

[0082] For example, the adaptability score of path E is 80 points, the energy consumption fluctuation graph shows that its power changes from 120W to 150W, and the load response vector group contains the path load range [60kg, 75kg]. Finally, a path weight map is generated, in which the weight of path E is 75%.

[0083] The steps of performing multi-scale interpolation on the feature compression sequence to obtain a scale direction distribution sequence, constructing a directional accessibility matrix using the scale direction distribution sequence and the directional response vector, and constructing a dynamic adaptation score table based on the directional accessibility matrix and the ternary path constraint vector specifically include: The feature compression sequence is divided into different levels according to the scale, and each level is interpolated using wavelet transform to obtain the scale direction distribution sequence at each scale.

[0084] By performing multi-scale decomposition on the feature compression sequence, the sequence is divided into multiple levels according to different spatial scales (such as micro, meso, and macro), allowing the specific characteristics of path direction changes at different scales to be discerned. Each level represents the details or trends of the path direction distribution at a specific scale. To ensure that the decomposed level sequences continuously and smoothly represent the path direction characteristics, wavelet transform is used to interpolate each level. Wavelet transform is a multi-resolution method that preserves both local and global signal characteristics, effectively smoothing discrete path direction data and supplementing missing values. Specifically, the feature compression sequence of each path is first decomposed into multiple layers of subsequences using discrete wavelet transform (DWT). Each layer of subsequence represents the distribution of path direction characteristics at different scales. Then, linear interpolation and nonlinear filtering are performed on each layer of subsequences using inverse wavelet transform (IDWT) to smooth the multi-scale path direction distribution. Ultimately, a scale direction distribution sequence corresponding to each level is generated, which preserves the path direction information at different scales.

[0085] For example, a wavelet decomposition is performed on the characteristic compression sequence [5°, 10°, 20°, 30°, 25°, 15°] of a particular path, dividing it into a low-frequency layer (macroscopic path trends) and a high-frequency layer (microscopic path fluctuation details). Linear interpolation is used to smooth the low-frequency layer into the sequence [8°, 15°, 25°], while quadratic nonlinear interpolation is used to obtain finer local details [2°, 4°, 3°]. Finally, an inverse wavelet transform is used to aggregate and generate a multi-scale directional distribution sequence. For example, the combined result of the macroscopic sequence [8°, 15°, 25°] and the microscopic sequence [2°, 4°, 3°] is [10°, 20°, 28°]. These data constitute a complete distribution description of path directions at different scales.

[0086] The change rate and fluctuation amplitude of the path direction are extracted according to the scale direction distribution sequence, and the change rate and fluctuation amplitude are weightedly fused with the direction response vector to obtain the directional accessibility matrix.

[0087] By analyzing the scaled directional distribution sequence, the rate of change and fluctuation amplitude of the path direction are extracted. The rate of change indicates the rate of change of the path direction with distance (e.g., whether the direction is stable), while the fluctuation amplitude indicates the severity of the directional fluctuation (e.g., the number and intensity of turning points). These data directly impact path accessibility, as drastic directional changes increase robot energy consumption and path complexity. The extracted rate of change and fluctuation amplitude are then combined with the directional response vector, a quantitative indicator of the alignment between path power consumption and path direction. Specifically, the rate of change is calculated from the scaled directional distribution sequence using derivative operations, such as normalizing the differences between adjacent directional points; the fluctuation amplitude is calculated using standard deviation or peak-to-valley analysis. The calculated rate of change and fluctuation amplitude are then weighted and fused with the directional response vector. The weights can be dynamically adjusted based on path design requirements (e.g., 50% each for the rate of change and the fluctuation amplitude). The resulting weighted fusion matrix is called the directional accessibility matrix. Each element in the matrix represents the accessibility score of a path direction within the current area, taking into account both the stability of the path direction and its adaptability to the power load.

[0088] For example, the scale-direction distribution sequence for a path is [5°, 10°, 20°, 15°, 10°]. The calculated directional change rate is [5°, 10°, -5°, -5°], and the directional fluctuation amplitude is 8°. Combined with the directional response vectors [3%, 5%, 2%], a weighted fusion is performed, with the change rate weighted at 60% and the fluctuation amplitude weighted at 40%. This ultimately generates a directional accessibility matrix. For example, a directional accessibility score for a region might be [0.85, 0.70, 0.60], indicating the path's adaptability in different directions.

[0089] The directional accessibility of the directional accessibility matrix is combined with the path constraints of the ternary path constraint vector to calculate the adaptability score of each path and obtain a dynamic adaptability score table.

[0090] By interactively calculating the accessibility data of the path direction in the directional accessibility matrix and the ternary path constraint vector, each path constraint vector consists of length, slope and surface material type, corresponding to different physical space restrictions of the path. First, the value range of the directional accessibility matrix and the path constraint vector is adjusted through normalization processing to make the two comparable. Then, a weighted combination calculation is performed to combine each constraint condition in the ternary path constraint (such as slope accounting for 40%, surface material accounting for 30%, and path length accounting for 30%) with the directional accessibility item by item to calculate the comprehensive adaptability score of each path. Specifically, the dynamic adaptability score is calculated using the weighted summation method, and the formula is: Adaptability score = Σ(directional reachability path constraint weight).

[0091] The results are recorded in a table, and the dynamic adaptation score table ranks the fitness scores of all path candidate sets for subsequent optimization.

[0092] For example, the directional accessibility matrix score of a path is [0.85, 0.70, 0.60], and the three-dimensional path constraint vectors are length 100m (weight 30%), slope 10° (weight 40%), and material is sand (weight 30%). The comprehensive adaptability score calculated by combining the weights is: Adaptability score = (0.8530%) + (0.7040%) + (0.6030%) = 0.745.

[0093] Finally, a dynamic adaptability score table is generated, in which the path has an overall score of 74.5 points and is prioritized with the candidate paths. The dynamic adaptability score table can also include the paths with the lowest and highest adaptability scores for comparative analysis.

[0094] In step S400, the current voltage and current current of the walking robot are obtained by using sensors, the electric energy input and electric energy output of the current voltage and current current are calculated, and the current density vector and power voltage drop value are generated based on the electric energy input and electric energy output.

[0095] Voltage and current sensors installed on the robot provide real-time data on the robot's current voltage (V) and current (A). Power input refers to the power the robot receives from the power supply or battery system, while power output refers to the power consumed by the robot while performing its tasks. The power input and output (unit: watts) are calculated by multiplying the voltage and current. Based on this, the power consumption of each part of the robot is analyzed by calculating the current density vector and power voltage drop. Specifically, power input is calculated as: Electric energy input = voltage (V) × current (A).

[0096] The power output can be calculated by monitoring the power consumption of the robot's drive system and various components. For example, the power consumption of the robot's drive motor is the power drop value, which indicates the power loss in the system.

[0097] For example, if the robot's current voltage is 24V and the current is 3A, the power input is: Power input = 24V × 3A = 72W.

[0098] Assuming that the power consumption of the robot drive motor and control system is 60W and 5W respectively, the power output is: Power output = 60W + 5W = 65W.

[0099] The current density vector can be expressed as the density of current distribution in different components of the robot, and the power voltage drop value reflects the loss of electrical energy in systems such as motors and sensors.

[0100] A short-term trend fitting analysis is performed on the current density vector to obtain the voltage fluctuation factor and load response delay. A multidimensional power stability matrix is constructed based on the power voltage drop value and load response delay. The multidimensional power stability matrix is integrated with the voltage fluctuation factor to generate the power stability distribution.

[0101] By performing short-term trend fitting on the current density vector (e.g., using linear regression or exponential smoothing algorithms), the changing trends of power input and output are predicted. Based on the current density changes, the voltage fluctuation factor can be calculated, which describes the stability of voltage changes over time. The load response delay is also calculated, representing the robot's response time to load changes. The power drop values and the load response delay are combined to construct a multidimensional power stability matrix, representing the robot's power stability under different loads. Finally, the multidimensional power stability matrix is combined with the voltage fluctuation factor to generate a power stability distribution, which is used to analyze the robot's power adaptability under different path conditions. Specifically, the current density vector is fitted with a short-term trend based on time series data. The voltage fluctuation factor can be calculated using the standard deviation or average rate of change of voltage changes. The load response delay is calculated based on the robot's response time to load changes. Based on this data, a power stability distribution map is generated.

[0102] For example, suppose short-term trend fitting calculates the current density vector's rate of change to be 0.2 A / s, the voltage fluctuation factor to be 5% (the ratio of the voltage fluctuation range to the average voltage), and the load response delay to be 0.3 seconds. The power dropout value is 5 W. Based on this data, a multidimensional power stability matrix can be constructed. This matrix reflects power stability under different loads and, combined with the voltage fluctuation factor, generates an energy stability distribution map, showing the robot's energy stability under different operating conditions.

[0103] The power stability distribution and path weight map are used to calculate the path load entropy value and power supply adaptation factor, and the path constraint adjustment table and alternative path index set are constructed based on the path load entropy value and power supply adaptation factor.

[0104] The generated power stability distribution is combined with the path weight map to calculate the load entropy and power adaptation factor for each path. The load entropy value indicates the degree of variability of different load conditions on the path. A high load entropy value indicates significant load fluctuations on the path. The power adaptation factor indicates the path's adaptability to power input. By comparing these metrics, it is possible to identify which paths are stable under load or power consumption changes, and which paths are affected by power shortages. Specifically, the load entropy value can be obtained using the Shannon entropy calculation method to measure the uncertainty of the load distribution on the path. The power adaptation factor can be calculated by calculating the ratio of the path power demand to the battery capacity, indicating the path's power adaptability. Based on these values, a path constraint adjustment table is constructed, which is used to optimize path selection and generate a set of alternative path indexes for subsequent use.

[0105] For example, the calculated load entropy value for path A is 0.45, and the power supply adaptation factor is 1.2 (indicating that path A is more adaptable to the battery), while the load entropy value for path B is 0.75, and the power supply adaptation factor is 0.85. Based on these values, path A is preferred due to its lower load entropy and higher adaptation factor. The path constraint adjustment table shows that path A is suitable for higher loads, while path B is more suitable for lighter loads.

[0106] Based on the path constraint adjustment table, the priority of the path weight map is corrected to obtain an updated path weight map. The alternative path index set is screened for compatibility to obtain a screened index set. The path adaptability result is extracted based on the updated path weight map and the screened index set.

[0107] By modifying the path weight map based on the path constraint adjustment table, the adjusted map can more accurately reflect the adaptability of the paths under different power conditions. The updated path weight map provides a comprehensive priority and stability score for each path. On this basis, the compatibility screening of the alternative path index set is used to select paths that are compatible with factors such as current power conditions and load requirements. Ultimately, the most suitable paths are extracted based on the updated path weight map and the filtered index set to generate a path adaptability result. Specifically, the path priority modification combines the power stability and load adaptability of the path through a weighted average method, ensuring that paths with strong adaptability and low power consumption are prioritized. Through compatibility screening, paths that do not meet the current conditions are eliminated to obtain the most suitable path set.

[0108] For example, the weight of path A is adjusted to 95 points, the weight of path B is adjusted to 80 points, and the weight of path C is adjusted to 70 points. Through the screening of the alternative path index set, path C is eliminated due to its high load entropy value and low adaptability factor, and path B is retained. Path A and path B are sorted according to their priorities, and path A is finally selected as the optimal path with an adaptability score of 92 points.

[0109] In step S500, segment disturbance injection is performed on the weight distribution in the path adaptability result and the historical path change trend to obtain a disturbance response sequence, and local polynomial fitting is performed on the disturbance response sequence to extract the segment fitting curvature map and the node offset scalar field; wherein, the historical path change trend refers to the change pattern of the path curvature change, slope fluctuation, and surface condition change data when the walking robot moves on the path.

[0110] By combining the weight distribution of the path adaptability results with historical path change trends, segmental disturbance injection is implemented. Historical path change trends can be obtained by monitoring and analyzing data such as changes in path curvature, slope fluctuations, and changes in surface conditions during robot movement. The purpose of segmental disturbance injection is to simulate environmental changes to the path, such as the presence of obstacles and terrain changes, to more accurately predict the stability and energy efficiency of the path during actual execution. Specifically, disturbance injection employs a random perturbation method to inject disturbances of varying intensities and types (such as changes in direction and slope) into each segment of the path. Next, a local polynomial fitting algorithm (such as quadratic or cubic polynomial regression) is used to fit the disturbance response sequence to extract local path variation characteristics. The resulting segmental curvature map and node offset scalar field can characterize path curvature changes and node offsets, providing a basis for subsequent path correction.

[0111] For example, historical data for Path A show a slope varying between 5° and 15° in a certain section, with rock obstacles along the path. During the simulation, the injected perturbation simulated the abrupt change in slope (from 15° to 30°) and the sudden appearance of the obstacle. A local polynomial fit generated a fitted curvature plot for this section, showing that the path's curvature increased from 3.2° to 10.5°. Furthermore, the node offset scalar field displays the displacement of the nodes in this section, providing data support for subsequent path optimization and correction.

[0112] The segment fitting curvature map and the node offset scalar field are compared segment by segment to generate a path stability vector group and a direction fusion value. The path stability vector group is used to construct a dynamic correction path set, and the stability of the dynamic correction path set is screened by the direction fusion value to obtain a path fitting map.

[0113] By comparing the segment-fitted curvature map with the node offset scalar field on a segment-by-segment basis, the path stability of each segment can be analyzed based on the curvature change and node offset magnitude. For example, by calculating the curvature change rate and offset for each segment of the path, a set of path stability vectors is generated, where each vector represents the stability of the path in a specific region. This is then weighted with the directional response vector (which represents the impact of path direction on energy consumption) to generate a directional fusion value. The directional fusion value represents the directional stability of the path in each segment and reflects the path's travel stability and feasibility. Using this set of path stability vectors, a set of dynamically corrected paths can be constructed to identify paths with high stability. Combining these paths with the directional fusion value yields a stability-screened path fitting map that demonstrates the adaptability and stability of each path under various disturbance conditions.

[0114] For example, in the fitted curvature graph for path B, the curvature of a certain section changes dramatically (increasing from 10° to 40°) and exhibits significant node offsets (maximum 20cm). By comparing the curvature change and node offsets for this section, the stability vector for this section is calculated to be 0.4, and the directional fusion value is 0.6, ultimately determining it as an unstable path. After stability screening, path B is corrected and combined with other stable paths to form an updated path fitting graph.

[0115] An energy consumption offset table and a travel efficiency distribution diagram are analyzed based on the path fitting graph and the path candidate set. An efficiency loss factor table is constructed based on the energy consumption offset table, and a path hierarchical screening group is constructed based on the travel efficiency distribution diagram.

[0116] An energy consumption offset table is generated by combining the stability data extracted from the path fitting graph with the energy consumption data in the path candidate set. This table shows the energy consumption offset of each path under different stability conditions, reflecting the energy efficiency level and energy consumption trend of the path. At the same time, the travel efficiency of the path is analyzed based on the travel efficiency distribution map. The travel efficiency distribution map shows the changes in efficiency when traveling on the path, taking into account factors such as path length, slope, and energy consumption. Based on the energy consumption offset table, an efficiency loss factor table can be constructed to quantify the energy efficiency loss of the path. The efficiency loss factor table shows the energy loss caused by the path during actual travel. Then, using the travel efficiency distribution map, a path hierarchical screening group is constructed to grade the paths according to their efficiency and sort them from high to low to ensure that the most efficient path is selected.

[0117] For example, the energy consumption deviation table for path C shows a 30% increase in energy consumption in a certain section, while the efficiency loss factor table shows an energy loss factor of 0.2 on the uphill section. The travel efficiency distribution diagram for path D shows that path D is most efficient when traveling in a straight line, with an efficiency value of 85%. Through the path classification screening, path D is marked as the preferred path and is listed as the most suitable path.

[0118] The path scoring comparison matrix and the priority sequence index table are generated through the efficiency loss factor table and the path classification screening group, and the optimal path output result is output according to the path scoring comparison matrix and the priority sequence index table.

[0119] The efficiency loss factor table and the path ranking filter group generate a path scoring comparison matrix, which compares different paths in terms of energy efficiency and travel efficiency. Each path's score reflects its overall performance across multiple dimensions, including energy efficiency, stability, and adaptability. A priority sequence index table is then used to sort all paths, and the optimal path is output based on the scoring comparison matrix and the priority sequence index table. This optimal path output ensures that the robot selects the most appropriate path when performing its task, thereby improving efficiency and saving energy.

[0120] For example, path A is scored 90, path B is scored 80, and path C is scored 75. Based on the score comparison matrix, path A is considered the optimal path and is selected for the actual task. The priority sequence index table shows that path A has the highest priority, and paths B and C are selected as alternative paths.

[0121] In this embodiment, a series of technical approaches are employed to optimize and screen the target path through in-depth processing and dynamic analysis of multidimensional data during the path planning process for a walking robot, ultimately outputting the optimal path. Specifically, sensors first acquire real-time data on air pressure, humidity, temperature difference, and surface material in the target area. These environmental parameters are spatially labeled using geographic reference coordinates to generate a multi-factor sequence. Subsequently, multiscale interpolation and sliding window processing techniques are used to extract environmental characteristics. Local stable regions and surface structure trends are analyzed from disturbance amplitude maps and terrain continuity data. Frequency decomposition and tensor fusion are then used to generate a multidimensional energy consumption map, providing a preliminary energy consumption reference for path planning. During the dynamic path planning phase, directional gradient decomposition and regional density clustering methods are used to mark the boundaries of abnormally high energy consumption areas using the multidimensional energy consumption map. A resistance vector field is then constructed. This is combined with historical energy consumption trajectory data to analyze power critical points and stable travel areas, generating a critical power map and a safety margin distribution map to quantify risk index and energy consumption stability. Combined with dynamic scoring technology, the directional accessibility matrix and the ternary path constraint vector are combined to generate a dynamic adaptation scoring table to prioritize candidate paths. The power stability distribution is further calculated based on the power input and output, and the path weight map is optimized. At the same time, the efficiency loss factor and the graded screening group are used to construct the path scoring matrix and priority index table, and finally the optimal path result with the highest priority is output according to the scoring comparison matrix. The solution of this embodiment effectively integrates environmental complexity, energy efficiency and path adaptability analysis through multi-angle disturbance injection, dynamic correction and stability screening, improves the intelligence level of path planning, and provides extremely high reliability and flexibility for the walking robot to perform tasks in complex dynamic environments.

[0122] Example 3: like Figure 2 As shown, the present application provides a walking robot path planning system 10 , which includes an acquisition module 11 , an evaluation module 12 , a generation module 13 , a fitting module 14 and an analysis module 15 .

[0123] The acquisition module 11 is mainly used to obtain the environmental parameters of the target area, bind the environmental parameters with spatial tags, obtain an environmental multi-factor sequence, and construct a multi-dimensional energy consumption map based on the environmental multi-factor sequence.

[0124] The evaluation module 12 is mainly used to use the multi-dimensional energy consumption map to analyze the travel resistance of the target area, generate a resistance vector field, perform power consumption risk assessment on the resistance vector field, obtain a risk classification label, and screen the path of the target area according to the risk classification label to obtain a path candidate set.

[0125] The generation module 13 is mainly used to perform tensor transformation and graph generation on the path candidate set and the preset power threshold mapping matrix to obtain a path weight graph.

[0126] The fitting module 14 is mainly used to obtain the current voltage and current current of the walking robot, and perform trend fitting on the current voltage and current current to obtain the power stability distribution, and perform path correction on the path weight map according to the power stability distribution to obtain the path adaptability result.

[0127] The analysis module 15 is mainly used to perform segment disturbance injection and local multinomial fitting on the path adaptability results, generate a path fitting map, and analyze the optimal path output result based on the path fitting map and the path candidate set.

[0128] In this embodiment, efficient path planning for a walking robot in complex environments is achieved by combining the real-time sensor data acquisition and geospatial tag binding capabilities of acquisition module 11, the path characteristic analysis and risk assessment capabilities of evaluation module 12, the weighted map generation technology of generation module 13, the power stability correction logic of fitting module 14, and finally the dynamic fitting and path optimization methods performed by analysis module 15. Specifically, acquisition module 11 acquires environmental parameters such as air pressure, humidity, temperature difference, and surface material of the target area in real time, and performs spatial tag binding based on geographic reference coordinates to generate a multi-factor sequence containing multidimensional environmental information. These sequences are used to construct a multidimensional energy consumption map, reflecting the resistance and energy consumption distribution of the target area, providing basic data support for subsequent path planning. In evaluation module 12, based on the data from the multidimensional energy consumption map, travel resistance analysis and power consumption risk assessment are performed, generating a resistance vector field and risk classification labels. Based on this data, a preliminary set of candidate paths is screened, eliminating high-resistance and high-risk path segments to ensure the reliability of path planning. The generation module 13 performs tensor transformation and graph generation on the path candidate set and the preset power threshold mapping matrix to obtain a path weight graph. The weight graph comprehensively considers factors such as path length, slope, and material type, providing accurate quantitative evaluation for path optimization. The fitting module 14 uses voltage and current sensors to obtain the real-time power state of the walking robot, performs trend fitting to generate a power stability distribution, and modifies the path in combination with the path weight graph to obtain a path adaptability result that better fits the power state. Finally, the analysis module 15 uses segment disturbance injection and local multi-factor fitting techniques to explore the local stability characteristics of the path, generate a path fitting graph, and analyze the comprehensive performance indicators of each path in combination with the path candidate set, and finally output the optimal path solution. This embodiment, through modular design, relies on an overall planning system with clear division of labor and collaborative work among modules, not only ensuring the efficiency and accuracy of data processing and decision-making processes, but also realizing real-time adjustment and optimization of path planning results in dynamic environments, providing a reliable solution for autonomous navigation of robots in complex environments.

[0129] It should be noted that, those skilled in the art will clearly understand that, for the sake of convenience and brevity of description, the specific working processes of the above-described system and each module can refer to the corresponding processes in the aforementioned embodiment 1 and will not be repeated here.

[0130] The structures, proportions, sizes, etc. depicted in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not intended to limit the conditions under which the present invention can be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportional relationships, or adjustments in size should still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and objectives that can be achieved by the present invention.

[0131] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A walking robot path planning method, characterized in that: include: Acquiring environmental parameters of a target area, binding the environmental parameters to spatial tags to obtain an environmental multi-factor sequence, and constructing a multidimensional energy consumption map based on the environmental multi-factor sequence; Performing a travel resistance analysis on a target area using the multi-dimensional energy consumption map to generate a resistance vector field, performing a power consumption risk assessment on the resistance vector field to obtain a risk classification label, and screening paths in the target area based on the risk classification label to obtain a set of candidate paths; Performing tensor transformation and graph generation on the candidate path set and a preset power threshold mapping matrix to obtain a path weight graph; Obtaining a current voltage and a current current of the walking robot, performing trend fitting on the current voltage and the current current to obtain a power stability distribution, and performing path correction on the path weight map according to the power stability distribution to obtain a path adaptability result; The path adaptability result is subjected to segment disturbance injection and local multinomial fitting to generate a path fitting map, and an optimal path output result is analyzed based on the path fitting map and the path candidate set.

2. The walking robot path planning method according to claim 1, characterized in that: The steps of obtaining environmental parameters of the target area, binding the environmental parameters to spatial tags to obtain an environmental multi-factor sequence, and constructing a multidimensional energy consumption map based on the environmental multi-factor sequence include: Obtaining the target area's air pressure, humidity, temperature difference, and surface material data to generate environmental parameters, and spatially tagging the environmental parameters using geographic reference coordinates to obtain an environmental multi-factor sequence; wherein the geographic reference coordinates refer to the current latitude, longitude, and altitude data; Performing sliding window processing on the environmental multi-factor sequence to extract response feature vectors and terrain switching marks, generating coupling nodes based on the response feature vectors, and constructing a disturbance amplitude map and a terrain continuity map through the coupling nodes and the terrain switching marks; Extracting local stable region features based on the disturbance amplitude map, extracting surface structure change trends based on the terrain continuity map, and fusing the local stable region features with the surface structure change trends to construct a response trend tensor and a material evolution sequence; The response trend tensor is subjected to frequency decomposition and multi-scale fusion to obtain a disturbance intensity map, and a multi-dimensional energy consumption map is generated based on the disturbance intensity map and the material evolution sequence.

3. The walking robot path planning method according to claim 1, characterized in that: The steps of analyzing the travel resistance of the target area using the multi-dimensional energy consumption map to generate a resistance vector field, performing a power consumption risk assessment on the resistance vector field to obtain a risk classification label, and screening paths in the target area according to the risk classification label to obtain a path candidate set include: Performing directional gradient decomposition and regional density clustering on the multidimensional energy consumption map to obtain energy consumption abnormal region boundaries, and constructing a resistance vector field using the energy consumption abnormal region boundaries; fusing the resistance vector field with historical energy consumption trajectory data to generate a path energy dissipation matrix and a power fluctuation graph, extracting power critical points based on the path energy dissipation matrix, and identifying stable travel areas based on the power fluctuation graph; wherein the historical energy consumption trajectory data refers to historical energy consumption and path data; Performing a cross-analysis on the power critical point and the stable travel area, constructing a critical power spectrum and a safety margin distribution diagram based on the analysis results, and calculating a risk index and energy consumption stability using the critical power spectrum and the safety margin distribution diagram; The risk index and the energy consumption stability are graded to generate a risk grading label, a resistance balance area is screened out based on the risk grading label, and the resistance balance area is integrated with a preset path accessibility constraint to obtain a path scoring table and an obstacle avoidance level matrix; A path candidate set is extracted according to the path scoring table and the obstacle avoidance level matrix.

4. The walking robot path planning method according to claim 1, characterized in that: The step of performing tensor transformation and graph generation on the path candidate set and the preset power threshold mapping matrix to obtain a path weight graph includes: Constructing a ternary path constraint vector based on the path length, slope, and surface material type of each path in the path candidate set, and performing a tensor transformation on the ternary path constraint vector and a preset power threshold mapping matrix to obtain a path energy consumption response value and a power distribution diagram; performing principal component decomposition on the path energy consumption response value to extract the main characteristic axis of energy consumption, and generating a power offset factor and a characteristic compression sequence according to the main characteristic axis of energy consumption and the power distribution diagram; Directionally reconstructing the power offset factor to obtain a direction response vector, and constructing an energy consumption fluctuation graph using the direction response vector and the power offset factor; Performing multi-scale interpolation on the feature compression sequence to obtain a scale direction distribution sequence, constructing a directional accessibility matrix using the scale direction distribution sequence and the directional response vector, and constructing a dynamic adaptation score table based on the directional accessibility matrix and the ternary path constraint vector; The dynamic adaptation score table is integrated with the energy consumption fluctuation graph to obtain a path energy consumption distribution feature set and a load response vector group, and a path weight map is constructed based on the path energy consumption distribution feature set and the load response vector group.

5. The walking robot path planning method according to claim 4, characterized in that: The steps of performing multi-scale interpolation on the feature compression sequence to obtain a scale direction distribution sequence, constructing a directional accessibility matrix using the scale direction distribution sequence and the directional response vector, and constructing a dynamic adaptation score table based on the directional accessibility matrix and the ternary path constraint vector include: The feature compression sequence is divided into different levels according to the scale, and each level is interpolated using wavelet transform to obtain the scale direction distribution sequence at each scale; Extracting the change rate and fluctuation amplitude of the path direction according to the scale direction distribution sequence, and weightedly fusing the change rate and fluctuation amplitude with the direction response vector to obtain a directional reachability matrix; The directional accessibility of the directional accessibility matrix is combined with the path constraint condition of the ternary path constraint vector to obtain a dynamic adaptation score table.

6. The walking robot path planning method according to claim 1, characterized in that: The steps of obtaining the current voltage and current current of the walking robot, performing trend fitting on the current voltage and current current to obtain a power stability distribution, and performing path correction on the path weight map according to the power stability distribution to obtain a path adaptability result include: Using sensors to obtain a current voltage and a current current of the walking robot, calculating an electric energy input and an electric energy output for the current voltage and the current current, and generating a current density vector and a power voltage drop value based on the electric energy input and the electric energy output; performing a short-term trend fitting analysis on the current density vector to obtain a voltage fluctuation factor and a load response delay, constructing a multidimensional power stability matrix based on the power voltage drop value and the load response delay, and fusing the multidimensional power stability matrix with the voltage fluctuation factor to generate an electric energy stability distribution; Calculating a path load entropy value and a power supply adaptation factor using the power energy stability distribution and the path weight map, and constructing a path constraint adjustment table and an alternative path index set based on the path load entropy value and the power supply adaptation factor; Based on the path constraint adjustment table, the priority of the path weight map is corrected to obtain an updated path weight map, the alternative path index set is screened for compatibility to obtain a screened index set, and the path adaptability result is extracted based on the updated path weight map and the screened index set.

7. The walking robot path planning method according to claim 1, characterized in that: The step of performing segment disturbance injection and local multinomial fitting on the path adaptability result to generate a path fitting graph, and analyzing the optimal path output result based on the path fitting graph and the path candidate set includes: Performing segment disturbance injection on the weight distribution and historical path change trend in the path adaptability result to obtain a disturbance response sequence, performing local polynomial fitting on the disturbance response sequence, and extracting a segment fitting curvature map and a node offset scalar field; wherein the historical path change trend refers to the change pattern of the curvature change, slope fluctuation, and surface condition change data of the path when the walking robot moves on the path; Performing segment-wise comparison on the segment fitting curvature map and the node offset scalar field to generate a path stability vector group and a direction fusion value, constructing a dynamic correction path set using the path stability vector group, and performing stability screening on the dynamic correction path set using the direction fusion value to obtain a path fitting map; Analyzing an energy consumption offset table and a travel efficiency distribution diagram based on the path fitting graph and the path candidate set, constructing an efficiency loss factor table based on the energy consumption offset table, and constructing a path hierarchical screening group based on the travel efficiency distribution diagram; A path scoring comparison matrix and a priority sequence index table are generated through the efficiency loss factor table and the path hierarchical screening group, and an optimal path output result is output according to the path scoring comparison matrix and the priority sequence index table.

8. A walking robot path planning system, characterized in that: include: An acquisition module is used to acquire environmental parameters of a target area, perform spatial tag binding on the environmental parameters to obtain an environmental multi-factor sequence, and construct a multi-dimensional energy consumption map based on the environmental multi-factor sequence; an evaluation module, configured to analyze the travel resistance of a target area using the multi-dimensional energy consumption map to generate a resistance vector field, perform a power consumption risk assessment on the resistance vector field to obtain a risk classification label, and perform path screening on the target area based on the risk classification label to obtain a set of candidate paths; A generation module, configured to perform tensor transformation and graph generation on the candidate path set and a preset power threshold mapping matrix to obtain a path weight graph; a fitting module, configured to obtain a current voltage and a current current of the walking robot, perform trend fitting on the current voltage and the current current to obtain a power stability distribution, and perform path correction on the path weight map according to the power stability distribution to obtain a path adaptability result; The analysis module is used to perform segment disturbance injection and local multinomial fitting on the path adaptability result to generate a path fitting map, and analyze the optimal path output result based on the path fitting map and the path candidate set.

Images