Petrochemical explosion-proof humanoid robot trajectory planning method, device and equipment

Abstract

The invention discloses a petrochemical explosion-proof humanoid robot trajectory planning method, device and equipment, and the method comprises the steps: collecting petrochemical environment gas concentration data, robot joint motion parameters and target position information, carrying out the dangerous region calibration of the gas concentration data, generating a forbidden map, and determining safety nodes and joint motion ranges; constructing an initial path skeleton based on target position information and safety node distribution, identifying a local dangerous section and a major dangerous section, and generating a motion constraint condition and a poison avoidance path; generating a motion primitive library based on the motion constraint condition and the joint motion range, performing stability filtering to generate stable motion primitives, and forming an advancing scheme set; acquiring a joint load peak value to construct a dynamic envelope, and judging execution fitness to form an execution probability field; and constructing a three-dimensional decision space based on the poison avoidance path, the advancing scheme set and the execution probability field, searching an optimal convergence point to generate a final planning trajectory, and providing a safe and efficient trajectory planning scheme for the petrochemical explosion-proof humanoid robot.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of robot trajectory planning, in particular to a petrochemical explosion-proof humanoid robot trajectory planning method, device and equipment. BACKGROUND

[0002] The petrochemical industry environment is a typical high-risk workplace, which has multiple safety threats such as toxic gas leakage, flammable gas accumulation, and corrosive substance distribution, which puts high requirements on the safety navigation of robots. The traditional robot path planning method is mainly designed for general industrial environments and lacks in-depth consideration of the special dangerous factors in petrochemical environments, making it difficult to effectively handle complex constraint conditions such as gas concentration changes, static electricity accumulation, and explosion-proof requirements.

[0003] Existing dangerous environment path planning techniques mostly use static maps and fixed constraints, which cannot adjust the path according to the dynamic changes of the environment danger level, and lack of coordinated consideration of the robot's own dynamics characteristics and environmental safety requirements. In particular, there are obvious deficiencies in handling multiple dangerous factors superimposed, balancing path safety and execution efficiency, and balancing action stability and explosion-proof requirements, resulting in planned paths that are either too conservative affecting work efficiency or have insufficient safety margins with potential risks. Therefore, there is an urgent need for a method to solve at least one of the above problems. SUMMARY

[0004] The present application discloses a petrochemical explosion-proof humanoid robot trajectory planning method, device and equipment, aiming to solve the technical problems of insufficient safety and poor adaptability of traditional trajectory planning methods in petrochemical dangerous environments. By establishing a coupling analysis model of path skeleton features and robot motion characteristics, multiple path risk factors such as path safety level mutation, geometric risk concentration, and connection vulnerability are comprehensively evaluated, and a safety constraint system is constructed. Using action primitive decomposition and stability filtering technology, combined with dynamics envelope analysis and execution probability field construction, the unified evaluation of trajectory safety and execution feasibility is realized, and the final planning trajectory that takes into account safety, feasibility and execution is generated.

[0005] The present application discloses a petrochemical explosion-proof humanoid robot trajectory planning method, device and equipment, aiming to solve the technical problems of insufficient safety and poor adaptability of traditional trajectory planning methods in petrochemical dangerous environments. By establishing a coupling analysis model of path skeleton features and robot motion characteristics, multiple path risk factors such as path safety level mutation, geometric risk concentration, and connection vulnerability are comprehensively evaluated, and a safety constraint system is constructed. Using action primitive decomposition and stability filtering technology, combined with dynamics envelope analysis and execution probability field construction, the unified evaluation of trajectory safety and execution feasibility is realized, and the final planning trajectory that takes into account safety, feasibility and execution is generated. Collecting petrochemical environment gas concentration data, robot joint motion parameters and target position information, calibrating the dangerous area of the gas concentration data to generate a forbidden map, determining safety nodes from the forbidden map, and determining joint activity range based on the joint motion parameters; Based on the target position information and the safety node distribution, a path is searched to construct an initial path skeleton, a safety analysis is performed on the initial path skeleton to identify local dangerous segments, and motion constraints and posture balance requirements are generated based on the local dangerous segments; The initial path skeleton is used to identify a major dangerous section, a threat level of the major dangerous section is evaluated to generate a risk weight, and the risk weight is used for path re-planning to generate a poison-avoiding path; A motion primitive library is generated based on the motion constraint condition and the joint range of motion, stable motion primitives are generated by stability filtering the motion primitive library according to the posture balance requirement, and a set of marching schemes is formed by path weaving the stable motion primitives between the safety nodes; Joint load peaks in the set of marching schemes are acquired, a dynamics envelope is constructed based on the joint load peaks, an execution fitness is determined according to the dynamics envelope, and an execution probability field is formed by spatial coordinate projection of the execution fitness; A three-dimensional decision space is constructed based on the poison-avoiding path, the set of marching schemes and the execution probability field, an optimal convergence point is found in the three-dimensional decision space, and a final planning trajectory is generated based on the optimal convergence point.

[0006] The second aspect of the present application proposes a petrochemical explosion-proof humanoid robot trajectory planning device, comprising: An environment perception module is used to collect petrochemical environment gas concentration data, robot joint motion parameters and target position information, to calibrate a dangerous area from the gas concentration data to generate a no-go map, to determine safety nodes from the no-go map, and to determine a joint range of motion based on the joint motion parameters; A path analysis module is used to search for a path based on the target position information and the safety node distribution to construct an initial path skeleton, to perform safety analysis on the initial path skeleton to identify a local dangerous section, and to generate a motion constraint condition and a posture balance requirement based on the local dangerous section; A risk avoidance module is used to identify a major dangerous section according to the initial path skeleton, to evaluate a threat level of the major dangerous section to generate a risk weight, and to use the risk weight for path re-planning to generate a poison-avoiding path; A motion generation module is used to generate a motion primitive library based on the motion constraint condition and the joint range of motion, to generate stable motion primitives by stability filtering the motion primitive library according to the posture balance requirement, and to form a set of marching schemes by path weaving the stable motion primitives between the safety nodes; An adaptability evaluation module is used to acquire joint load peaks in the set of marching schemes, to construct a dynamics envelope based on the joint load peaks, to determine an execution fitness according to the dynamics envelope, and to form an execution probability field by spatial coordinate projection of the execution fitness; A trajectory decision module is configured to establish a three-dimensional decision space based on the poison-avoiding path, the set of travel schemes and the execution probability field, find an optimal convergence point in the three-dimensional decision space, and generate a final planning trajectory based on the optimal convergence point.

[0007] A third aspect of the present application provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the petrochemical explosion-proof humanoid robot trajectory planning method disclosed in the first aspect when executing the program.

[0008] The beneficial effects of the present application are embodied in the following aspects: first, the correlation mechanism between petrochemical environment gas concentration and robot safe navigation is established, through real-time gas concentration monitoring and dangerous area demarcation, combined with static electricity accumulation tendency analysis and joint friction avoidance strategy, which helps to prevent the risk of ignition and explosion danger in the robot movement process, and improves the intrinsic safety level of robot operation in petrochemical environment. Second, a complete technical link from dangerous segment identification to poison-avoiding path generation is constructed, through path risk feature quantization and threat level evaluation, a standardized risk weight system is established, the accurate response of trajectory planning to different poison threats is realized, and through action primitive library construction and stability filtering, the execution stability and safety reliability of the avoidance action are guaranteed. Finally, by using three-dimensional decision space fusion and multi-level search technology, the safety, feasibility and execution of the three dimensions are unified into the decision framework, through dynamics envelope analysis and execution probability field construction, the deep integration of robot dynamics characteristics and environmental safety requirements is realized, and the final planning trajectory generated effectively takes into account the execution efficiency and action stability while ensuring high safety standards.

[0009] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0010] The drawings herein show specific examples of the technical solutions described in the present application, and constitute part of the specification together with the specific embodiments, for explaining the technical solutions, principles and effects of the present application.

[0011] Unless specifically stated or otherwise, the same reference signs in different drawings represent the same or similar technical features, and different reference signs may also be used to represent the same or similar technical features.

[0012] Figure 1 is a flowchart of a petrochemical explosion-proof humanoid robot trajectory planning method according to the present application.

[0013] Figure 2This is a structural block diagram of a petrochemical explosion-proof humanoid robot trajectory planning device according to the present invention.

[0014] Figure 3 This is a schematic diagram of the structure of a computer device according to the present invention. Detailed Implementation

[0015] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0016] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0017] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0018] The technical solutions of the embodiments of this application are described below.

[0019] like Figure 1 As shown, this embodiment of the invention provides a trajectory planning method for a petrochemical explosion-proof humanoid robot, including the following steps S110-S160: Step S110: Collect petrochemical environment gas concentration data, robot joint motion parameters and target position information, mark dangerous areas from gas concentration data to generate a restricted area map, determine safe nodes from the restricted area map, and determine the joint range of motion based on joint motion parameters.

[0020] Specifically, petrochemical environment gas concentration data, robot joint motion parameters and target position information are collected. The petrochemical environment gas concentration data collection is realized by deploying multiple types of gas sensors around the robot. The sensor configuration includes three main devices: combustible gas detector, toxic gas detector and oxygen concentration meter. The combustible gas detector uses catalytic combustion type sensor, the detection range is 0-100% LEL (lower explosive limit), the response time is less than 10 seconds, and the accuracy is ±3% LEL. The toxic gas detector is aimed at common petrochemical harmful gases such as hydrogen sulfide, carbon monoxide and benzene series, and uses electrochemical principle, with detection accuracy reaching ppm level. The oxygen concentration meter measures 0-25% Vol, with an accuracy of ±0.1% Vol, and is used to monitor the oxygen deficiency in a closed space. The sensor arrangement adopts a three-dimensional array method, with sensor nodes installed in front of the robot, on both sides and above, forming a 360-degree omnidirectional monitoring coverage. The data collection frequency is set to 1 Hz to ensure timely response to environmental changes. The gas concentration data preprocessing includes temperature compensation, humidity correction and baseline drift correction to eliminate the influence of environmental factors on measurement accuracy. The robot joint motion parameter collection is realized by built-in encoders and inertial measurement units, and the parameters include joint angle, angular velocity, angular acceleration and joint torque in four dimensions. The encoder resolution is 0.1°, and the sampling frequency is 100 Hz, providing high-precision position feedback. The inertial measurement unit integrates three-axis gyroscope and three-axis accelerometer to measure the dynamic characteristics of joint motion. The target position information is obtained by total station, GPS or indoor positioning label, with positioning accuracy better than 0.5 m, providing accurate spatial reference for path planning.

[0021] The gas concentration data is marked as a dangerous area to generate a forbidden map. A hierarchical standard is adopted to divide the area into four levels according to the gas concentration level: safe zone, warning zone, dangerous zone and forbidden zone. The safe zone is the area where all gas concentrations are below the safety threshold, the combustible gas concentration is less than 10% LEL, the toxic gas concentration is below the occupational exposure limit, and the oxygen concentration is in the range of 19.5%-23.5%. The warning zone is the area where part of the gas concentration is close to but does not exceed the dangerous threshold, which needs to be monitored but allows passage. The dangerous zone is the area where the gas concentration exceeds the safety limit, the combustible gas concentration is between 10%-25% LEL, or the toxic gas concentration exceeds the short-term exposure limit. The forbidden zone is the area where the gas concentration is seriously over-standard, the combustible gas concentration is greater than 25% LEL, or the toxic gas concentration reaches the level of immediate threat to life and health. The area marking algorithm uses the contour interpolation method to generate a continuous concentration distribution map based on discrete sensor measurement point data. The interpolation technique selects Kriging interpolation to establish a spatial autocorrelation model of gas concentration, and the interpolation accuracy is ensured by cross-validation to ensure reliability. The forbidden map generation adopts raster representation, with a map resolution of 1 m x 1 m, and each grid records the danger level and main danger factor at that location.

[0022] Safe nodes are determined from the no-go map. First, only the grid locations marked as safe zones are considered as candidate nodes based on the hazard level constraint. The spatial constraint requires that the 3m x 3m range around a node is all safe areas, ensuring that the robot has sufficient safety margin at this location. Connectivity analysis uses graph theory methods to model the safe areas as an undirected graph, with nodes representing safe grids and edges representing the connectivity between adjacent grids. Connectivity component identification is achieved through a depth-first search algorithm to determine each independent safe area block. Node importance evaluation considers the centrality, connectivity, and spatial distribution characteristics of the node. Centrality reflects the central position of the node in the connected graph, and connectivity represents the number of other nodes reachable from this node. Spatial distribution analysis ensures that safe nodes are evenly distributed in the map, avoiding excessive concentration or sparsity of nodes. Node optimization uses a greedy algorithm to select the optimal node set, with the objective function considering node coverage, connectivity, and distribution uniformity. Node labeling uses a hierarchical numbering scheme, with primary safe nodes labeled as S+ followed by a number, and secondary nodes labeled as SS+ followed by a number.

[0023] Joint range of motion is determined based on joint motion parameters. The joint range of motion is the allowed variation interval of each joint angle under the premise of ensuring work efficiency and safety. Parameter analysis first performs statistical analysis on historical joint motion data to calculate basic statistical quantities such as mean, variance, maximum, and minimum of each joint angle. Motion pattern recognition uses clustering analysis methods to divide joint motion into four typical patterns: walking pattern, operation pattern, obstacle avoidance pattern, and standby pattern. Each pattern corresponds to different joint activity characteristics and range requirements. Range of motion calculation uses probability statistics method, with 95% confidence interval as the normal range of joint motion: θ_range=[μ-1.96σ,μ+1.96σ], where μ is the mean of joint angle and σ is the standard deviation. Safety margin setting considers the special requirements of petrochemical environment, leaving a 20% safety margin based on the statistical range to avoid joint motion reaching the limit position. Joint coupling analysis studies the motion coordination relationship between different joints, and identifies strongly coupled joint pairs through correlation analysis. Workspace calculation is based on forward kinematics to determine the reachable space of the robot end effector within the given joint range. Range optimization uses multi-objective optimization method to minimize joint motion amplitude under the premise of meeting work requirements, reducing mechanical wear and energy consumption. The constraints include joint physical limits, collision avoidance, singularity avoidance, and multiple constraints.

[0024] In step S120, an initial path skeleton is constructed based on the target position information and the distribution of safe nodes, and a safety analysis is performed on the initial path skeleton to identify local dangerous sections. Motion constraints and posture balance requirements are generated based on the local dangerous sections.

[0025] Specifically, an initial path skeleton is constructed based on target position information and the distribution of safety nodes. The path search is based on a safety node connectivity graph, with safety nodes as graph vertices and safety channels between nodes as graph edges. The edge weight considers both distance cost and safety index. The search algorithm uses the A* algorithm, with a heuristic function combining Euclidean distance and risk assessment. The cost function is designed as: f(n) = g(n) + h(n) + s(n), where n is the search node, g(n) is the actual cost, h(n) is the heuristic cost, and s(n) is the safety cost. The safety cost is calculated based on the safety level of safety nodes, with lower levels resulting in higher costs. The path search process uses a bidirectional search strategy, starting from both the start and end points and converging at the intermediate position to form a complete path. The search space is limited to the safety area to avoid passing through dangerous and no-entry areas. Path smoothing uses spline curve fitting technology to eliminate jagged corners in the search path and generate a smooth trajectory suitable for robot motion. The initial path skeleton is represented by a key point sequence, including path start and end points, turning points, and important safety nodes. The skeleton data structure records the coordinates, arrival time, motion direction, and local safety level of each key point. Path segmentation divides long paths into multiple sub-segments, with each sub-segment length controlled within 50m for subsequent analysis.

[0026] The initial path skeleton is analyzed for safety to identify local dangerous segments. Based on the geometric characteristics of the initial path skeleton, the safety node level passed, and the path connection relationship, the path safety is evaluated. Path geometry analysis identifies sharp turns, long straight segments, and node-intensive areas, which may increase the risk of motion, lack alternative paths, or have path conflicts. Safety node level analysis evaluates the safety level distribution of the nodes passed along the path direction, identifying segments with low safety levels and areas with drastic level changes. Path connection relationship analysis checks the connection stability of the path with safety nodes, identifying fragile segments that are weakly connected or dependent on a single node. The risk value accumulation method is used to identify dangerous segments by weighting and summing the geometric risk, node risk, and connection risk: Risk = w1 x Geometry + w2 x NodeLevel + w3 x Connection, where Geometry is the geometric feature risk value, NodeLevel is the node level risk value, and Connection is the connection relationship risk value. The weight coefficients are set to w1 = 0.3, w2 = 0.5, and w3 = 0.2 according to path planning requirements. The risk threshold is set to the 80th percentile of the risk distribution, and segments exceeding the threshold are marked as local dangerous segments. Dangerous segment classification includes high-risk segments, medium-risk segments, and special-risk segments, with different categories corresponding to different processing strategies. Dangerous segment geometric feature extraction includes dangerous segment length, width, center position, and main risk type information.

[0027] In some embodiments, the generating the motion constraint condition and the posture balance requirement based on the local danger segment comprises: identifying an electrostatic accumulation tendency area from the local danger segment; establishing a charge flow path map based on the electrostatic accumulation tendency area; setting a joint friction avoidance point by using the charge flow path map; and establishing the motion constraint condition and the posture balance requirement according to the joint friction avoidance point.

[0028] An electrostatic accumulation tendency area is identified from the local danger segment. A local danger segment feature analysis method is adopted, and key parameters such as danger segment type, geometric complexity, risk level, and spatial distribution density are comprehensively considered. Danger segment type analysis divides the local danger segment into three categories according to risk nature, namely geometric risk type, node risk type, and connection risk type. The geometric risk type danger segment has a higher electrostatic accumulation tendency due to complex path twists, frequent robot motion changes, and increased friction contact. The node risk type danger segment has greater electrostatic accumulation safety hazards due to passing through low safety level nodes and small environmental safety margin. The connection risk type danger segment has unstable path connection and may need to frequently adjust the motion trajectory, increasing the electrostatic generation opportunity. Geometric complexity evaluation is based on the length, width, turning frequency, and angle change of the danger segment. The higher the complexity of the danger segment, the higher the electrostatic accumulation tendency. A straight danger segment with a length of more than 30 meters has a higher accumulation tendency due to continuous friction. A narrow danger segment with a width of less than 2 meters has an increased contact opportunity due to space limitations. A tortuous danger segment with a turning frequency of more than 3 times per 10 meters has increased friction due to motion complexity. Risk level mapping maps the high, medium, and special three risk levels of the local danger segment to the strong, medium, and weak three tendency levels of electrostatic accumulation. Spatial distribution density analysis identifies areas with dense distribution of danger segments. The electrostatic accumulation tendency is significantly enhanced in dense areas due to multiple risk superposition. Accumulation tendency evaluation adopts a weighted scoring method to comprehensively calculate the danger segment type, geometric complexity, risk level, and distribution density according to the weight proportion of 4:3:2:1. The scoring threshold is set to the 70th percentile of the scoring distribution. Local danger segment areas exceeding the threshold are marked as electrostatic accumulation tendency areas.

[0029] The charge flow path map is established based on the electrostatic accumulation prone area. The charge generation source identification includes robot joint friction, foot bottom contact with the ground, air relative motion of the fuselage, and other main sources. The generation rate is estimated by the triboelectric formula, considering factors such as contact area, relative speed, material properties, etc. The charge transmission path is described by a resistance network model, which divides the space into a resistance grid, and the resistance value of each grid is calculated according to the material resistivity and geometric size. The transmission direction is determined by the potential gradient, and the charge flows from the high potential area to the low potential area. The accumulation node identification is analyzed by charge balance, and the accumulation node is formed when the charge inflow rate is greater than the outflow rate. The dissipation path includes natural dissipation mechanisms such as leakage to the ground, air ionization, and corona discharge. The path map data structure uses a directed graph representation, with nodes representing charge accumulation points and edges representing charge flow paths, and edge weights representing resistance sizes. The resolution of the path map is set to the size of the spatial grid, generally 1m x 1m to ensure calculation accuracy.

[0030] For example, the joint friction avoidance point is set using the charge flow path map, including: extracting a high-risk charge convergence node from the charge flow path map; establishing a friction contact forbidden zone based on the high-risk charge convergence node; generating a joint motion offset point using the friction contact forbidden zone; and converting the joint motion offset point to a joint friction avoidance point.

[0031] The high-risk charge convergence node is extracted from the charge flow path map. The high-risk charge convergence node identification is based on charge density distribution and potential gradient analysis to determine the spatial position with the most serious charge accumulation and the highest discharge risk. First, calculate the charge density of each node in the path map, and the density calculation uses the charge balance equation: ∇·D=ρ, where D is the electric displacement vector and ρ is the charge density. The judgment criteria for high-risk nodes include three conditions: charge density exceeding the safety threshold, potential gradient greater than the critical value, and discharge probability higher than the risk limit. The charge density threshold is set according to the explosion-proof requirements of the petrochemical environment, generally taking the charge density corresponding to the minimum ignition energy. The potential gradient critical value is determined based on the air breakdown strength, which is about 3×10 6 V / m for dry air. The discharge probability is calculated by statistical methods, considering factors such as environmental humidity, temperature, and air pressure. The node risk level evaluation uses a fuzzy comprehensive evaluation method to convert quantitative indicators into qualitative risk levels. The high-risk nodes are sorted in descending order of comprehensive risk value, with the highest risk nodes being processed first. The node influence range is determined by electric field distribution calculation, and the space within the influence range needs special protection. Time duration analysis evaluates the time stability of high-risk nodes, with temporary protection for short-lived nodes and permanent protection measures for long-term nodes.

[0032] Frictional contact exclusion zones are established based on high-risk charge convergence nodes. These zones are spatial areas where robot joints should not engage in frictional contact to avoid electrostatic discharge and potential explosion risks. The exclusion zones are established using a safe distance envelope method, creating spherical or ellipsoidal spatial exclusion zones centered on the high-risk charge convergence nodes. The safe distance is calculated based on the electric field strength decay law and minimum safe distance requirements: d_safe = k × √(Q / E_threshold), where Q is the node charge, E_threshold is the safe electric field strength, and k is the safety factor. The shape of the exclusion zone is determined according to the electric field distribution characteristics; isotropic electric fields correspond to spherical exclusion zones, and anisotropic electric fields correspond to ellipsoidal exclusion zones. Exclusion zones are classified into three levels: absolute exclusion zone, restricted zone, and warning zone. Absolute exclusion zones strictly prohibit any contact; restricted zones allow contact under special protection; and warning zones require enhanced monitoring. The merging of exclusion zones for multiple high-risk nodes uses a union operation, with overlapping areas treated according to the strictest restriction level. The fuzzy processing of exclusion zone boundaries uses a gradient function to avoid abrupt constraints at the boundaries. The restricted area parameters include geometric descriptions such as center coordinates, radius or axis length, and orientation angle. The restricted area data structure records information such as center coordinates, geometric parameters, restriction level, and effective time. Visualization uses a 3D transparent volume, with different colors corresponding to different restriction levels.

[0033] Joint motion offset points are generated using friction contact restricted zones. These offset points are new target points set to avoid the friction contact restricted zones, based on the original motion trajectory. Geometric calculation methods are used to generate these offset points, finding the alternative trajectory with the smallest deviation from the original trajectory while satisfying the restricted zone avoidance constraints. The calculation objective is to minimize the weighted sum of squared trajectory deviations, with weights allocated according to joint importance and motion accuracy requirements. Constraints include restricted zone avoidance constraints, joint limit constraints, motion continuity constraints, and task completion constraints. Avoidance constraints are expressed as inequalities, requiring the distance between the offset point and the center of the restricted zone to be greater than a safe distance. The offset direction is selected using a gradient method, moving away from the center of the restricted zone. The offset magnitude is determined through a binary search to find the minimum offset amount that satisfies the constraints. Multi-joint coordinated offset considers the kinematic coupling between joints, using the Jacobian matrix to analyze the impact of joint motion on the end effector position. The timing of the offset points is synchronized with the overall motion planning to ensure smooth execution of the offset movements. Trajectory reconstruction uses spline interpolation to generate smooth motion trajectories between the offset points.

[0034] The joint motion offset point is converted into a joint friction avoidance point. First, the kinematics inverse solution calculation is performed to convert the offset point coordinates in Cartesian space into angle values in joint space. A numerical iteration method is used to solve the nonlinear equations by Newton-Raphson iteration. The multi-solution processing adopts the optimality criterion selection, which preferentially selects the solution that is close to the current joint angle and meets the physical constraints. The reachability analysis checks whether the inverse solution result is within the joint physical limit range, and the solution that exceeds the limit is marked as unreachable. Singularity detection is performed through the condition number analysis of the Jacobian matrix to avoid the robot entering a singular pose. The avoidance point precision evaluation is performed through forward kinematics calculation to determine the deviation of the end position corresponding to the inverse solution result from the target position. The avoidance point with unsatisfactory precision is improved through iteration to adjust the joint angle to improve the precision. The avoidance point time marker records the execution time and duration of each avoidance point to provide timing information for motion control. The avoidance strategy storage contains complete information such as the original trajectory, offset trajectory, avoidance point coordinates, and execution timing. The coordinate conversion uses the standard robot kinematics transformation matrix to ensure conversion accuracy and consistency.

[0035] Motion constraint conditions and posture balance requirements are established based on the joint friction avoidance points. The establishment of motion constraint conditions adopts a hierarchical constraint structure, including two levels of hard constraints and soft constraints. Hard constraints are safety requirements that must be strictly met, including avoidance point position constraints, velocity limit constraints, and acceleration limit constraints. The avoidance point position constraint requires that the joint motion strictly follows the avoidance trajectory, with a position deviation of no more than ±2 cm. The velocity limit constraint limits the joint angular velocity within a safe range, with an angular velocity in the avoidance region not exceeding 50% of the normal value. The acceleration limit constraint prevents sharp motion changes, with an angular acceleration limit of 30% of the normal value. Soft constraints are performance requirements, including motion smoothness constraints, energy consumption control constraints, and task efficiency constraints. Posture balance requirements consider the impact of avoidance actions on the overall stability of the robot, and develop corresponding center of gravity control and support strategies. Center of gravity offset compensation is achieved by adjusting the posture of non-avoidance joints to maintain the overall center of gravity within the support polygon. Support strategies include increasing the support base, reducing the center of gravity height, and adjusting gait parameters. Balance control parameters include key indicators such as zero-moment point position, stability margin, and posture angle. The mathematical representation of constraint conditions uses a combination of linear and nonlinear inequalities, and the feasible motion control parameters are determined through constraint solving algorithms. The constraint priority setting places safety constraints at the highest priority, balance constraints at the second priority, and performance constraints at the lowest priority.

[0036] In step S130, based on the initial path skeleton, a major hazard segment is identified, a threat level assessment is performed on the major hazard segment to generate a risk weight, and a path re-planning is performed using the risk weight to generate a poison avoidance path.

[0037] Specifically, the major dangerous segment is identified based on the initial path skeleton. The major dangerous segment is the key segment with the highest threat level and the widest impact range in the local dangerous segment. Along the trajectory of the initial path skeleton, the major dangerous segment is identified based on the spatial distribution characteristics of the path skeleton and the safety node level information. The major dangerous segment is a segment on the path that simultaneously meets multiple high-risk conditions and has a threat level exceeding the safety threshold, which poses a serious safety threat to the robot and the surrounding environment. The identification method uses a multi-factor comprehensive analysis, mainly considering four key factors: path safety level mutation, path geometric risk concentration, path connection vulnerability, and path proximity to forbidden areas. The safety level mutation analysis identifies segments with a sharp drop in safety node level on the path, and segments with a level drop exceeding 2 levels are marked as high-risk. The path geometric risk concentration analysis identifies dangerous segments with dense sharp turns and long distances without alternative paths, and segments with turn angles greater than 90 degrees and turn spacings less than 20 meters are considered geometrically high-risk areas. The connection vulnerability analysis identifies segments that rely on a single connection or unstable connections, and segments with a connection degree less than 2 or a connection weight less than 50% of the average value are considered connection vulnerable areas. The forbidden area proximity analysis identifies segments of the path that are too close to the dangerous areas in the S110 identified forbidden map, and segments with a distance to the forbidden area boundary less than 10 meters are considered close to high-risk areas. The threat concentration degree calculation uses spatial overlay analysis to overlay multiple path risk factors in space, and areas with an overlay value exceeding the set threshold are considered major dangerous segments. The dangerous segment feature extraction focuses on the maximum threat intensity and the main path risk type.

[0038] In some embodiments, the threat level assessment of the major dangerous segment generates a risk weight, including: determining a quantitative scale based on the major dangerous segment assessment of chemical hazard intensity; setting a dangerous degree calibration coefficient according to the quantitative scale; and using the dangerous degree calibration coefficient to convert the threat level of the major dangerous segment to generate a risk weight.

[0039] A quantitative scale was established to determine the intensity of chemical hazards based on the criticality of hazardous segments. Four key factors of criticality were analyzed, including the severity of safety level mutation, the concentration of geometric risk, the intensity of connection vulnerability, and the proximity of forbidden areas. The severity of safety level mutation was evaluated based on the drop in node level and the frequency of mutation. A road segment was marked as extremely high mutation if the drop in node level was more than 3 levels, high mutation if the drop was between 2 and 3 levels, medium mutation if the drop was between 1 and 2 levels, and low mutation if the drop was less than 1 level. The concentration of geometric risk was evaluated considering the density of sharp turns and the complexity of the path. A road segment was marked as extremely high geometric risk if the density of sharp turns was more than 5 per 100 meters and the angle of the turn was more than 120 degrees. The intensity of connection vulnerability was evaluated based on the redundancy of connections and the distribution of connection weights. A road segment was marked as extremely high vulnerability if the degree of connection was 1 and the weight was less than 70% of the average value. The proximity of forbidden areas was evaluated based on the nearest distance to the forbidden area and the impact range. A road segment was marked as extremely high proximity risk if the distance was less than 5 meters. The quantitative scale was established using a normalization method to convert different levels of path risk data into a unified quantitative scale. The scale range was set to 1-10, with 1 representing the lowest hazard and 10 representing the highest hazard. The quantitative conversion used a piecewise linear function, with different conversion weights corresponding to different risk factors. The multi-factor synthesis used a weighted arithmetic mean method, with weights assigned based on the importance of the factors: safety level mutation 40%, geometric risk 25%, connection vulnerability 20%, and proximity 15%.

[0040] A calibration coefficient for the degree of danger was set according to the quantitative scale. The calibration coefficient K_calibration was used to adjust the weight proportion of path hazard intensity in risk assessment, establishing an accurate mapping relationship between the quantitative scale and the actual risk level. The calibration coefficient was set based on the numerical distribution characteristics and statistical laws of the quantitative scale, with a piecewise function used to determine the calibration coefficient corresponding to different quantitative scale intervals. The calibration coefficient for extremely high hazard path segments (quantitative scale 8-10) was set to 1.2-1.5, reflecting their extraordinary danger. The calibration coefficient for high hazard path segments (quantitative scale 6-8) was 1.0-1.2, the coefficient for medium hazard path segments (quantitative scale 4-6) was 0.8-1.0. The coefficient for low hazard path segments (quantitative scale 2-4) was 0.6-0.8, and the coefficient for micro-hazard path segments (quantitative scale 0-2) was 0.4-0.6. The coefficient distribution used a linear interpolation method to ensure smooth transition between different quantitative scale intervals. The stability of the calibration coefficient was evaluated by variance analysis, and coefficients with a variance less than 0.1 were considered stable and usable.

[0041] The risk weight is generated by converting the threat level of the major hazard segment using the calibration coefficient of the hazard degree. The threat level conversion converts the qualitative hazard description and quantitative hazard degree index into a standardized risk weight value. A multi-step calculation method is adopted, which first calculates the basic risk value and then adjusts it using the calibration coefficient. The basic risk value R_base is calculated based on the comprehensive characteristics of the major hazard segment: R_base = W1 x S_level + W2 x G_risk + W3 x C_weak + W4 x P_close, where S_level is the safety level mutation value, G_risk is the geometric risk value, C_weak is the connection vulnerability value, P_close is the proximity value, and the weight coefficients W1-W4 are consistent with the weights in the quantitative scale calculation. The adjusted risk value R_adjusted is obtained by modifying the calibration coefficient: R_adjusted = R_base x K_calibration, where K_calibration is the calibration coefficient of the hazard degree. The threat level classification is based on the adjusted risk value, and a five-level classification standard is adopted: extremely high threat (R≥8.0), high threat (6.0≤R<8.0), medium threat (4.0≤R<6.0), low threat (2.0≤R<4.0), and micro threat (R<2.0). The risk weight W is generated by normalization processing, which maps the risk value to the 0-1 interval: W = R_adjusted / R_max, where R_max is the maximum risk value of the system.

[0042] In some embodiments, the path re-planning using the risk weight generates a poison avoidance path, including: constructing a threat avoidance area and a safety guidance area based on the risk weight; generating a path repulsion factor using the threat avoidance area, the path repulsion factor including poison repulsion and explosion repulsion; and path fusion of the path repulsion factor and the safety guidance area to generate a poison avoidance path.

[0043] The threat avoidance region and the safety guidance region are constructed based on the risk weight. The construction of the threat avoidance region and the safety guidance region provides spatial constraints and guidance information for path re-planning, realizes effective avoidance of dangerous areas and full use of safe areas. The threat avoidance region is a spatial region whose risk weight exceeds a set threshold and needs to be avoided by the robot. The region division adopts the risk weight contour method to convert the continuous risk weight distribution into discrete region labels. The avoidance region threshold is set to the 75th percentile of the risk weight, ensuring that the high-risk region is effectively identified. The region expansion process considers the physical size and safety margin of the robot, and expands 2 meters outward from the original high-risk region as a safety buffer. The safety guidance region is a spatial region with low risk weight and suitable for robot passage, which is the preferred channel for path planning. The guidance region identification is based on the reverse analysis of the risk weight. The region whose risk weight is lower than the 25th percentile is marked as the safety guidance region. The region connectivity analysis adopts the graph theory method to identify the connected safety region block, providing a feasible channel network for path search. The spatial distribution characteristics of the threat avoidance region and the safety guidance region affect the effectiveness of subsequent path planning, and the accuracy of the region boundary determines the safety of the avoidance path.

[0044] The path repulsion factor is generated using the threat avoidance region. The poison repulsion degree calculation considers the area size and shape complexity of the threat avoidance region. The larger the area and the more irregular the shape, the higher the repulsion degree. The repulsion degree calculation uses a distance decay function. The closer to the center of the threat avoidance region, the higher the repulsion degree. The region density analysis considers the number of threat avoidance regions per unit area, and the high-density region increases the repulsion degree weight. The explosive repulsion degree calculation considers the spatial distribution pattern and connectivity characteristics of the threat avoidance region. The large-area avoidance region with strong connectivity has a higher explosive repulsion degree. The region boundary complexity analysis evaluates the tortuosity of the avoidance region boundary. The region with complex boundary increases the navigation difficulty. The spatial gradient analysis calculates the risk weight gradient of the threat avoidance region boundary. The region with a large gradient indicates a sharp change in risk. The spatial distribution of the repulsion factor uses a distance decay model. The closer to the threat center, the larger the repulsion factor. The repulsion factor standardization process unifies the numerical range to the 0-10 interval, which is convenient for comprehensive calculation with other path planning parameters.

[0045] The path repulsion factor is fused with the safety-oriented area to generate a poison-avoiding path. The path fusion algorithm considers the blocking effect of the repulsion factor and the guiding effect of the safety-oriented area to generate an optimal path that can effectively avoid danger and efficiently reach the target. The fusion calculation uses the potential field method principle to convert the repulsion factor into a repulsion potential field and the safety-oriented area into an attractive potential field. The strength of the repulsion potential field is proportional to the value of the repulsion factor, and a strong repulsion potential is formed in the high repulsion factor area. The strength of the attractive potential field is related to the safety level and the distance from the target, and a strong attractive potential is generated in the area with high safety level and close distance to the target. The vector synthesis method is used for potential field superposition, and the vector sum of the repulsion potential and the attractive potential is calculated at each spatial position. The path search is performed along the gradient direction of the synthesized potential field, automatically avoiding high potential energy areas (dangerous areas) and tending to low potential energy areas (safe areas). The path smoothing processing uses spline curve fitting technology to eliminate the path discontinuity caused by the gradient change of the potential field. The path feasibility check includes geometric constraint check, kinematics constraint check and safety constraint check. The geometric constraint ensures that the path does not pass through physical obstacles, and the kinematics constraint ensures that the robot can perform path movement. The quality evaluation of the poison-avoiding path uses a multi-objective evaluation method, which considers the path length, safety level, travel time and energy consumption level. The path data output includes path coordinate sequence, risk level distribution, recommended travel speed and safety protection requirements.

[0046] In step S140, a motion primitive library is generated based on the motion constraint conditions and the joint activity range, the motion primitive library is filtered for stability according to the posture balance requirement to generate stable motion primitives, and the stable motion primitives are woven between the safety nodes to form a set of marching schemes.

[0047] Specifically, the motion primitive library is generated based on the motion constraints and joint range of motion. Motion primitive is the minimum unit of motion for robots to complete specific functions, with characteristics such as independence, combinability and repeatability. A constraint-oriented generation method is used to construct the primitive library. According to the type of motion constraint, the motion primitives are divided into four categories: avoidance constraint type, speed limit type, acceleration limit type and position constraint type. The avoidance constraint type primitive is generated based on the avoidance point position constraint, which ensures that the joint motion strictly follows the avoidance trajectory and the position deviation does not exceed the specified range. The speed limit type primitive is generated based on the angular velocity limit constraint, which limits the angular velocity in the avoidance area to no more than a specified proportion of the normal value. The acceleration limit type primitive is generated based on the acceleration limit constraint, which prevents sharp motion changes and limits the angular acceleration within a safe range. The position constraint type primitive is generated based on the joint range of motion constraint, which ensures that all joint motions are within the determined effective range. The primitive parameterization representation uses a joint space description method, and each primitive contains parameters such as joint angle sequence, angular velocity constraint, angular acceleration limit and execution time. Constraint condition mapping applies hard constraints and soft constraints to the primitive generation process, with hard constraints as boundary conditions that must be met and soft constraints as optimization targets. The joint range of motion check uses the aforementioned θ_range parameter, and actions that exceed the range are automatically excluded. The primitive combination rule establishes the connection method and conversion conditions between different primitives based on constraint compatibility, supporting the construction of complex action sequences. The primitive library data structure is organized by constraint classification, with the top layer being the constraint type, the middle layer being the constraint strength, and the bottom layer being the specific motion primitive. Each primitive contains complete information such as constraint source, parameter list, execution condition and compatible primitive list.

[0048] In some embodiments, the stability filtering of the motion primitive library according to the posture balance requirement generates stable motion primitives, including: stability grading of the motion primitive library according to the posture balance requirement to form a stability gradient; extracting stability peak points and stability valley points in the stability gradient; partitioning and constructing based on the stability peak points and the stability valley points as boundaries to generate a motion stability system; and performing optimal judgment on the motion stability system to generate stable motion primitives.

[0049] The stability gradient is formed by ranking the motion primitives in the library according to the stability requirements of the posture balancing demand. The stability evaluation criteria are established based on the key indicators in the posture balancing demand, such as the zero moment point position, stability margin, and posture angle. The stability evaluation first converts the posture balancing demand into quantifiable scoring criteria. The zero moment point position evaluation is based on the zero moment point position requirement in the posture balancing demand and uses ZMP trajectory deviation to calculate the score. The stability margin evaluation is based on the stability margin parameter in the posture balancing demand and uses the ratio of the actual margin to the required margin to calculate the score. The posture angle evaluation is based on the posture angle limit in the posture balancing demand and uses the ratio of the angle deviation to the limit range to calculate the score. The comprehensive stability score uses the weight distribution method in the posture balancing demand: S_total = W_zmp x S_zmp + W_margin x S_margin + W_posture x S_posture, where the weight coefficients W_zmp, W_margin, and W_posture come from the parameter settings in the posture balancing demand. The stability score of each motion primitive is calculated, and the ZMP trajectory, stability margin change, and posture angle sequence during the execution of the primitive are obtained through simulation or experimental data. The scoring results are standardized according to the evaluation criteria of the posture balancing demand, making it easy to compare and rank. The sorting algorithm arranges all primitives from high to low according to the stability score. The hierarchical processing divides the sorted primitives into corresponding stability levels based on the hierarchical criteria of the posture balancing demand. The gradient construction forms a smooth stability change curve through the continuous score distribution, reflecting the overall stability characteristics of the primitive library. Primitives with the same score are further sorted based on secondary indicators in the posture balancing demand.

[0050] The stability peak points and stability valley points are extracted in the stability gradient. The peak point is the position of the highest local stability primitive, corresponding to the local maximum point of the gradient curve. The valley point is the position of the lowest local stability primitive, corresponding to the local minimum point of the gradient curve. The extraction algorithm uses the first derivative analysis method to calculate the derivative sequence of the stability gradient, and the position of the sign change of the derivative corresponds to the peak point or valley point. The peak identification condition is the position where the derivative changes from positive to negative, and the valley identification condition is the position where the derivative changes from negative to positive. The significance screening retains only significant peak points and valley points by setting a threshold condition, and eliminates minor fluctuations caused by noise. The threshold is set based on the statistical analysis of the gradient change amplitude, taking the 90th percentile of the change amplitude distribution as the significance threshold. The peak-valley correspondence analysis establishes the pairing relationship between adjacent peak points and valley points, and each peak-valley pair corresponds to a stability change interval. The multi-peak and multi-valley processing considers the complex situation where multiple peak points and valley points exist in the gradient, and uses clustering analysis to merge similar peak and valley points.

[0051] The stability peak point and the stability valley point are used as boundaries to construct the action stability system. The peak point and the valley point are used as the main boundary line, and the continuity of the stability score and the rationality of the interval are considered. The threshold segmentation and boundary adjustment strategy are used in the partition method. First, the initial segmentation is based on the peak and valley points, and then the boundary position is adjusted according to the stability distribution characteristics. The interval naming uses the stability level identifier, such as high stability zone, medium-high stability zone, medium stability zone, medium-low stability zone, and low stability zone. The number of primitives contained in each partition is controlled within a reasonable range to avoid affecting the balance of the system. The partition feature analysis calculates the stability mean, variance, primitive number, and typical representative of each partition. The interval boundary optimization adjusts the boundary position to maximize the consistency of the stability within the interval through boundary sensitivity analysis. The stability system establishment includes the partition hierarchy, interval boundary, feature parameter, and application rule, etc. The partition standardization establishes a unified partition standard and naming specification to support the application of stability system in different scenarios.

[0052] The stability action primitives are generated by optimizing the action stability system. Based on the partition structure and stability evaluation results of the action stability system, the selection and judgment are carried out. A multi-level selection mechanism is adopted, which first selects the partition with higher evaluation level in the stability system, and then selects the primitives from the selected partition according to the stability score. The partition selection standard is based on the partition level of the stability system, and the primitives in the high stability zone and the medium-high stability zone are preferred. The primitive selection uses the score ranking of the stability system to select the primitives with high scores. The constraint compatibility check ensures that the selected primitives are consistent with the original constraint conditions, and maintains the constraint integrity of the primitive library. The repeated primitive elimination process eliminates the primitives with similar functions in the stability system, and the optimal version is retained by comparing the stability scores. The optimization algorithm is based on the hierarchical structure of the stability system, and the primitives with high stability and high partition level are preferred. The number of selected primitives is determined according to the distribution characteristics of the stability system, and the reasonable proportion of primitives at each level is maintained. The optimization result record contains the selected primitive list, stability level, partition attribution, and score information. The stable action primitive database establishment contains the complete archives of primitive parameters, stability information, and system attribution, etc.

[0053] The stable action primitives are used to weave paths between safety nodes to form a set of travel plans. The path weaving is based on the constraint characteristics of stable action primitives and the spatial distribution characteristics of safety nodes. A method combining graph search and action sequence planning is used, with safety nodes as key control points of the path and stable action primitives as the connection method between nodes. The weaving algorithm first establishes the reachability analysis between nodes, based on the movement range and execution characteristics of stable action primitives, to determine which safety nodes can be connected by existing stable primitives. The connection feasibility evaluation is based on the constraint type of the action primitive and the safety level of the safety node, and only the connection with compatible constraints and matching safety levels is considered feasible. The multi-primitive combination processing is based on the constraint compatibility of the primitives, and when a single primitive cannot complete the connection, a sequence combination of constraint-compatible primitives is used to achieve it. The sequence connection rule is based on the constraint continuity of the primitives, to ensure the continuity of adjacent primitives in terms of constraint satisfaction, state transition, etc. The path weaving uses a depth-first search algorithm, starting from the starting safety node and gradually expanding to adjacent nodes until the target node is reached. The branch processing strategy retains multiple branches when multiple optional paths are encountered, generating diversified travel plans. After weaving, a set of travel plans containing all feasible paths is generated, with each plan recording the complete sequence of action primitives and node paths. The number of plan sets is determined based on the number of actual reachable paths, and contains all feasible plans that meet the constraint conditions from the starting node to the target node. The plan data structure records the action primitive sequence, node path, constraint satisfaction degree, and connection relationship.

[0054] In step S150, the joint load peak value in the travel plan set is obtained, the dynamics envelope is constructed based on the joint load peak value, the execution adaptability is determined according to the dynamics envelope, and the execution probability field is formed by projecting the execution adaptability in the spatial coordinates.

[0055] The joint load peaks in the travel scheme set are obtained. The joint load peak is the maximum moment or maximum power consumption point that the joint bears during the execution of the action sequence. The load calculation adopts an inverse dynamics method to calculate the driving moment required by each joint based on the dynamics model of the robot and the motion parameters. The calculation formula is based on the Lagrange equation: τ = M(q) × q_ddot + C(q, q_dot) × q_dot + G(q), where τ is the joint torque vector, M is the inertia matrix, C is the Coriolis matrix, G is the gravity term, q is the joint angle, q_dot is the joint angular velocity, and q_ddot is the joint angular acceleration. The node path provides the attitude information of the robot at different spatial positions, affecting the calculation of the gravity term G(q) and the inertia matrix M(q). The constraint satisfaction degree information determines the constraint boundary conditions during the execution of each scheme, affecting the calculation range and safety margin evaluation of the joint moment. The connection relationship information ensures the motion continuity between adjacent action primitives, ensures the smooth transition of velocity and acceleration at the primitive connection, and avoids false load peaks caused by motion mutation. Load peak identification is achieved by peak detection on time series moment data, and a local maximum value search algorithm is used to locate the peak point. Peak screening sets threshold conditions to only retain significant peaks that are 1.5 times the average load. Multi-scheme load analysis is based on the complete data structure of each scheme, and the action primitive sequence is extracted one by one for dynamics modeling. The load distribution is calculated in combination with the spatial position of the node path, the influence of the constraint satisfaction degree on the load boundary is considered, the continuity of the calculation is ensured by using the connection relationship, and an accurate correspondence between the scheme and the load peak is established.

[0056] In some embodiments, constructing a dynamics envelope based on the joint load peak includes: performing torque distribution identification based on the joint load peak to generate a high torque concentration band and a low torque dispersion band; performing load balancing adjustment on the low torque dispersion band using the high torque concentration band to form an equalized torque band; performing spatial envelope recombination on the equalized torque band to generate a recombined envelope unit; and performing contour locking on the recombined envelope unit to construct a dynamics envelope.

[0057] The torque distribution identification based on joint load peaks generates high torque concentration zones and low torque dispersion zones. First, the torque distribution density function is established, and the kernel density estimation method is used to calculate the distribution density of torque on the joint space and time axis. The high torque concentration zone is the area where the torque density is more than twice the average density, representing the space or time period with high torque load concentration and intensity. The low torque dispersion zone is the area where the torque density is less than half the average density, representing the area with low torque load dispersion and intensity. The distribution identification uses clustering analysis method to classify similar torque characteristics of load peaks into the same concentration zone or dispersion zone. The multi-joint collaborative analysis considers the mutual influence of torque distribution in different joints, and identifies the torque transmission and compensation relationship between joints. The distribution boundary determination uses threshold segmentation method to determine the boundary of concentration zone and dispersion zone according to the change rate of torque density. The spatiotemporal distribution mapping establishes a two-dimensional mapping of torque distribution in time and space dimensions, supporting multi-dimensional distribution analysis.

[0058] The load balancing deployment of low torque dispersion zone by high torque concentration zone forms balanced torque zone. Load balancing deployment uses torque redistribution technology to transfer part of the load of high torque concentration zone to low torque dispersion zone, achieving spatial and temporal balance of torque load. The deployment principle is based on the characteristics of robot redundant degrees of freedom, and realizes load redistribution by adjusting joint motion coordination relationship. The deployment algorithm uses constrained least squares method to minimize the unevenness of torque distribution under the premise of meeting kinematic constraints. The objective function is designed to minimize torque variance, and the constraint conditions include joint limit constraint, speed constraint and task completion constraint. The load decomposition of high torque concentration zone uses principal component analysis method to decompose the composite torque into multiple independent components. The load transfer strategy determines the transfer amount and transfer direction according to the carrying capacity of low torque dispersion zone. The transfer path planning establishes the load transfer path from concentration zone to dispersion zone to ensure the smoothness of the transfer process. The balance degree evaluation is calculated by the statistical characteristics of torque distribution, including variance, skewness, kurtosis and other indicators. The deployment effect is measured by the improvement degree of balance degree, and the greater the improvement degree, the better the deployment effect. Multi-round deployment handles complex load distribution, and gradually improves torque balance through iterative deployment.

[0059] The uniform torque band is spatially enveloped to generate recombined envelope units. A spatial segmentation and merging strategy is adopted. Firstly, the torque band is spatially segmented according to geometric characteristics. Then, segments with similar characteristics are merged to form envelope units. The segmentation criterion is based on torque gradient and geometric continuity. Locations with dramatic gradient changes are taken as segmentation boundaries. The merging criterion considers the shape regularity and functional integrity of envelope units. Units with similar shapes and complementary functions are preferentially merged. The geometric shape of the envelope unit is described by regular polyhedrons, including cuboids, cylinders, ellipsoids, and other basic geometric bodies. Shape selection is based on the geometric characteristics of torque distribution. Slender distribution corresponds to a cylinder, and compact distribution corresponds to an ellipsoid. The size of the unit is determined by the statistical characteristics of the torque distribution. The envelope volume is proportional to the total torque. The recombination process preserves the basic characteristics of the torque distribution, ensuring that the recombined envelope units accurately reflect the original torque characteristics. The connection relationship between units is established through adjacency analysis. The interface connection mode and transmission characteristics between adjacent units are determined.

[0060] The recombined envelope units are profile-locked to construct the dynamic envelope. The discrete recombined envelope units are connected to form a continuous dynamic envelope boundary through boundary extraction and surface fitting methods. First, the outer profile of each envelope unit is extracted, and then the profiles of adjacent units are connected to form the overall envelope through surface fitting technology. The profile extraction uses the isosurface algorithm to generate an isotorque surface as the profile line at the envelope unit boundary. The surface fitting selects B-spline surfaces to achieve smooth connection between different units through control points and weights adjustment. Boundary continuity check ensures the geometric continuity and torque continuity of adjacent unit profiles at the connection. Envelope closure processing identifies open boundaries of the envelope through topological analysis and realizes the closure of the envelope using surface extension technology. The geometric characteristics of the dynamic envelope include the envelope volume and the length of the principal axis, which will be used for subsequent stability margin calculation. The mathematical representation of the envelope boundary adopts a parametric equation or an implicit function form, which supports the analytical calculation of the envelope.

[0061] The adaptability is determined according to the dynamic envelope. The adaptability is a quantitative indicator of the matching degree of the travel scheme and the dynamic envelope. The adaptability evaluation mainly considers four key dimensions: envelope geometric features, envelope distribution uniformity, envelope boundary stability and envelope volume efficiency. The envelope geometric feature evaluation is based on the shape regularity and symmetry of the dynamic envelope. The more regular the shape and the better the symmetry, the more reasonable the load distribution. The envelope distribution uniformity is calculated by the variance of the torque distribution in the envelope. The smaller the variance, the more uniform the load distribution. The envelope boundary stability is evaluated based on the curvature change of the envelope boundary. The smoother the boundary, the more stable the load transfer. The envelope volume efficiency is calculated by the ratio of the envelope volume to the total load. The smaller the ratio, the higher the load concentration and the better the efficiency. The comprehensive score of the adaptability Adaptability is calculated using the weighted average method: Adaptability = w1 x Geometry + w2 x Uniformity + w3 x Boundary + w4 x Volume, where Geometry is the geometric feature score, Uniformity is the distribution uniformity score, Boundary is the boundary stability score, and Volume is the volume efficiency score. The weight coefficients are assigned according to the importance of the dynamic envelope. The score is standardized to map the adaptability to the 0-1 interval, which facilitates comparison between different schemes. The threshold determination sets the adaptability acceptance threshold. The scheme higher than the threshold is marked as an executable scheme. The multi-scheme sorting arranges the schemes from high to low according to the adaptability score, providing priority reference for scheme selection.

[0062] The execution adaptability is projected into a spatial coordinate to form an execution probability field. The execution adaptability is mapped from the scheme space to the physical space through spatial projection technology to construct a continuous field distribution reflecting the execution probability of different spatial positions. The spatial coordinate projection establishes a mapping relationship between the execution adaptability and the robot workspace position, and each spatial point corresponds to an execution probability value. The projection method uses radial basis function interpolation to extend the discrete adaptability data to a continuous spatial distribution. The interpolation kernel function selects a Gaussian function, and the kernel parameters are adaptively adjusted according to the data distribution characteristics. The execution probability P_exec is calculated based on the normalization of the adaptability: P_exec = Adaptability / Σ(Adaptability), which ensures that the probability value meets the probability distribution requirements. The field distribution smoothing process uses spatial filtering technology to eliminate discontinuous points and outliers in the interpolation process. The probability field gradient calculation identifies the area with the most drastic change in execution probability, providing spatial guidance information for path planning. The multi-scale field representation supports probability field analysis at different resolutions. Coarse scale is used for global planning, and fine scale is used for local path adjustment. The field data structure uses a regular grid to store the coordinates and probability values of each grid point.

[0063] At step S160, a three-dimensional decision space is established based on the poison-avoiding path, the set of travel schemes, and the execution probability field configuration, an optimal convergence point is found in the three-dimensional decision space, and a final planning trajectory is generated based on the optimal convergence point.

[0064] Specifically, a three-dimensional decision space is established based on the poison-avoiding path, the set of travel schemes, and the execution probability field configuration. The three-dimensional decision space is a decision analysis framework containing three dimensions of safety, feasibility, and execution, and provides comprehensive decision support for robot path planning. The space is constructed using a hierarchical architecture, with the bottom layer being physical space coordinates, the middle layer being decision attributes, and the top layer being evaluation indexes. The physical space uses a three-dimensional rectangular coordinate system, covering the entire working area of the robot, and the spatial resolution is set to 0.5m×0.5m×0.2m to ensure decision accuracy. The safety dimension is based on the risk assessment results of the poison-avoiding path, and each position in the space is marked with the corresponding safety level and risk weight. The feasibility dimension is based on the action primitive distribution of the set of travel schemes, and identifies the types of actions that can be executed and the execution difficulty at different positions. The execution dimension is based on the probability distribution of the execution probability field, reflecting the execution success rate of the robot at each position. Data fusion uses tensor operation methods to combine the information of the three dimensions to form a four-dimensional decision tensor: D(x,y,z,t)=Safety(x,y,z)⊗Feasibility(x,y,z)⊗Execution(x,y,z), where D(x,y,z,t) is the four-dimensional decision tensor, x, y, z are the physical space coordinates, t is the time coordinate, Safety(x,y,z) is the safety value, Feasibility(x,y,z) is the feasibility value, and Execution(x,y,z) is the execution value, and ⊗ represents tensor product operation. Space discretization divides the continuous decision space into regular grids, and each grid point records the decision attributes and evaluation indexes. Attribute weight allocation is determined according to task priority, with safety weight 0.5, feasibility weight 0.3, and execution weight 0.2. Space constraints include boundary constraints, obstacle constraints, and dangerous area constraints, ensuring the physical reasonableness of the decision space.

[0065] In some embodiments, finding the optimal convergence point in the three-dimensional decision space includes: determining a search depth based on the safety convergence difficulty of the three-dimensional decision space, the safety convergence difficulty including safety distribution density, feasibility space connectivity, and execution probability gradient; determining a priority configuration according to the search depth; formulating a hierarchical search execution scheme using the priority configuration; and generating the optimal convergence point through the hierarchical search execution scheme.

[0066] Based on the three-dimensional decision space, the search difficulty is evaluated and the search depth is determined. The convergence difficulty of search is evaluated based on the distribution characteristics and spatial complexity of the three dimensions in the three-dimensional decision space. The safety convergence difficulty evaluation establishes a quantitative index of search complexity, providing a basis for parameter setting of search algorithm. The evaluation method uses decision space feature analysis, mainly considering three key factors: safety distribution density, feasibility space connectivity and execution probability gradient. The safety distribution density is calculated based on the variance and extreme value distribution of the safety dimension in the three-dimensional decision space. The convergence difficulty is high in the area with drastic density changes. The feasibility space connectivity is analyzed by the number and size of the connected regions in the feasibility dimension of the decision space. The search is difficult in the area with low connectivity. The execution probability gradient is based on the gradient distribution of the execution probability field in the decision space. The area with large gradient changes needs fine search. The comprehensive difficulty evaluation uses a weighted scoring method: Difficulty=α×Safety_density+β×Feasibility_connectivity+γ×Execution_gradient, where Safety_density is the safety distribution density, Feasibility_connectivity is the feasibility connectivity, Execution_gradient is the execution gradient, and the weight coefficients are α=0.4, β=0.3, γ=0.3, respectively. The search depth determination is based on the mapping relationship of difficulty level. High difficulty corresponds to deep search, and low difficulty corresponds to shallow search. The depth level is divided into five levels: extremely deep search (difficulty 0.8-1.0), deep search (difficulty 0.6-0.8), moderate search (difficulty 0.4-0.6), shallow search (difficulty 0.2-0.4), and extremely shallow search (difficulty 0-0.2). The search depth parameters include search layer number, grid density, iteration number, and convergence precision, etc.

[0067] The priority configuration is determined according to the search depth. Different priority strategies are formulated according to the search depth level. In the extreme depth search mode, the safety priority is set to the highest, and the search algorithm gives priority to safety constraints, followed by feasibility and execution. In the deep search mode, safety and feasibility are equally important, and both are ensured by balancing the weights. In the moderate search mode, an equal configuration is adopted, and the weights of the three target dimensions are equal. In the shallow search mode, execution is focused on, and execution efficiency is pursued under the premise of ensuring basic safety. In the extremely shallow search mode, fast convergence is the goal, and the search speed is reduced in exchange for precision requirements. The priority configuration parameters include target weight vector, constraint importance ranking, search resource allocation, and termination condition setting. Resource allocation considers three aspects of calculation time, storage space, and search range, and high-priority tasks obtain more resources. The dynamic priority adjustment mechanism adjusts the priority configuration according to the search progress and intermediate results, improving the adaptability of the search. The effectiveness of the configuration is evaluated through statistical analysis of historical search results, and the configuration with a high success rate obtains a higher credibility. The priority configuration data structure records configuration parameters, applicable conditions, performance indicators, and usage history, etc.

[0068] A hierarchical search execution scheme is formulated using priority configuration. A hierarchical design approach is adopted to decompose complex search problems into multiple relatively simple sub-problems. The search hierarchy is divided into three levels: global search layer, regional search layer, and local search layer. The global search layer performs coarse-grained search in the entire decision space to identify potential excellent regions. The regional search layer performs medium-precision search in the identified excellent regions to narrow the search range. The local search layer performs high-precision search in the most promising local region to determine the precise convergence point position. Inter-layer information transfer establishes a data flow from the upper layer to the lower layer to ensure the continuity of the search process. The search strategy configuration sets the search algorithm, parameter setting, and termination condition for each layer according to the priority. The execution order adopts a serial execution mode, and each layer search is executed in order from global to local. The search parameter adaptation adjusts the search parameters of the next layer according to the intermediate results of each layer search. The exception handling mechanism handles possible exceptions such as no solution, multiple solutions, or convergence failure in the search process. The scheme execution monitoring records key information of the search process, including search trajectory, intermediate results, resource consumption, and performance indicators. The execution scheme data contains complete information such as hierarchical structure, algorithm configuration, parameter setting, execution flow, and monitoring indicators.

[0069] The optimal convergence point is generated by performing a hierarchical search. Based on the developed hierarchical search execution scheme, the optimal position is searched in the three-dimensional decision space according to the established search strategy and hierarchy. The hierarchical search algorithm is executed to search for the optimal position layer by layer in the three-dimensional decision space according to the developed scheme. In the global search stage, the coarse-grained search is performed according to the global layer configuration of the execution scheme, using the search method and grid density specified in the scheme. In the regional search stage, the search algorithm and precision requirement specified in the scheme are used to search within the candidate region according to the regional layer configuration of the execution scheme. In the local search stage, the precise positioning method determined by the scheme is used to quickly converge to the optimal point according to the local layer configuration of the execution scheme. The global optimal point is selected from the local optimal points found in each candidate region by multi-point comparison as the final optimal convergence point. The convergence judgment condition is based on the termination criteria of the hierarchical search execution scheme, including the search precision requirement and the position change threshold. The optimal convergence point is verified by checking its performance in the original three dimensions (safety, feasibility, and executability) to ensure that each dimension meets the minimum requirement. The convergence point record contains information such as spatial coordinates, decision value, dimension scores, and search path.

[0070] The final planning trajectory is generated based on the optimal convergence point. The piecewise planning method is used to plan the path from the starting point to the convergence point and from the convergence point to the target point. The path connection is based on the spatial characteristics and decision attributes of the convergence point, and the smooth curve fitting technology is used to ensure the continuity and differentiability of the trajectory at the convergence point. The trajectory constraints are based on the decision dimension information of the optimal convergence point, including the safety requirements, feasibility limitations, and executability conditions of the convergence point. The kinematic constraints are determined according to the feasibility dimension of the convergence point, including the maximum speed and maximum acceleration limitations near the point. The dynamic constraints are determined according to the executability dimension of the convergence point to ensure that the trajectory meets the load requirements of the point. The safety constraints are determined according to the safety dimension of the convergence point, requiring the trajectory to maintain the corresponding safety level. The trajectory parameterization uses B-spline curves to represent the trajectory shape, and the control points and weights are adjusted to control the trajectory shape. The time allocation is based on the decision value and spatial position of the convergence point, considering the time requirements and execution efficiency of reaching the convergence point. The trajectory resolution is set to a time step of 0.1 seconds and a spatial step of 0.05 meters to ensure control accuracy. The execution parameter labeling is based on the decision attributes of the convergence point, labeling the execution parameters such as speed, acceleration, and safety level for each point on the trajectory. The trajectory quality evaluation is a comprehensive evaluation of the three indicators of shortest length, least time, and highest safety. The emergency plan is designed for sudden situations during trajectory execution. The final planning trajectory output contains complete information such as trajectory coordinate sequence, time sequence, execution parameters, and safety measures In order to perform the above-mentioned method embodiment corresponding to a petrochemical explosion-proof humanoid robot trajectory planning method, to realize the corresponding functions and technical effects. Referring to Figure 2 , Figure 2A structural block diagram of a petrochemical explosion-proof humanoid robot trajectory planning device 200 provided by an embodiment of the present application is shown. For ease of illustration, only parts related to the present embodiment are shown. The petrochemical explosion-proof humanoid robot trajectory planning device 200 provided by an embodiment of the present application includes: An environment perception module 201 is configured to collect petrochemical environment gas concentration data, robot joint motion parameters and target position information, perform dangerous area calibration on the gas concentration data to generate a forbidden path map, determine a safe node from the forbidden path map, and determine a joint activity range based on the joint motion parameters; A path analysis module 202 is configured to perform path search based on the target position information and the safe node distribution to construct an initial path skeleton, perform safety analysis on the initial path skeleton to identify a local dangerous section, generate motion constraint conditions and posture balance requirements based on the local dangerous section; A risk avoidance module 203 is configured to identify a major dangerous section according to the initial path skeleton, perform threat level evaluation on the major dangerous section to generate a risk weight, and perform path re-planning using the risk weight to generate a poison avoidance path; An action generation module 204 is configured to generate an action primitive library based on the motion constraint conditions and the joint activity range, perform stability filtering on the action primitive library according to the posture balance requirements to generate stable action primitives, and weave the stable action primitives between the safe nodes to form a set of marching schemes; An adaptability evaluation module 205 is configured to obtain joint load peaks in the set of marching schemes, construct a dynamics envelope based on the joint load peaks, determine an execution adaptability according to the dynamics envelope, and project the execution adaptability into a spatial coordinate to form an execution probability field; A trajectory decision module 206 is configured to construct a three-dimensional decision space based on the poison avoidance path, the set of marching schemes and the execution probability field, find an optimal convergence point in the three-dimensional decision space, and generate a final planning trajectory based on the optimal convergence point.

[0071] The petrochemical explosion-proof humanoid robot trajectory planning device 200 described above can implement a petrochemical explosion-proof humanoid robot trajectory planning method described above. The optional items in the method embodiment described above are also applicable to the present embodiment, and will not be described in detail here. The remaining content of the present embodiment can be referred to the content of the method embodiment described above, and will not be described in detail in the present embodiment.

[0072] As Figure 3As shown, the third embodiment of the present application further provides a computer device, comprising a memory 301, a processor 302, and a computer program stored in the memory 301 and capable of running on the processor 302, characterized in that the processor 302 implements the steps of the petrochemical explosion-proof humanoid robot trajectory planning method according to the first embodiment of the present application when executing the program.

[0073] The above embodiments are intended to exemplarily reproduce and deduce the technical solutions of the present application, and to completely describe the technical solutions, objects and effects of the present application, so as to make the public more thoroughly and comprehensively understand the disclosed content of the present application, and not to limit the protection scope of the present application. The above embodiments are not based on an exhaustive enumeration of the present application, and there can be many other unlisted embodiments. Any substitution and improvement made without violating the concept of the present application shall fall within the protection scope of the present application.

Claims

1. A petrochemical explosion-proof humanoid robot trajectory planning method, characterized in that, The method comprises the following steps: Collecting petrochemical environmental gas concentration data, robot joint motion parameters and target position information, calibrating the dangerous area of the gas concentration data to generate a forbidden map, determining the safe nodes from the forbidden map, and determining the joint activity range based on the joint motion parameters; Based on the target position information and the distribution of the safe nodes, a path search is performed to construct an initial path skeleton, a safety analysis is performed on the initial path skeleton to identify local dangerous sections, and motion constraints and posture balance requirements are generated based on the local dangerous sections; According to the initial path skeleton, a major dangerous section is identified, a threat level assessment is performed on the major dangerous section to generate a risk weight, and a poison avoidance path is generated by path re-planning using the risk weight; Based on the motion constraints and the joint activity range, a motion primitive library is generated, the motion primitive library is filtered for stability according to the posture balance requirements to generate stable motion primitives, and the stable motion primitives are woven between the safe nodes to form a set of marching schemes; Obtain the joint load peak value in the marching scheme set, construct a dynamics envelope based on the joint load peak value, determine the execution adaptability according to the dynamics envelope, and project the execution adaptability into a spatial coordinate to form an execution probability field; Based on the poison avoidance path, the marching scheme set and the execution probability field, a three-dimensional decision space is constructed, an optimal convergence point is found in the three-dimensional decision space, and a final planning trajectory is generated based on the optimal convergence point.

2. The method of claim 1, wherein, The method comprises the following steps: From the local dangerous section, an electrostatic accumulation tendency area is identified; Based on the electrostatic accumulation tendency area, an electric charge flow path diagram is established; Using the electric charge flow path diagram, a joint friction avoidance point is set; According to the joint friction avoidance point, the motion constraints and the posture balance requirements are established.

3. The method of claim 1, wherein, The method comprises the following steps: Based on the major dangerous section, the chemical hazard intensity is evaluated to determine the quantitative scale; According to the quantitative scale, a hazard degree calibration coefficient is set; Using the hazard degree calibration coefficient, the threat level of the major dangerous section is converted to generate a risk weight.

4. The method of claim 1, wherein, The method comprises the following steps: According to the posture balance requirements, the motion primitive library is sorted for stability to form a stability gradient; In the stability gradient, the stability peak point and the stability valley point are extracted; Partitioning is performed with the stability peak point and the stability valley point as boundaries to construct a motion stability system; The motion stability system is optimized to generate stable motion primitives.

5. The method of claim 1, wherein, The method comprises the following steps: Based on the joint load peak value, the torque distribution is identified to generate a high torque concentration zone and a low torque dispersion zone; Using the high torque concentration zone, the low torque dispersion zone is balanced to form an equal torque zone; The equal torque zone is recombined to form a recombined envelope unit; The profile-locked construction dynamics envelope of the recombination envelope unit is performed.

6. The method of claim 1, wherein, The path re-planning using the risk weight generates a poison-avoiding path, including: Based on the risk weight, a threat-avoiding area and a safety-guiding area are constructed; A path repulsion factor is generated using the threat-avoiding area, including poison repulsion and explosion repulsion; The path repulsion factor is fused with the safety-guiding area to generate a poison-avoiding path.

7. The method of claim 1, wherein, The optimal convergence point in the three-dimensional decision space is searched, including: Based on the three-dimensional decision space, the safety convergence difficulty is evaluated to determine the search depth, including safety distribution density, feasibility space connectivity, and execution probability gradient; According to the search depth, a priority configuration is determined; A hierarchical search execution scheme is developed using the priority configuration; An optimal convergence point is generated through the hierarchical search execution scheme.

8. The method of claim 2, wherein, The joint friction avoidance point is set using the charge flow path diagram, including: High-risk charge convergence nodes are extracted from the charge flow path diagram; Based on the high-risk charge convergence nodes, a friction contact forbidden zone is established; A joint motion offset point is generated using the friction contact forbidden zone; The joint motion offset point is converted into a joint friction avoidance point.

9. A petrochemical explosion-proof humanoid robot trajectory planning device, characterized by, It includes: An environment perception module is used to collect petrochemical environment gas concentration data, robot joint motion parameters and target position information, to calibrate the dangerous area of the gas concentration data to generate a forbidden map, to determine the safe nodes from the forbidden map, and to determine the joint activity range based on the joint motion parameters; A path analysis module is used to search for an initial path skeleton based on the target position information and the safe node distribution, to identify local dangerous sections through safety analysis of the initial path skeleton, and to generate motion constraint conditions and posture balance requirements based on the local dangerous sections; A risk avoidance module is used to identify major dangerous sections according to the initial path skeleton, to generate risk weights through threat level evaluation of the major dangerous sections, and to generate a poison-avoiding path through path re-planning using the risk weights; An action generation module is used to generate an action primitive library based on the motion constraint conditions and the joint activity range, to generate stable action primitives through stability filtering of the action primitive library according to the posture balance requirements, and to form a set of marching schemes by weaving the stable action primitives between the safe nodes; An adaptability evaluation module is used to obtain the joint load peak value in the marching scheme set, to construct a dynamics envelope based on the joint load peak value, to determine the execution adaptability according to the dynamics envelope, and to form an execution probability field by projecting the execution adaptability in space coordinates; A trajectory decision module is used to construct a three-dimensional decision space based on the poison-avoiding path, the marching scheme set and the execution probability field, to search for an optimal convergence point in the three-dimensional decision space, and to generate a final planning trajectory based on the optimal convergence point.

10. A computer device, comprising: A computer program product comprising a computer readable medium, the computer readable medium having stored thereon the computer program of claim 9. A computer program comprising program code adapted to perform the method of any one of claims 1 to 8 when the program is executed on a computer. A computer program comprising program code adapted to perform the method of any one of claims 1 to 8 when the program is executed on a computer. A computer program comprising program code adapted to perform the method of any one of claims 1 to 8 when the program is executed on a computer. A computer

Images