Industrial robot intelligent inspection method and system

Abstract

The invention discloses an industrial robot intelligent inspection method and system, and the method comprises the steps: collecting task point distribution data and robot pose information to construct an inspection state mapping table, and correcting a pose-task point mapping relation in combination with obstacle avoidance parameters; performing strong and weak constraint grouping on the inevitable node sequence, and generating a path planning instruction after conflict resolution through constraint conflict detection and priority arbitration; a reachable path network is obtained through multi-branch path searching, and main paths and alternative paths are screened based on path cost evaluation and overlapping road section passing conflict analysis to form a main and standby path set; generating a path execution parameter according to the traffic difficulty coefficient, performing regional risk assessment, and arranging high-risk road sections and safe road sections at intervals through buffer capacity assessment and a risk attenuation coefficient to form a risk balance path; and real-time pose comparison and deviation trend prediction are carried out based on the track reference, and a path correction amount is generated by combining tolerance dynamic adjustment to output an inspection control instruction, so that intelligent autonomous inspection of the industrial robot is realized.

Description

Technical Field

[0001] This invention relates to the field of industrial automation technology, and in particular to an intelligent inspection method and system for industrial robots. Background Technology

[0002] Industrial robot inspection, as an important component of intelligent manufacturing and automated production, plays a crucial role in equipment status monitoring and safety hazard investigation in scenarios such as substations, petrochemical plants, and intelligent warehouses. With the expansion of inspection areas and the increase in inspection points, how to rationally plan the robot's travel route, coordinate the access sequence of each inspection point, and ensure the safety and reliability of the inspection process has become a core issue restricting inspection efficiency and quality.

[0003] Existing robot inspection route planning methods mostly employ fixed routes or simple shortest path strategies, making it difficult to dynamically adjust based on real-time environmental changes and equipment operating status. On one hand, these methods lack comprehensive consideration of access constraints at inspection points, resulting in some points with time window requirements or sequential dependencies failing to complete inspection on time. On the other hand, route planning fails to adequately assess the risk status of each passage area, leaving robots without targeted safety protection measures in high-risk sections such as narrow passages and densely populated equipment areas. Furthermore, trajectory tracking deviations during inspection execution are often not corrected in a timely manner, and accumulated errors may cause the robot to deviate from the planned route or even collide with obstacles. Summary of the Invention

[0004] This invention discloses an intelligent inspection method and system for industrial robots. It aims to construct an inspection state mapping table by collecting task point distribution and robot pose information, perform constraint conflict detection and priority arbitration on necessary nodes to generate path planning instructions, and employ multi-branch path search combined with cost evaluation and overlap detection to select main and alternative paths to form a main-backup path set. Based on road segment geometric features and motion state, it calculates the passage difficulty coefficient to classify risk levels, and configures high-risk road segments and safety buffer sections according to risk attenuation coefficients to form risk-balanced paths. Finally, it combines deviation trend prediction with tolerance dynamic adjustment to generate path correction amounts, providing execution control support for industrial robot inspection.

[0005] The first aspect of this invention proposes an intelligent inspection method for industrial robots, comprising the following steps: Collect inspection task point distribution data, task point constraints and current pose information of industrial robot, and construct inspection state mapping table according to the inspection task point distribution data and current pose information; Based on the inspection status mapping table, a task coverage requirement analysis is performed to generate a sequence of necessary nodes, and a path planning instruction is generated according to the degree of matching between the sequence of necessary nodes and the constraints of the task points. Based on the path planning instructions, a multi-branch path search is performed to obtain a reachable path network. The reachable path network is divided into a main path and alternative paths to generate a main and alternative path set. Based on the main and alternative path set and the current pose information, the passage difficulty coefficient is determined by road segment decomposition. Path execution parameters are generated based on the passage difficulty coefficient. Based on the path execution parameters, a risk level assessment is performed to generate a regional risk identifier. Based on the regional risk identifier, high-risk road segments and safe road segments are divided. The high-risk road segments and the safe road segments are interleaved and arranged to generate a risk-balanced path. Based on the risk-balanced path, a time sequence is arranged to generate an inspection and scheduling sequence. Based on the inspection scheduling sequence, path deviation is detected to generate path correction amount, and inspection control instructions are generated according to the path correction amount and the primary and backup path sets.

[0006] A second aspect of this invention provides an intelligent inspection system for industrial robots, comprising: The data acquisition module is used to collect inspection task point distribution data, task point constraints and current pose information of the industrial robot, and to construct an inspection status mapping table based on the inspection task point distribution data and the current pose information. The path planning module is used to perform task coverage requirement analysis on the inspection status mapping table to generate a sequence of necessary nodes, and generate path planning instructions according to the degree of matching between the sequence of necessary nodes and the constraints of the task points. The path search module is used to perform multi-branch path search to obtain a reachable path network based on the path planning instructions, divide the reachable path network into main paths and alternative paths to generate a main and alternative path set, determine the passage difficulty coefficient by decomposing the road segments according to the main and alternative path set and the current pose information, and generate path execution parameters based on the passage difficulty coefficient. The risk scheduling module is used to perform risk level assessment and generate regional risk identifiers according to the path execution parameters, divide high-risk road segments and safe road segments based on the regional risk identifiers, insert and arrange the high-risk road segments and the safe road segments at intervals to generate risk balance paths, and generate inspection scheduling sequences based on the risk balance paths. The instruction generation module is used to generate a path correction amount based on the path deviation detection of the inspection scheduling sequence, and generate an inspection control instruction based on the path correction amount and the primary and backup path sets.

[0007] The beneficial effects of this invention are reflected in the following points: First, the necessary node sequence is divided into strong constraint node groups and weak constraint node groups according to the degree of constraint rigidity. Time window conflict detection is performed on strong constraint nodes to generate conflict node pairs. Priority arbitration is then conducted based on equipment safety level and the degree of inspection cycle overdue. This allows for automatic conflict resolution when there are constraint contradictions at multiple detection points, avoiding task execution failure due to irreconcilable constraints. Second, after candidate paths in the reachable path network are extracted and intersection calculations determine overlapping road segments, a distribution map is constructed by tracking shared nodes and identifying highly overlapping areas. Traffic conflict markers are generated by combining historical obstacle frequency and logistics busyness, enabling the selection of alternative paths passing through low-conflict road segments. When the main path is blocked, a reliable alternative can be quickly switched to. Finally, the passage time of the safety buffer section is paired with the risk exposure time of high-risk road segments to calculate the time margin. A risk attenuation coefficient is generated based on the carrying capacity value to guide the road segment ratio configuration. Combined with the cumulative analysis of deviation feature vectors, the deviation trend is predicted and the tolerance range is dynamically tightened. This can form an effective buffer before and after the risk area and correct trajectory deviations in a timely manner, reducing safety hazards during the inspection process. Attached Figure Description

[0008] The accompanying drawings illustrate specific examples of the technical solutions described in this invention and, together with the detailed embodiments, form part of the specification, serving to explain the technical solutions, principles, and effects of this invention.

[0009] Figure 1 This is a flowchart illustrating an intelligent inspection method for industrial robots according to the present invention.

[0010] Figure 2 This is a structural block diagram of an intelligent inspection system for industrial robots according to the present invention. Detailed Implementation

[0011] 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.

[0012] 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.

[0013] 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.

[0014] The technical solutions of the embodiments of this application will be described below.

[0015] like Figure 1 As shown, this embodiment of the invention provides an intelligent inspection method for industrial robots, including the following steps S110-S150: Step S110: Collect inspection task point distribution data, task point constraints and current pose information of the industrial robot, and construct an inspection state mapping table based on the inspection task point distribution data and current pose information.

[0016] Specifically, the system collects inspection task point distribution data, task point constraints, and the current pose information of the industrial robot. The inspection task point distribution data is obtained through a pre-defined inspection task configuration file. This file stores the 3D spatial coordinates and inspection priority information of each task point. A single inspection scenario typically contains 20 to 150 task points. Task point constraints are read from the task management module and include three types: time window constraints, access order constraints, and mutual exclusion constraints. Time window constraints specify the effective access period for each task point, access order constraints specify the sequential access relationship between specific task points, and mutual exclusion constraints specify combinations of task points that cannot be active simultaneously. The current pose information is collected in real-time by the industrial robot's encoder and inertial measurement unit. This pose information includes the 3D position coordinates of the robot base, attitude quaternions, and angle values ​​of each joint. Each task point in the inspection task point distribution data includes a detection area radius parameter. The robot's end effector is considered to have reached the task point when it enters this radius. The detection area radius is set according to the type of inspection task; the detection area radius for vision-based inspection tasks is typically larger than that for contact-based inspection tasks. The current pose information also includes the robot's velocity vector and acceleration vector. The magnitude of the velocity vector determines the robot's instantaneous velocity, while the magnitude of the acceleration vector reflects the robot's dynamic response capability. Together, they constrain the robot's maneuverability boundaries during obstacle avoidance. The time window in the task point constraint is recorded in seconds. Task point accesses that exceed the time window will be marked as invalid. The time window constraint is mainly used to coordinate access conflicts in multi-robot collaborative inspection scenarios.

[0017] In some embodiments, constructing an inspection state mapping table based on the inspection task point distribution data and the current pose information includes: extracting task point spatial coordinates based on the inspection task point distribution data to construct a coordinate feature set; associating the coordinate feature set with the current pose information to construct a pose-task point mapping relationship; parsing the robot's motion state from the current pose information to generate obstacle avoidance parameters; and correcting the pose-task point mapping relationship based on the obstacle avoidance parameters to generate the inspection state mapping table.

[0018] A coordinate feature set is constructed by extracting the spatial coordinates of inspection task points from the inspection task point distribution data. The three-dimensional spatial coordinates of each task point are sequentially read from the inspection task point distribution data and transferred to the coordinate feature set. The coordinate feature set standardizes the original coordinates, mapping the coordinate values ​​to a local coordinate system with the geometric center of the inspection area as the origin. This standardization eliminates differences in coordinate references between different inspection scenarios. The detection area radius parameter from the inspection task point distribution data is synchronously written into the coordinate feature set, forming a paired record with the spatial coordinates. The paired record format is a combination of position coordinates and detection radius. The detection area radius is used to determine the accessibility range of the task point. The inspection priority information of each task point in the coordinate feature set is synchronously extracted from the inspection task point distribution data. The priority information is recorded in integer form, with a value range of 1 to 10. A larger value indicates a higher urgency of the task point's inspection. The data structure of the coordinate feature set includes four fields: task point number, local coordinates, detection area radius, and inspection priority. These four fields together describe the spatial characteristics and task attributes of each task point.

[0019] A pose-task point mapping relationship is constructed by associating the coordinate feature set with the current pose information. The local coordinates of each task point in the coordinate feature set are subtracted from the robot's position coordinates in the current pose information. The difference result is converted into relative coordinates with the robot's current position as the origin. The relative coordinates intuitively reflect the spatial distance and direction of the robot to each task point. During the coordinate transformation, the rotation matrix corresponding to the pose quaternion in the current pose information is used to complete the coordinate system transformation. The X-axis of the transformed coordinate system is aligned with the robot's forward direction. The orientation angle in the current pose information divides each task point in the coordinate feature set into a front region, a side region, and a rear region. The front region corresponds to task points within a range of ±60 degrees of the orientation angle, the side region corresponds to a range of ±60 to ±120 degrees, and the rear region corresponds to a range of ±120 to ±180 degrees. The pose-task point mapping records the region category of each task point and the distance value from the robot's current position. The distance value is calculated and corrected using the detection area radius from the coordinate feature set. The corrected distance equals the Euclidean distance from the robot's position to the center of the task point minus the detection area radius. The corrected distance reflects the actual travel required for the robot to enter the effective detection range. The pose-task point mapping also uses the joint angle values ​​in the current pose information to determine workspace accessibility. Task points that exceed the working radius of the robot's end effector are marked as requiring movement to reach.

[0020] Obstacle avoidance parameters are generated by parsing the robot's motion state from the current pose information. Velocity and acceleration vectors are extracted from the current pose information, and the braking distance is calculated based on the velocity magnitude and the maximum permissible deceleration. The braking distance determines the minimum clearance required for the robot to stop safely after detecting an obstacle. The obstacle avoidance parameters are derived from the velocity and acceleration data of the current pose information, and the braking distance is calculated using the formula d_brake=v 2 / (2×a_max), where d_brake is the braking distance, v is the magnitude of the current velocity vector, and a_max is the robot's maximum permissible deceleration, which is typically set to 1.5 to 3.0 m / s² from the robot's parameter configuration. 2 The rate of change of the acceleration vector in the current pose information reflects the robot's motion stability. When motion stability is poor, the safety margin coefficient in the obstacle avoidance parameters increases from the default value of 1.2 to 1.5. The safety margin coefficient is used to add an extra safety margin on top of the braking distance. The minimum safe distance in the obstacle avoidance parameters is dynamically adjusted according to the velocity of the current pose information. The minimum safe distance decreases when the velocity is low and increases when the velocity is high. The dynamic adjustment strategy takes into account both motion efficiency and safety. The joint angle values ​​in the current pose information are used to calculate the extension state of the robot arm. When the arm is in the extended state, the avoidance radius increases accordingly to reflect the actual space occupation. The final output of the obstacle avoidance parameters includes four items: braking distance, safety margin coefficient, minimum safe distance, and avoidance radius. These four items together constitute the spatial safety constraints during the robot's motion.

[0021] Based on obstacle avoidance parameters, the pose-task point mapping relationship is corrected to generate an inspection status mapping table. Task points in the pose-task point mapping relationship whose distance value is less than the minimum safe distance in the obstacle avoidance parameters are marked as temporarily unreachable in the inspection status mapping table. These temporarily unreachable task points must wait for the robot to move beyond the safe distance before being accessed. The avoidance radius in the obstacle avoidance parameters is added to the reachability determination condition of the pose-task point mapping relationship. When there are static obstacles around a task point or the passage width is insufficient, the reachability status of that task point in the inspection status mapping table is adjusted to restricted reachability. For task points in the rear area of ​​the pose-task point mapping relationship, the arrival time is estimated by combining the braking distance of the obstacle avoidance parameters. The arrival time estimation formula is T_arrive=d_adj / v_cruise+T_turn, where T_arrive is the estimated arrival time, d_adj is the corrected distance, v_cruise is the cruising speed, and T_turn is the additional turning time. The T_turn value for task points in the rear area is 3 to 5 seconds, and the T_turn value for task points in the front area is 0. The inspection status mapping table discretizes the continuous distance values ​​in the pose-task point mapping relationship into five levels. The level classification thresholds are: 0 to 2 meters for Level 1, 2 to 5 meters for Level 2, 5 to 10 meters for Level 3, 10 to 20 meters for Level 4, and over 20 meters for Level 5. The threshold boundaries are dynamically adjusted based on the safety margin coefficient in the obstacle avoidance parameters. The inspection status mapping table ultimately includes six fields: task point number, distance level, accessibility status, area category, priority weighting factor, and estimated arrival time. The priority weighting factor is derived from the inspection priority information in the coordinate feature set, and the mapping rule is to divide the priority value by 10 to obtain a weighting factor in the range of 0.1 to 1.0. The inspection status mapping table uses a hierarchical index structure to organize the data. The index is divided into three sub-tables according to area category: front area index, side area index, and rear area index. Within each sub-table, the data is arranged in ascending order of distance level.

[0022] Step S120: Perform task coverage requirement analysis on the inspection status mapping table to generate a sequence of necessary nodes, and generate path planning instructions based on the degree of matching between the sequence of necessary nodes and the constraints of task points.

[0023] Specifically, a task coverage requirement analysis is performed on the inspection status mapping table to generate a sequence of necessary nodes. Task points in the inspection status mapping table that are directly reachable are prioritized for inclusion in the task coverage requirement analysis. Task points that are temporarily unreachable or have limited reachability are temporarily excluded and re-evaluated after the robot's pose changes. The task coverage requirement analysis traverses all task point records in the inspection status mapping table, comprehensively evaluating the access urgency of each task point based on its priority weighting factor and distance level. The access urgency score is calculated using the formula: S_urgency = w1 × P_weight + w2 × (6 - D_level) / 5, where S_urgency is the access urgency score, P_weight is the priority weighting factor (range 0.1 to 1.0), D_level is the distance level (range 1 to 5), w1 and w2 are weighting coefficients set to 0.6 and 0.4 respectively, and the score range is 0.14 to 1.0. Task coverage requirement analysis marks task points with an access urgency score exceeding 0.6 as mandatory nodes. Mandatory nodes are task points that must be traversed on the inspection path and cannot be skipped or delayed. Task points in the inspection status mapping table categorized as "Ahead Area" are prioritized as mandatory nodes under the same access urgency condition, as these have the shortest arrival paths and require no additional turning actions. The mandatory node sequence is arranged in descending order of access urgency score, forming an ordered task point access queue. Each node in the queue records three attributes: node number, access urgency score, and estimated arrival time. The estimated arrival time is synchronously extracted from the inspection status mapping table, and the access urgency score is used for priority determination during subsequent constraint grouping.

[0024] In some embodiments, generating path planning instructions based on the degree of matching between the required node sequence and the task point constraints includes: dividing the required node sequence into a strongly constrained node group and a weakly constrained node group according to the task point constraints; performing constraint conflict detection on the strongly constrained node group and the weakly constrained node group to generate conflict node pairs; performing priority arbitration processing on the conflict node pairs to generate a conflict-resolved node sequence; and generating path planning instructions based on the conflict-resolved node sequence.

[0025] Based on task point constraints, the sequence of necessary nodes is divided into strongly constrained node groups and weakly constrained node groups. Nodes with both time window and access order constraints in their task point constraints are assigned to the strongly constrained node group, while nodes with only a single constraint type or no constraint are assigned to the weakly constrained node group. Nodes in the strongly constrained node group have strict requirements on both access time and access order. Nodes with a time window duration of less than 60 seconds in their task point constraints are automatically assigned to the strongly constrained node group, as short time windows mean less tolerance for access timing. Nodes with access order constraints marked in their task point constraints and having the highest constraint priority are assigned to the strongly constrained node group; the access order of these nodes cannot be adjusted. Nodes with only mutual exclusion access constraints in their task point constraints are assigned to the weakly constrained node group. Mutual exclusion access constraints can be satisfied by adjusting node intervals, offering greater constraint flexibility. Nodes in the weakly constrained node group have greater scheduling flexibility during path planning, and their order can be adjusted based on actual path costs without violating the task point constraints. Strongly constrained node groups and weakly constrained node groups each have their own independent index structures. These index structures use node numbers as keys and constraint type identifiers and access urgency scores as values. The access urgency score is inherited from the node attributes of the required node sequence and is used to prioritize nodes within the same group. The number of nodes in the strongly constrained node group typically accounts for 20% to 40% of the total number of nodes in the required node sequence. When the proportion of strongly constrained nodes exceeds 50%, the system issues a scheduling complexity warning indicating potential over-constraint settings.

[0026] Constraint conflict detection is performed on strongly constrained node groups and weakly constrained node groups to generate conflicting node pairs. Any two nodes in a strongly constrained node group are considered to have conflicting access orders, determined by the inconsistency between the access order constraints in the task point constraints and the actual order in the required node sequence. Conflict pair detection employs a pairwise comparison strategy: full pairing detection is performed within each strongly constrained node group, and cross-group pairing detection is performed between strongly constrained and weakly constrained node groups. The detection process records the conflict type and severity of each pair of nodes. In a strongly constrained node group, node A's time window is from second 100 to second 130, and node B's time window is from second 80 to second 110. If node A precedes node B in the required node sequence and the travel time between the two nodes exceeds 20 seconds (calculated by subtracting node A's estimated arrival time from node B's estimated arrival time), then node A and node B constitute a time-conflicting node pair. When a node in a weakly constrained node group encounters a mutual exclusion access conflict with a node in a strongly constrained node group, the weakly constrained node in the conflicting node pair is marked as the adjustable party, and the strongly constrained node is marked as the fixed party. Subsequent conflict resolution prioritizes adjusting the position of the adjustable party. Conflicting node pairs are sorted according to conflict severity, with time conflicts and sequence conflicts being more severe than mutual exclusion access conflicts. Higher-severity conflict pairs are resolved first. Conflicting node pairs are stored in a list, where each element contains four items: node number pair, conflict type, conflict severity, and adjustable party mark. The list is sorted in descending order of conflict severity to support sequential processing during priority arbitration.

[0027] Conflicting node pairs are prioritized for arbitration to generate a reconciled node sequence. Each element in the conflict pair list is arbitrated sequentially according to its conflict severity. The access urgency scores of the two nodes in a conflict pair are compared. The node with the higher access urgency score retains its original position in the reconciled node sequence, while the node with the lower access urgency score is repositioned to eliminate the conflict. Priority arbitration follows a strong constraint priority principle; nodes in the strongly constrained node group have a higher priority than nodes in the weakly constrained node group by default, even if the original access urgency score of the weakly constrained node is higher. The reconciled node sequence is generated through iterative adjustment. Each iteration processes a pair of conflicting nodes in the conflict pair list, and after adjustment, it checks for new conflicts. If new conflicts arise, iteration continues until there are no conflicts or the maximum number of iterations (100) is reached. If the maximum number of iterations is exceeded, the intermediate result with the fewest conflicts is selected as the output. When time conflicts exist between conflicting node pairs, priority arbitration calculates the time cost of two adjustment schemes and selects the scheme with the smaller total time cost. The time cost calculation formula is C_time = ΔT_travel + T_wait, where C_time is the time cost, ΔT_travel is the change in travel time caused by the node position adjustment, and T_wait is the additional waiting time required after the adjustment. After conflict resolution, the node sequence retains all nodes in the mandatory node sequence, only adjusting the order of the nodes. The adjustment range aims to minimize the change in the total path length. After conflict resolution, each node in the node sequence is renumbered according to the adjusted access order, with the numbering starting from 1 and incrementing continuously. This numbering information is used to generate path planning instructions. After priority arbitration, there are no order conflicts, time conflicts, or mutual exclusion access conflicts between adjacent nodes in the node sequence after conflict resolution, and the sequence satisfies all the requirements of the task point constraints.

[0028] Path planning instructions are generated based on the node sequence after conflict resolution. Nodes in the conflict-resolved node sequence are sequentially converted into path planning instructions, with each node corresponding to a target pose instruction. The target pose instruction includes target position coordinates and target attitude angles. The target position coordinates of the target pose instruction are extracted from the inspection status mapping table record corresponding to the conflict-resolved node sequence, and the target attitude angle is determined based on the detection task type and the orientation of the detected object. Nodes marked as strongly constrained in the conflict-resolved node sequence have hard time constraints set in the path planning instructions; the robot must reach the node within a specified time window. Arriving after the timeout will trigger an alarm and record the abnormal event. Path planning instructions corresponding to weakly constrained nodes in the conflict-resolved node sequence have soft time constraints, allowing for a certain degree of time deviation without triggering an alarm. The total length of the path planning instruction queue is equal to the number of nodes in the conflict-resolved node sequence, and the instructions are arranged according to the node access order. After the path planning instructions are generated, an integrity check is performed. The check rules include: instruction queue continuity check to confirm that there are feasible paths between the target positions of adjacent instructions; target position reachability check to confirm that each target position is within the robot's workspace; and time constraint feasibility check to confirm that the time window corresponding to the hard time constraint matches the estimated arrival time. After the check passes, the path planning instructions are output to the path execution module. The path planning instructions ultimately contain three types of content: node access order, target pose information, and time constraint information. These three types of content serve as input constraints for subsequent multi-branch path search.

[0029] Step S130: Based on the path planning instructions, perform multi-branch path search to obtain the reachable path network. Divide the reachable path network into main paths and alternative paths to generate a set of main and alternative paths. Based on the set of main and alternative paths and the current pose information, determine the passage difficulty coefficient through road segment decomposition. Generate path execution parameters based on the passage difficulty coefficient.

[0030] Specifically, a reachable path network is obtained through multi-branch path search based on path planning instructions. The node access order and target pose information in the path planning instructions serve as input constraints for the multi-branch path search. The search algorithm explores multiple feasible paths between adjacent nodes while satisfying the access order. The multi-branch path search employs an improved A* algorithm, incorporating the time constraint information of the path planning instructions into the heuristic function. For nodes with hard time constraints, the search prioritizes path branches with lower time costs. The design of the heuristic function considers both path length and time feasibility. The path search results between adjacent nodes in the path planning instructions typically contain 2 to 5 feasible paths. These feasible paths differ in terms of distance, number of turns, and obstacle distribution. In automated assembly lines, there are often multiple path choices between adjacent workstations, such as going around the equipment edge or going straight through the aisle. The reachable path network stores the search results in a directed graph structure. Nodes in the graph correspond to the mandatory nodes in the path planning instructions, and edges correspond to feasible paths between adjacent nodes. Each edge includes attribute information such as path length, turning point location, and area identifiers. The edge attribute information is recorded synchronously during the path search process. The total number of edges in a reachable path network is typically 3 to 4 times the number of nodes in the path planning instruction. When there are too many edges, low-quality paths are pruned to control the network size. The pruning strategy prioritizes removing edges whose path length exceeds 1.5 times that of the shortest path.

[0031] In some embodiments, the step of dividing the reachable path network into primary and alternative paths to generate a primary and backup path set includes: evaluating and ranking the path cost of the reachable path network to identify an optimal path candidate set; performing path overlap detection on the optimal path candidate set to generate an overlapping road segment distribution; performing obstacle accessibility analysis on the overlapping road segment distribution to generate traffic conflict markers; and retaining low-conflict alternative paths based on the traffic conflict markers to form a primary and backup path set.

[0032] Path cost evaluation and ranking are performed on reachable path networks to identify the optimal path candidate set. The path length, number of turns, and type of area traversed by each edge in the reachable path network are used as basic indicators for cost evaluation. These three indicators quantify the distance cost, maneuver cost, and environmental cost of the path, respectively. The cost evaluation uses a weighted summation method to calculate the comprehensive cost value, with the formula C_total = w1 × C_dist + w2 × C_turn + w3 × C_env, where C_total is the comprehensive cost value, C_dist is the distance cost equal to the path length divided by a reference length, C_turn is the maneuver cost equal to the number of turns multiplied by the single-turn cost coefficient, and C_env is the environmental cost assessed based on the equipment density of the traversed areas. The weighting coefficients w1, w2, and w3 are set to 0.4, 0.3, and 0.3, respectively. Paths passing through narrow passages or densely populated equipment areas in the reachable path network receive higher environmental cost scores. In the inspection scenario of a CNC machining center, the environmental cost of paths passing through the machine tool spindle area is significantly higher than that of paths passing through the walkway area. After cost evaluation, multiple paths connecting the same node pair in the reachable path network are sorted in ascending order of comprehensive cost. The path with the lowest cost after sorting is marked as the optimal path for that node pair. The optimal path candidate set includes the optimal and second-best paths for each node pair. The cost difference between the second-best and optimal paths is usually controlled within 15%. Paths with excessively large cost differences are not included in the optimal path candidate set. Each path in the optimal path candidate set records two attributes: path number and comprehensive cost value, which are used as the sorting criteria for subsequent overlap detection. Some node pairs in the reachable path network have only a single feasible path. For these node pairs, only the optimal path is recorded in the optimal path candidate set, without a second-best path. Single-path node pairs are common in processing units with compact equipment layouts.

[0033] For example, the step of performing path overlap detection on the optimal path candidate set to generate an overlapping road segment distribution includes: determining overlapping road segments by performing path overlap detection on the optimal path candidate set; tracing shared nodes in the overlapping road segments to construct a shared node distribution map; identifying high overlap areas and low overlap areas based on the shared node distribution map to generate overlap classification labels; and jointly generating an overlapping road segment distribution based on the overlap classification labels and the shared node distribution map.

[0034] Overlapping road segments are determined by path overlap detection based on the optimal path candidate set. The trajectory point sequences of any two paths in the optimal path candidate set are compared point-by-point. Points with a distance less than a set threshold are considered overlapping. This threshold is determined based on the robot's outer envelope width, typically between 0.3 and 0.5 meters. A threshold that is too large will lead to false positives, while a threshold that is too small will lead to false negatives. Consecutive overlapping points constitute overlapping road segments. The start and end positions of an overlapping road segment are determined by the coordinates of the first and last overlapping points. The length of the overlapping road segment is calculated from the arc length of the path between the start and end positions. The arc length is calculated by summing the distances between adjacent trajectory points segment by segment. When the number of paths in the optimal path candidate set is n, path overlap detection requires n×(n-1) / 2 pairwise comparisons. For large-scale path sets, a spatial partitioning acceleration strategy is used to reduce the number of comparisons. Spatial partitioning divides the inspection area into grids, and only path segments located in the same or adjacent grids are compared. During the detection process, overlapping road segments are marked with the numbers of the two paths to which they belong. The path numbers are a combination of node pair numbers and path sequence numbers, in the format "starting point number - ending point number - path sequence number". This format allows for reverse location of the specific path to which an overlapping road segment belongs. Paths with lower overall cost value in the optimal path candidate set are marked as priority retention paths during overlap detection. When two paths overlap, the priority retention path has a higher retention priority. After the overlapping road segment detection is completed, an overlapping road segment set is output. Each element in the set contains three pieces of information: the coordinates of the overlapping road segment, the overlap length, and the associated path number.

[0035] In overlapping road segments, shared nodes are tracked to construct a shared node distribution map. Points on the trajectory of an overlapping road segment that are path turning points or transit points are marked as shared nodes. Shared nodes are key locations traversed by multiple paths. Turning points are determined by an angle exceeding 10 degrees between the directions of the two trajectories before and after the point. Transit points are determined by a distance less than the radius of the detection area between the point and a necessary transit node. The shared node distribution map represents the connections between shared nodes using an undirected graph structure. Nodes correspond to shared nodes, and edges correspond to connections between adjacent shared nodes in the overlapping road segment. The edge weight is set as the path distance between adjacent shared nodes, calculated segmentally from the overlap length of the overlapping road segment. The density of shared nodes in an overlapping road segment reflects the degree of path convergence in that area. Areas with high shared node density typically correspond to passageway intersections or concentrated equipment areas in inspection scenarios. In automated welding workshops, the central aisle of welding station clusters is often a necessary transit point for multiple inspection paths, forming a high-density shared node area. Each node in the shared node distribution map is accompanied by attribute information, including node coordinates and the number of associated paths. When overlapping road segments are long, they may contain multiple shared nodes. The spacing between shared nodes reflects the structural characteristics of the overlapping road segments. Adjacent shared nodes with a spacing of less than 1 meter are merged into a single node to simplify the structure of the shared node distribution map. The number of connected components in the shared node distribution map reflects the degree of concentration of the overlapping area. A single connected component indicates that the overlapping area is concentrated, while multiple connected components indicate that the overlapping area is dispersed across different functional blocks in the workshop.

[0036] Based on a shared node distribution map, high-overlap and low-overlap regions are identified, generating overlap classification labels. Shared nodes with a degree greater than 3 in the shared node distribution map are marked as core nodes in high-overlap regions. The node degree represents the number of paths associated with that node. Core nodes have a high degree of path convergence, which easily leads to traffic conflicts. Core node identification uses a degree threshold method, with a threshold set to 3 based on the empirical judgment that three-way intersections in typical industrial scenarios constitute a significant conflict risk. The overlap classification labels divide the shared node distribution map into several regions, expanding outward from the core nodes to form high-overlap regions. The expansion radius is dynamically adjusted according to the degree of the core nodes, calculated using the formula R_expand=R_base×(1+0.2×(d-3)), where R_base is set to a base expansion radius of 2 meters, and d is the degree of the core node. Regions outside the coverage area of ​​the core nodes in the shared node distribution map are classified as low-overlap regions. Low-overlap regions have a lower degree of path convergence, and the risk of traffic conflicts is relatively controllable. The overlap classification marker uses a combination of region number and overlap level. The region number uniquely identifies each partitioned region, and the overlap level is either high or low. The region number uses the format "H" or "L" prefix followed by a three-digit serial number, where H indicates a high overlap region and L indicates a low overlap region. The area ratio of high overlap regions in the shared node distribution map is typically 15% to 25%. When the ratio is too high, the path distribution of the reachable path network needs to be optimized to reduce the overlap. When the ratio exceeds 30%, the system issues a path planning quality warning indicating an excessive overlap risk. The overlap classification marker outputs a list of high overlap regions and a list of low overlap regions. The lists record three items for each region: region number, included shared node number, and overlap level.

[0037] The overlapping road segment distribution is generated jointly based on the overlap classification markers and the shared node distribution map. Overlapping road segments within high-overlap areas in the overlap classification markers are marked with a high-priority processing identifier in the overlapping road segment distribution. These road segments have a higher risk of traffic conflicts and require priority conflict resolution. The high-priority processing identifier is represented by an integer ranging from 1 to 3, with higher values ​​indicating higher processing priority. Road segments located within the direct coverage area of ​​core nodes have a priority of 3. The coordinate information of each shared node in the shared node distribution map is written into the overlapping road segment distribution, forming a correlation with the start and end coordinates of the overlapping road segments to fully describe the spatial structure of the overlapping area. The correlation method is to add a field for a list of shared node numbers passed through the overlapping road segment records, with the list sorted by the direction of travel. The overlapping road segment distribution integrates the regional division results of the overlap classification markers and the node connection information of the shared node distribution map to form a complete data structure containing the location of overlapping road segments, overlap level, and processing priority identifier. In the overlapping classification, overlapping road segments within low-overlap areas are marked with a low-priority processing identifier in the overlapping road segment distribution. These road segments have a low risk of traffic conflict and can be processed when computational resources are sufficient. A low-priority processing identifier value of 0 indicates that no priority processing is required. The overlapping road segment distribution uses a spatial index structure to organize the data. The spatial index uses an R-tree structure, and the leaf nodes of the R-tree store the mapping relationship between the minimum bounding rectangle of the overlapping road segment and the road segment number, supporting fast retrieval of overlapping road segment information within a specified area by coordinate range.

[0038] Obstacle accessibility analysis is performed on overlapping road segments to generate traffic conflict markers. Road segments with high overlap levels are prioritized for obstacle accessibility analysis, which includes whether the road segment width meets the robot's passage requirements and whether there are dynamic obstacles on both sides of the road segment. Traffic conflict markers record the traffic conflict risk level of each road segment in the overlapping road segment distribution. The risk level is comprehensively assessed based on the road segment width, obstacle distance, and processing priority indicators in the overlapping road segment distribution. Road segments with a processing priority indicator of 3 receive a higher base risk score in the risk assessment. Road segments in the overlapping road segment distribution whose width is less than 1.5 times the robot's outer envelope width are marked as high conflict risk in the traffic conflict markers; these types of road segments are difficult to support simultaneous passage by multiple robots. The traffic conflict markers classify conflict risk into three levels: high, medium, and low. For high conflict risk road segments, it is recommended to retain only a single path; for medium conflict risk road segments, two paths (primary and backup) can be retained; and for low conflict risk road segments, multiple alternative paths can be retained. In overlapping road segment distributions, road segments located in densely populated equipment areas tend to be marked with higher conflict risk levels in traffic conflict markings. Narrow passageways between injection molding machines in an injection molding workshop are typical high-conflict-risk road segments. Traffic conflict markings are stored in a mapping structure with road segment number as the key and conflict risk level and conflict type as the value. Conflict types include three categories: width restriction, obstacle interference, and path intersection.

[0039] Based on traffic conflict markers, low-conflict alternative paths are retained to form a primary and backup path set. For alternative paths corresponding to high-conflict-risk road segments in the traffic conflict markers, only the lowest-cost alternative path is retained in the primary and backup path set; the remaining alternative paths are eliminated to avoid the superposition of traffic conflict risks. The lowest-cost path is determined by the comprehensive cost recorded in the optimal path candidate set. The primary path in the primary and backup path set is formed by connecting the optimal paths of each node pair. The road segments traversed by the primary path are all at low or medium conflict-risk levels in the traffic conflict markers, and the primary path constitutes the default execution path for inspection tasks. All alternative paths corresponding to low-conflict-risk road segments in the traffic conflict markers are retained in the primary and backup path set. Low-conflict road segments have good conditions for multi-path parallel passage, and alternative paths provide alternative choices when the primary path is blocked. The number of alternative paths for each road segment in the primary and backup path set is dynamically adjusted according to the traffic conflict markers: 0 to 1 alternative paths for high-conflict road segments, 1 to 2 for medium-conflict road segments, and 2 to 3 for low-conflict road segments. This differentiated configuration of the number of alternative paths balances path redundancy and storage overhead. The conflict type information marked in the traffic conflict marker is written into the road segment attribute field of the primary and backup path sets. When the conflict type is width-restricted, the maximum permissible speed limit is added to the road segment attribute; when the conflict type is obstacle interference, the sensor-enhanced acquisition identifier is added to the road segment attribute; when the conflict type is path intersection, the intersection point coordinate information is added to the road segment attribute. After the primary and backup path sets are generated, connectivity verification is performed to ensure that the primary path can connect all necessary nodes and that there are no breaks between road segments. If the verification fails, the path will backtrack to the reachable path network to search for and supplement the path.

[0040] In some embodiments, determining the passage difficulty coefficient by decomposing road segments based on the primary and backup path sets and the current pose information includes: generating road segmentation points based on the primary and backup path sets; dividing road segments based on the road segmentation points and extracting turning angles and road segment lengths to generate geometric feature parameters; analyzing the robot's motion state based on the current pose information to generate motion state parameters; and performing kinematic constraint analysis on the geometric feature parameters and the motion state parameters to form the passage difficulty coefficient.

[0041] Road segmentation points are generated based on the primary and backup path sets. Locations in the primary path trajectory where the curvature change exceeds a set threshold are marked as road segmentation points. Abrupt curvature changes typically correspond to curves or directional adjustments on the path. These road segmentation points divide the primary and backup path sets into several relatively straight road segments with minimal curvature changes within each segment, allowing the robot to maintain a relatively stable motion state. The bifurcation and merging points between the primary and backup paths in the primary and backup path sets are automatically marked as road segmentation points. The bifurcation point is where the primary path and backup path begin to separate, and the merging point is where the backup path rejoins the primary path. The spacing between road segmentation points is typically controlled within the range of 3 to 8 meters. Too small a spacing results in overly fragmented road segment divisions, while too large a spacing leads to significant differences in features within individual road segments, affecting the accuracy of the difficulty assessment. Locations of essential nodes along the primary and backup path sets are also marked as road segmentation points. At these essential nodes, the robot needs to pause and perform detection tasks, forming natural road segment boundaries. The conflict type information in the segment attribute fields of the primary and backup path sets is synchronously written to the corresponding segment split point records. The conflict type information is used for speed limit determination when generating subsequent path execution parameters. The segment split points are numbered sequentially along the primary path, with the numbers starting from 1 and increasing continuously. The numbers and coordinate information of the segment split points are stored in pairs, and the paired data structure is a list of segment split points.

[0042] The system divides the road into segments based on segment division points and extracts turning angles and segment lengths to generate geometric feature parameters. Segment division points divide the primary and backup path sets into continuous segment sequences. The trajectories between adjacent segment division points constitute an independent segment, which is the basic unit for assessing traffic difficulty. The total number of segment sequences equals the number of segment division points minus one. Geometric feature parameters include three indicators for each segment: length, start and end direction angles, and turning angle. The segment length is calculated from the arc length of the trajectory between segment division points, using a cumulative method of accumulating trajectory points, i.e., summing the distance between adjacent trajectory points segment by segment. The turning angle at a segment division point is defined as the angle between the ending direction of the previous segment and the starting direction of the next segment. The turning angle reflects the directional adjustment range required by the robot at that position, and is calculated using a vector angle formula. In the geometric feature parameters, segment division points with turning angles less than 15 degrees are marked as gentle turns, those between 15 and 45 degrees as medium turns, and those greater than 45 degrees as sharp turns. In the inspection paths of precision assembly workshops, sharp turns typically occur at equipment corners. The type of turn at road segment division points affects the robot's travel speed and attitude adjustment strategy. Sharp turns require speed reduction and may require turning on the spot, while medium turns can be smoothly traversed using an arc transition method. Geometric feature parameters are organized using road segment numbers as indices. Each road segment corresponds to a set of length, direction angle, and turning angle values. The data structure uses array format to support fast access by road segment number.

[0043] The robot's motion state is analyzed based on its current pose information to generate motion state parameters. The velocity and acceleration vectors in the current pose information reflect the robot's instantaneous motion state. The magnitude and direction of the velocity vector determine the robot's motion capability boundaries, while the acceleration vector reflects the robot's current dynamic response state. The motion state parameters extract four pieces of information from the current pose information: the robot's current velocity, current orientation angle, maximum permissible speed, and minimum turning radius. These four pieces of information together describe the kinematic constraints the robot faces when traversing each road segment. The current velocity and current orientation angle are real-time variables, while the maximum permissible speed and minimum turning radius are inherent robot parameters. When the velocity in the current pose information is high, the robot needs a longer deceleration interval before sharp turns to ensure safe passage. The length of the deceleration interval is proportional to the square of the current velocity. The orientation angle in the current pose information is compared with the initial orientation angle of each road segment in the geometric feature parameters. The difference between the two is the orientation deviation angle. The orientation deviation angle reflects the steering adjustment required when the robot enters that road segment. When the orientation deviation angle is large, the robot needs to complete the steering action before entering the road segment. Motion state parameters are used in conjunction with geometric feature parameters to assess the ease with which the robot traverses each road segment in its current motion state.

[0044] A kinematic constraint analysis is performed on the geometric feature parameters and motion state parameters to form a passage difficulty coefficient. The turning angle in the geometric feature parameters is compared with the minimum turning radius in the motion state parameters. Road segments with turning angles exceeding the robot's turning ability are marked as high-difficulty road segments. The passage difficulty coefficient comprehensively evaluates the ease of passage for each road segment. The evaluation indicators include the matching degree between road segment length and current speed, the matching degree between turning angle and turning ability, and the deviation degree between road segment direction and current orientation. The calculation formula is D=w1×(L / v_max) / t_ref+w2×min(θ / θ_max,1)+w3×(Δφ / 180), where D is the passage difficulty coefficient, L is the road segment length, v_max is the maximum allowable speed, t_ref is the reference passage time set to 10 seconds, θ is the turning angle, θ_max is the maximum allowable turning angle set to 90 degrees, Δφ is the deviation angle between road segment direction and current orientation, and w1, w2, and w3 are weighting coefficients set to 0.4, 0.35, and 0.25, respectively. When the current speed is high in the motion state parameters, the difficulty coefficients for short road segments and sharp turns in the geometric feature parameters increase accordingly, requiring the robot to spend more time and distance to decelerate and adjust. The difficulty coefficient is represented by a normalized value from 0 to 1, where 0 indicates no difficulty and 1 indicates extremely high difficulty requiring special handling. Normalization facilitates comparison of difficulty between different road segments and threshold determination. The difficulty coefficients are output organized by road segment number, and each road segment's difficulty coefficient is paired with its corresponding road segment number to form a difficulty coefficient list, the length of which is equal to the total number of road segments.

[0045] Path execution parameters are generated based on the difficulty coefficient of the route. For road segments with higher difficulty coefficients, a lower target speed is set in the path execution parameters; for road segments with lower difficulty coefficients, a higher target speed is set. Speed ​​settings prioritize safety while also considering inspection efficiency. Target speed and difficulty coefficient are negatively correlated. The path execution parameters include four parts: target speed, difficulty coefficient, caution marking, and travel time budget for each road segment. These four parts work together to control the robot's road segment movement. The target speed determines the robot's movement speed, and the caution marking triggers a special movement mode. Road segments with a difficulty coefficient exceeding 0.7 are marked as caution sections in the path execution parameters. In these segments, the robot needs to activate a low-speed mode and increase the sensor sampling frequency. Caution in areas with high-temperature equipment or precision instruments can effectively reduce the risk of accidental collisions. In low-speed mode, the target speed is reduced to 30% of the maximum permissible speed. The travel time budget for each road segment in the path execution parameters is determined based on the segment length and target speed. The travel time budget T_budget = L / v_target × (1 + k_margin), where L is the segment length, v_target is the target speed, and k_margin is a time margin coefficient set to 0.1. The travel time budget is increased by 10% on top of the theoretical travel time to cope with unforeseen circumstances. Road segments with a difficulty coefficient between 0.3 and 0.7 adopt the standard travel mode in the path execution parameters, requiring no special speed limits or attitude adjustments. In the standard travel mode, the target speed is set to 70% of the maximum permissible speed. The path execution parameters are organized in the form of instruction sequences, which correspond one-to-one with the segment sequences in the primary and backup path sets. Each instruction includes two parts: a segment number and execution parameters.

[0046] Step S140: Based on the path execution parameters, perform risk level assessment to generate regional risk labels. Based on the regional risk labels, divide high-risk road segments and safe road segments. Insert high-risk road segments and safe road segments at intervals to generate risk-balanced paths. Based on the risk-balanced paths, perform time-series arrangement to generate inspection and scheduling sequences.

[0047] Specifically, risk level assessments are conducted based on route execution parameters to generate regional risk labels. The target speed, difficulty coefficient, and caution markings for each road segment within the route execution parameters serve as inputs for the risk level assessment. These three parameters reflect the traffic risk characteristics of each road segment from different dimensions. The risk level assessment quantifies the risk of each road segment in the route execution parameters, with quantification indicators including three sub-items: speed risk, difficulty risk, and environmental risk. Speed ​​risk is positively correlated with the target speed, difficulty risk is positively correlated with the difficulty coefficient, and environmental risk is determined based on the equipment distribution density of the areas traversed by the road segment. The risk quantification calculation formula is R_total = w1 × (v_target / v_max) + w2 × D_pass + w3 × ρ_device, where R_total is the comprehensive risk value, v_target is the target speed, v_max is the maximum permissible speed, D_pass is the passage difficulty coefficient, and ρ_device is the normalized value of device density, which is equal to the number of devices within a 2-meter radius of the road segment divided by the number of reference devices, with a value ranging from 0 to 1. w1, w2, and w3 are weighting coefficients set to 0.3, 0.4, and 0.3, respectively. Road segments marked as "cautious passage" in the route execution parameters automatically receive a higher base risk score in the risk level assessment. The base risk score for cautious passage road segments is set to 0.5 and is added to the comprehensive risk value. Regional risk identification attaches the risk level assessment results to each road segment in the form of a label, which includes three parts: risk level, main risk type, and risk score. In the path execution parameters, road segments with tighter travel time budgets tend to be marked with higher risk levels in the regional risk labeling. Time constraints mean the robot needs to travel faster, thus increasing movement risks. The regional risk labeling divides risk levels into three levels: high risk, medium risk, and low risk. Road segments with a risk score above 0.7 are classified as high risk, those with a risk score between 0.4 and 0.7 as medium risk, and those with a risk score below 0.4 as low risk.

[0048] Based on regional risk markers, road sections are divided into high-risk and safe road sections. Road sections with a high-risk level in the regional risk markers are directly classified into the high-risk road section set, which are areas requiring special attention during inspections. Safe road sections are selected from road sections with a low-risk level in the regional risk markers. Selection criteria include sufficient road width, no surrounding dynamic obstacles, and flat, water-free ground. The criterion for sufficient road width is that the width is greater than twice the outer envelope width of the robot. Road sections with a medium-risk level in the regional risk markers are further classified according to the main risk type. Those with speed risk as the main risk type are classified into safe road sections, while those with environmental risk as the main risk type are classified into high-risk road sections. This secondary classification refines the classification of medium-risk road sections. In intelligent manufacturing workshops, high-risk road sections typically correspond to areas requiring special attention, such as equipment operating areas, material storage areas, and personnel work areas. In stamping workshops, the passageways around the stamping machine group are typical high-risk road sections. Safe road sections typically correspond to areas with good traffic conditions, such as dedicated inspection lanes, equipment interval areas, and open passageways. The results of the high-risk road section and safe road section division are output in the form of a road section number list. Each element in the high-risk road section list records two items: road section number and risk score. Each element in the safe road section list records two items: road section number and road section length. The union of the two lists covers all road sections in the primary and backup path sets.

[0049] In some embodiments, the step of interleaving high-risk road segments and safe road segments to generate a risk-balanced path includes: generating a risk exposure duration based on the high-risk road segments; selecting a safe buffer segment adjacent to the high-risk road segment from the safe road segments; assessing the buffer capacity of the safe buffer segment in conjunction with the risk exposure duration to generate a risk attenuation coefficient; and configuring the high-risk road segment and the safe buffer segment in proportion according to the risk attenuation coefficient to form a risk-balanced path.

[0050] Risk exposure duration is generated based on high-risk road segments. The length of each segment in the high-risk road segment list and the target travel speed are used to calculate the time required for the robot to traverse that segment. The calculated time is the risk exposure duration for that segment, calculated as T_exp = L_risk / v_target, where T_exp is the risk exposure duration, L_risk is the length of the high-risk road segment, and v_target is the target travel speed. Risk exposure duration reflects the duration of continuous operation of the robot in a high-risk environment; a longer duration means more cumulative exposure time. Road segments marked as "use with caution" in the high-risk road segment list have relatively longer risk exposure durations due to their lower target speeds. For example, the risk exposure duration of a segment traversing a high-temperature area in a heat treatment workshop is significantly longer than that of a segment of the same length in a normal-temperature area. The risk exposure durations are organized into a duration sequence according to the order of the high-risk road segments, with each element in the duration sequence corresponding to the exposure duration value of a high-risk road segment. The risk score in the high-risk road segment list is used as a weighting factor to adjust the risk exposure duration. The adjustment formula is T_adj = T_exp × (1 + R_total), where T_adj is the adjusted risk exposure duration and R_total is the comprehensive risk value of the road segment. The adjusted risk exposure duration more accurately reflects the actual degree of risk accumulation. The risk exposure duration sequence is organized using the high-risk road segment number as an index, and the sequence length is equal to the number of road segments in the high-risk road segment set.

[0051] Safe buffer zones adjacent to high-risk road sections are selected from the safe road sections. Road sections spatially adjacent to high-risk road sections in the safe road section list meet the criteria for buffer zones. Spatially adjacent means the distance between the endpoints of the two road sections is less than a set threshold or there is a direct connection; the set threshold is 0.5 meters. The selection of safe buffer zones involves checking the adjacency relationship between each road section in the safe road section list and the high-risk road section. Safe road sections that meet the adjacency criteria are included in the candidate set of safe buffer zones. The length of a road section in the safe road section list is used to determine whether the road section meets the buffer requirements. Safe road sections with a length of less than 2 meters are not included in the candidate set of safe buffer zones. The length threshold of 2 meters is determined based on the robot's braking distance and state transition time. The candidate set of safe buffer zones is organized according to their association with high-risk road sections, with each high-risk road section corresponding to a set of adjacent candidate safe buffer zones. The passage conditions of the safe buffer zone should match the risk type of the adjacent high-risk road section. For high-risk road sections with environmental risk types, safe road sections with significantly different environmental conditions are prioritized as buffer zones. Each element in the candidate set of safety buffer sections records three items: buffer section number, road section length, and associated high-risk road section number.

[0052] For example, the step of assessing the buffer capacity of the safety buffer segment in conjunction with the risk exposure duration to generate a risk attenuation coefficient includes: estimating the passage time of the safety buffer segment to generate a buffer duration distribution; identifying locations with sufficient time margin in the buffer duration distribution in conjunction with the risk exposure duration to construct a margin location set; assessing the risk carrying capacity of the margin location set to generate a carrying capacity value; and determining the risk overflow threshold of the carrying capacity value to generate a risk attenuation coefficient.

[0053] The process of estimating the travel time of safety buffer sections generates a buffer time distribution. The segment length and target travel speed of each buffer section in the candidate set are used to estimate the time required for the robot to traverse that segment. The estimation result is the buffer time of that safety buffer section, calculated using the formula T_buf = L_buf / v_buf, where T_buf is the buffer time, L_buf is the buffer section length read from the candidate set, and v_buf is the target travel speed read from the path execution parameters. The buffer time distribution summarizes the estimated travel time results of all safety buffer sections. The distribution data is organized using the safety buffer section number as an index, and the index structure supports quick lookup of the corresponding buffer time value by buffer section number. The target travel speed of a safety buffer section is typically higher than that of adjacent high-risk sections. This speed difference allows the robot to accelerate its passage within the safety buffer section to compensate for time lost in the high-risk section; the speed difference ratio is typically between 1.2 and 1.5 times. The duration values ​​of each buffer segment in the buffer duration distribution vary, stemming from different combinations of road segment length and traffic speed. The duration values ​​typically range from 2 to 15 seconds. The buffer duration distribution is grouped according to the association relationships of high-risk road segments. The safe buffer segment duration data corresponding to each high-risk road segment is grouped into the same group. The grouping information is obtained from the associated high-risk road segment number field in the safe buffer segment candidate set. This grouping structure facilitates pairing analysis when there is sufficient time margin for subsequent identification.

[0054] In the buffer duration distribution, locations with sufficient time margin are identified by combining the risk exposure duration, thus constructing a margin location set. Safe buffer segments in the buffer duration distribution whose buffer duration exceeds the risk exposure duration by a certain proportion are marked as having sufficient time margin. The proportion threshold is typically set to 0.8 times, meaning a buffer duration exceeding 80% of the risk exposure duration is considered sufficient. This threshold is based on the minimum time margin required for robot state transitions. Safe buffer segments with sufficient time margin have a good foundation for risk attenuation, providing sufficient time for the robot to adjust its state, including speed recovery, sensor mode switching, and path confirmation. The margin location set includes all safe buffer segment numbers with sufficient time margin. Each element records both the buffer segment number and the buffer duration; the buffer duration value is directly copied from the buffer duration distribution to avoid duplicate calculations. Safe buffer segments with insufficient time margin in the buffer duration distribution are not included in the margin location set. The risk attenuation capability of such buffer segments is limited and needs to be supplemented through other methods, such as cascading multiple buffer segments or extending the passage time of a single buffer segment. High-risk road sections with longer exposure times require greater time margins for adjacent safety buffer zones. In electronic component manufacturing workshops, high-risk road sections in electrostatic discharge (ESD) protection areas typically need to be configured with longer safety buffer zones to ensure sufficient time margins. Insufficient margins may lead to incomplete ESD elimination and product quality issues.

[0055] Risk-bearing capacity is assessed for each safety buffer segment in the margin location set to generate carrying capacity values. The risk-bearing capacity of each safety buffer segment in the margin location set represents the cumulative amount of risk that the segment can absorb and mitigate. The carrying capacity is related to the buffer duration and the segment length. The carrying capacity value quantifies the risk-bearing capacity of each safety buffer segment in the margin location set. The calculation formula is C_load=α×(T_buf / T_ref)+β×(L_buf / L_ref), where C_load is the carrying capacity value, T_buf is the buffer duration read from the margin location set, L_buf is the segment length read from the candidate safety buffer segment set, α and β are weighting coefficients, and the carrying capacity value is a dimensionless normalized value. Safety buffer segments with good traffic conditions and long segment lengths in the margin location set obtain higher carrying capacity values. These segments provide an ideal state recovery environment for the robot. The assessment of carrying capacity also considers dynamic factors of the safety buffer zone, including personnel activity density and logistics vehicle traffic frequency during that period. These dynamic factors are adjusted to the basic carrying capacity value in the form of correction coefficients, with the correction coefficients ranging from 0.8 to 1.2. Safety buffer zones with concentrated margins, located in equipment intervals or dedicated inspection lanes, typically have higher carrying capacity values. In automobile assembly workshops, dedicated robot lanes next to the assembly line are typical areas with high carrying capacity.

[0056] A risk attenuation coefficient is generated by determining the risk spillover threshold based on the carrying capacity value. The carrying capacity value is compared with the accumulated risk corresponding to the risk exposure duration. The accumulated risk is calculated using the formula Q_risk = T_adj × γ, where Q_risk is the accumulated risk, T_adj is the adjusted risk exposure duration read from the risk exposure duration sequence, and γ is set to a base risk accumulation rate of 0.1. The comparison result is used to determine whether there is risk spillover in the safety buffer section. Risk spillover refers to the safety buffer section's carrying capacity being insufficient to completely absorb the accumulated risk generated by adjacent high-risk road sections. The spillover risk will continue to subsequent road sections, forming residual risk, which will reduce the safety margin of subsequent road sections. The risk spillover threshold is set at 90% of the carrying capacity value. Safety buffer sections with a carrying capacity value exceeding 90% of the accumulated risk are considered to have no spillover risk, while those below 90% are considered to have spillover risk. The 90% threshold is set as the safety boundary for risk management. The risk attenuation coefficient is calculated using the formula η = min(C_load / Q_risk, 1) × λ, where η is the risk attenuation coefficient ranging from 0 to 1, C_load is the carrying capacity value, Q_risk is the accumulated risk, and λ is the overflow attenuation factor. When there is no overflow risk, λ is 1, indicating complete attenuation; when there is overflow risk, λ is 0.8, indicating a reduced attenuation effect. The risk attenuation coefficient is output using the safety buffer segment number as an index. The output data covers all safety buffer segments participating in the buffer capacity assessment. The index structure is consistent with the margin location set for easy subsequent pairing. Safety buffer segments with a risk attenuation coefficient lower than 0.6 require increased configuration weight or the introduction of additional buffering measures in subsequent road segment proportion configurations. 0.6 is the lower limit of the effectiveness of risk attenuation.

[0057] Based on the risk attenuation coefficient, a risk-balanced path is formed by configuring the proportion of high-risk road sections and safety buffer sections. Safety buffer sections with higher risk attenuation coefficients receive a larger allocation weight in the road section proportion configuration. This allocation weight determines the actual usable length of the buffer section in the risk-balanced path, and is inversely proportional to the risk attenuation coefficient. The road section proportion configuration determines the length ratio between high-risk road sections and safety buffer sections, following the risk balance principle. The required buffer section length per unit length of high-risk road section is inversely proportional to the risk attenuation coefficient; the lower the risk attenuation coefficient, the longer the buffer section needs to be configured to compensate for insufficient attenuation capacity. Safety buffer sections with a risk attenuation coefficient below 0.6 require increased buffer section length or multiple buffer sections connected in series in the road section proportion configuration to improve the overall attenuation effect. The risk-balanced path connects the configured high-risk road sections and safety buffer sections in spatial order, following a basic pattern of safety buffer section-high-risk road section-safety buffer section. This pattern ensures a buffer transition when the robot enters and leaves high-risk areas. The segment attribute field of the risk balancing path records three items: segment number, segment type, and risk attenuation coefficient. The risk attenuation coefficient is used for time allocation adjustment during subsequent time-series arrangement. After the risk balancing path is generated, a risk distribution balance check is performed. The check standard is that there must be at least one safe buffer segment with a risk attenuation coefficient greater than 0.7 between adjacent high-risk segments.

[0058] The inspection scheduling sequence is generated through time-series orchestration based on the risk balancing path. Each segment of the risk balancing path is converted into a time-dimensional execution sequence according to its spatial arrangement. This execution sequence determines the timing and time allocation for the robot's visits to each segment. The inspection scheduling sequence transforms the spatial information of the risk balancing path into temporal scheduling information, which includes the start time, end time, and duration of each segment. The time-series orchestration allocates execution time based on the travel time budget for each segment in the risk balancing path. The risk attenuation coefficient in the segment attribute field of the risk balancing path is used to adjust the time allocation; segments with lower risk attenuation coefficients have their travel time appropriately increased to reduce actual operating speed. Segments marked as requiring cautious travel in the risk balancing path have an additional time margin reserved in the inspection scheduling sequence; this margin is typically set to 15% of the travel time budget. The total duration of the inspection scheduling sequence equals the sum of the travel times for all segments in the risk balancing path, and the total duration must meet the overall requirements of the time window in the task point constraints. In the risk balancing path, a group of consecutive high-risk road segments is treated as a whole and its execution time is allocated in the inspection scheduling sequence. The execution time of each segment within the group is arranged continuously to avoid interruption. The inspection scheduling sequence is output in the form of a timestamp-indexed instruction list. Each entry in the instruction list includes two items: execution time and target road segment. In inspection tasks in discrete manufacturing workshops, the inspection scheduling sequence typically contains 30 to 100 scheduling instructions.

[0059] Step S150: Path deviation detection is performed based on the inspection scheduling sequence to generate path correction amount, and inspection control instructions are generated based on the path correction amount and the primary and backup path sets.

[0060] In some embodiments, the step of generating a path correction amount based on the inspection scheduling sequence by detecting path deviation includes: extracting preset trajectory points to generate a trajectory reference based on the inspection scheduling sequence; performing real-time pose comparison on the trajectory reference to generate a deviation feature vector; performing cumulative deviation analysis based on the deviation feature vector to generate a deviation trend prediction value; and dynamically adjusting the tolerance range according to the deviation trend prediction value and the inspection scheduling sequence to generate a path correction amount.

[0061] A trajectory baseline is generated by extracting preset trajectory points based on the inspection and dispatch sequence. The start and end coordinates of the target road segments corresponding to each dispatch instruction in the inspection and dispatch sequence serve as the primary source of preset trajectory points, determining the endpoint positions of each road segment. The trajectory baseline expands the discrete target positions in the inspection and dispatch sequence into a continuous reference trajectory. The expansion method uses cubic spline interpolation to generate smooth curves between adjacent preset trajectory points. Execution time information from the inspection and dispatch sequence is synchronously written into the trajectory baseline, giving it both spatiotemporal attributes. These spatiotemporal attributes allow for querying the corresponding expected position and expected attitude at any given time. The density of preset trajectory points is dynamically adjusted based on road segment characteristics. The spacing between trajectory points on straight road segments is set to 1 to 2 meters, while the spacing on turning road segments is shortened to 0.3 to 0.5 meters to improve the fitting accuracy of curve segments. Dispatch instructions originating from high-risk road segments in the inspection and dispatch sequence are marked as high-risk trajectory segments in the trajectory baseline. The marking information of high-risk trajectory segments is used for weighted processing during subsequent deviation feature vector calculation. The trajectory reference is stored in the form of a time-indexed coordinate sequence. Each element in the sequence contains four items: timestamp, position coordinates, attitude angle, and risk marker. In the inspection task of the flexible circuit board production workshop, the trajectory reference usually contains 500 to 2000 preset trajectory points.

[0062] A deviation feature vector is generated by real-time pose comparison with the trajectory reference. The preset trajectory point closest to the current moment in the trajectory reference is used as the reference point for pose comparison. The position coordinates and attitude angles of the reference point are compared with the robot's current pose. The deviation feature vector contains two components: position deviation and attitude deviation. The position deviation is a three-dimensional vector representing the distance difference between the robot and the reference point in the X, Y, and Z directions. The attitude deviation is the difference between the robot's current orientation angle and the expected orientation angle of the reference point. The execution frequency of real-time pose comparison is consistent with the robot's pose sampling frequency; comparison calculation and updating of the deviation feature vector are performed immediately after each sampling. The robot positions between adjacent preset trajectory points in the trajectory reference are determined using linear interpolation to determine the corresponding reference positions. The interpolation calculation is based on the relative positions of the current moment within the time interval of adjacent trajectory points. The formula for calculating the magnitude of the deviation feature vector is as follows: The deviation feature vector is calculated using the following formula: e_x, e_y, and e_z represent the components of the positional deviation along the three coordinate axes, with the magnitude reflecting the degree to which the robot deviates from the expected trajectory. Reference points marked as high-risk in the trajectory baseline are weighted during deviation feature vector calculation. The weighting coefficient is set to 1.2 to 1.5 to amplify the deviation sensitivity of that section. This weighting ensures the robot maintains higher trajectory tracking accuracy in high-risk areas. The deviation feature vectors are stored in a deviation history queue in chronological order, with the queue length set to the data from the most recent 50 to 100 sampling periods. This historical data is used for cumulative deviation analysis.

[0063] Cumulative deviation analysis is performed based on the deviation feature vector to generate deviation trend prediction values. The historical sequence of the deviation feature vector reflects the evolution of robot trajectory deviation over time, and cumulative deviation analysis extracts the changing trend and periodic characteristics of the deviation from the historical sequence. The deviation trend prediction value is obtained by performing linear regression or exponential smoothing on the historical sequence of the deviation feature vector, and the prediction value represents the expected deviation amount after several sampling periods under the current trend. Cumulative deviation analysis distinguishes between two components: systematic deviation and random deviation. Systematic deviation is characterized by the mean of the deviation feature vector continuously deviating from zero, while random deviation is characterized by the fluctuation of the deviation feature vector around the mean. The identification of systematic deviation in the historical sequence of the deviation feature vector adopts the sliding window mean test method. When the mean deviation within the window continuously exceeds the threshold, a systematic deviation is determined to exist. The deviation trend prediction value is calculated using the weighted moving average method, with the weight of recent deviation data higher than that of earlier data. The weight decays exponentially with time intervals, and the decay factor is set to 0.9 to 0.95. Cumulative deviation analysis also calculates the divergence rate of the deviation, defined as the increase in the deviation magnitude per unit time. An excessively high divergence rate triggers an early warning, indicating a potential risk of trajectory tracking loss of control during the inspection task. The deviation trend prediction includes three parts: position trend prediction, attitude trend prediction, and divergence rate. These three parts guide position correction, attitude correction, and tolerance range adjustment, respectively.

[0064] The path correction amount is generated by dynamically adjusting the tolerance interval based on the deviation trend prediction value and the inspection scheduling sequence. The deviation trend prediction value, indicating the future deviation direction and magnitude, is used to plan correction actions in advance, making the correction actions proactive rather than just responding to the current deviation. Each segment in the inspection scheduling sequence has a preset standard tolerance interval, which defines the maximum allowable position deviation and maximum attitude deviation. When the deviation exceeds the tolerance interval, a correction action is triggered. The tolerance interval is dynamically adjusted by shrinking or expanding the standard tolerance interval according to the magnitude and direction of the deviation trend prediction value. When the deviation trend prediction value is large, the tolerance interval is shrunk to trigger correction in advance; when the deviation trend prediction value is small, the tolerance interval can be appropriately expanded to reduce unnecessary correction actions. The tolerance interval corresponding to the scheduling instructions from high-risk segments in the inspection scheduling sequence is pre-shrunken. The shrinkage magnitude is related to the divergence rate in the deviation trend prediction value; the higher the divergence rate, the larger the shrinkage magnitude, ensuring that the robot enters the high-risk area with a small initial deviation. The formula for calculating the path correction amount is as follows: Where ΔP is the path correction, e is the current deviation, and K_p, K_d, and K_i are the proportional, derivative, and integral gain coefficients, respectively. These three gain coefficients are adjusted according to the robot's dynamic response characteristics. The path correction is output after kinematic constraint verification, which includes checking whether the correction exceeds the robot's maximum speed and maximum acceleration limits. If the limits are exceeded, the correction is truncated.

[0065] Inspection control commands are generated based on path correction values ​​and the primary and backup path sets. The position and attitude correction components in the path correction values ​​are converted into control commands executable by the robot, taking into account the robot's kinematic and dynamic constraints. The main path information of the current segment in the primary and backup path sets is combined with the path correction values ​​to generate a corrected target trajectory. The corrected target trajectory eliminates accumulated deviations while maintaining path continuity. Inspection control commands include three categories: speed control commands, steering control commands, and attitude adjustment commands. These three types of commands control the robot's forward speed, direction of travel, and end effector attitude, respectively. When the path correction value is large, the inspection control commands adopt a step-by-step correction strategy, breaking down large corrections into multiple smaller corrections for gradual execution. Step-by-step correction avoids abrupt trajectory changes that could lead to robot instability. Information on alternative paths in the primary and backup path sets serves as a contingency plan for path switching during inspection control command generation. When the path correction value exceeds the correctable range, a path switching command is triggered to guide the robot to an alternative path. Inspection control commands are output to the robot motion controller in the form of command packets. The sending frequency of the command packets is synchronized with the path deviation detection cycle. In wafer handling inspection in semiconductor manufacturing workshops, the sending frequency of command packets usually reaches more than 20 times per second to ensure the real-time performance of trajectory tracking.

[0066] To implement the above-described method embodiments, an intelligent inspection method for industrial robots is proposed to achieve the corresponding functions and technical effects. See also... Figure 2 , Figure 2 This diagram illustrates a structural block diagram of an intelligent inspection system 200 for an industrial robot according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The intelligent inspection system 200 for an industrial robot according to an embodiment of this application includes: The data acquisition module 201 is used to collect inspection task point distribution data, task point constraints and current pose information of the industrial robot, and to construct an inspection status mapping table according to the inspection task point distribution data and the current pose information. The path planning module 202 is used to perform task coverage requirement analysis on the inspection status mapping table to generate a sequence of necessary nodes, and generate path planning instructions according to the degree of matching between the sequence of necessary nodes and the constraints of the task points. The path search module 203 is used to perform multi-branch path search to obtain a reachable path network according to the path planning instructions, divide the reachable path network into main paths and alternative paths to generate a main and alternative path set, determine the passage difficulty coefficient by road segment decomposition according to the main and alternative path set and the current pose information, and generate path execution parameters according to the passage difficulty coefficient. Risk scheduling module 204 is used to perform risk level assessment and generate regional risk identifiers according to the path execution parameters, divide high-risk road segments and safe road segments based on the regional risk identifiers, insert and arrange the high-risk road segments and the safe road segments at intervals to generate risk balance paths, and perform time-series arrangement based on the risk balance paths to generate inspection scheduling sequences. The instruction generation module 205 is used to generate a path correction amount based on the path deviation detection of the inspection scheduling sequence, and generate an inspection control instruction based on the path correction amount and the primary and backup path set.

[0067] The aforementioned intelligent inspection system 200 for industrial robots can implement an intelligent inspection method for industrial robots according to the above-described method embodiments. The options described in the above method embodiments are also applicable to this embodiment and will not be detailed here. The remaining content of this application's embodiments can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0068] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

Claims

1. An intelligent inspection method for industrial robots, characterized in that, include: Collect inspection task point distribution data, task point constraints and current pose information of industrial robot, and construct inspection state mapping table according to the inspection task point distribution data and current pose information; Based on the inspection status mapping table, a task coverage requirement analysis is performed to generate a sequence of necessary nodes, and a path planning instruction is generated according to the degree of matching between the sequence of necessary nodes and the constraints of the task points. Based on the path planning instructions, a multi-branch path search is performed to obtain a reachable path network. The reachable path network is divided into a main path and alternative paths to generate a main and alternative path set. Based on the main and alternative path set and the current pose information, the passage difficulty coefficient is determined by road segment decomposition. Path execution parameters are generated based on the passage difficulty coefficient. Based on the path execution parameters, a risk level assessment is performed to generate a regional risk identifier. Based on the regional risk identifier, high-risk road segments and safe road segments are divided. The high-risk road segments and the safe road segments are interleaved and arranged to generate a risk-balanced path. Based on the risk-balanced path, a time sequence is arranged to generate an inspection and scheduling sequence. Based on the inspection scheduling sequence, path deviation is detected to generate path correction amount, and inspection control instructions are generated according to the path correction amount and the primary and backup path sets.

2. The method according to claim 1, characterized in that, The step of constructing an inspection status mapping table based on the inspection task point distribution data and the current pose information includes: Based on the inspection task point distribution data, spatial coordinates of the task points are extracted to construct a coordinate feature set; The coordinate feature set is associated with the current pose information to construct a pose-task point mapping relationship; The robot's motion state is parsed from the current pose information to generate obstacle avoidance parameters; Based on the obstacle avoidance parameters, the pose-task point mapping relationship is corrected to generate an inspection status mapping table.

3. The method according to claim 1, characterized in that, The step of generating path planning instructions based on the degree of matching between the required node sequence and the task point constraints includes: Based on the task point constraints, the sequence of necessary nodes is divided into a strongly constrained node group and a weakly constrained node group. Constraint conflict detection is performed on the strong constraint node group and the weak constraint node group to generate conflict node pairs; The conflicting node pairs are subjected to priority arbitration to generate a node sequence after conflict resolution; Path planning instructions are generated based on the node sequence after conflict resolution.

4. The method according to claim 1, characterized in that, The step of dividing the reachable path network into primary and alternative paths to generate a primary and alternative path set includes: For the reachable path network, path cost evaluation and ranking are performed to identify the optimal path candidate set; The optimal path candidate set is subjected to path overlap detection to generate an overlapping road segment distribution; Obstacle accessibility analysis is performed on the overlapping road segment distribution to generate traffic conflict markers; Based on the traffic conflict markers, low-conflict alternative paths are retained to form a primary and backup path set.

5. The method according to claim 1, characterized in that, The step of determining the passage difficulty coefficient by decomposing road segments based on the primary and backup path sets and the current pose information includes: Based on the primary and backup path sets, generate road segmentation points; The road segments are divided based on the road segmentation points, and the turning angles and road segment lengths are extracted to generate geometric feature parameters; Based on the current pose information, the robot's motion state is analyzed to generate motion state parameters; A kinematic constraint analysis is performed on the geometric feature parameters and the motion state parameters to form a passage difficulty coefficient.

6. The method according to claim 1, characterized in that, The step of interleaving high-risk road segments with safe road segments to generate a risk-balanced path includes: Risk exposure duration is generated based on the high-risk road sections; Select safe buffer zones adjacent to the high-risk road sections from the safe road sections; The buffer capacity of the safety buffer segment is assessed in conjunction with the risk exposure duration to generate a risk attenuation coefficient. Based on the risk attenuation coefficient, the high-risk road segment and the safety buffer segment are configured in proportion to form a risk-balanced path.

7. The method according to claim 1, characterized in that, The step of generating path correction amount based on the inspection scheduling sequence includes: Based on the inspection and scheduling sequence, preset trajectory points are extracted to generate a trajectory reference; Real-time pose comparison of the trajectory reference is performed to generate a deviation feature vector; Based on the aforementioned deviation feature vector, cumulative deviation analysis is performed to generate a deviation trend prediction value; Based on the predicted deviation trend value and the inspection scheduling sequence, the tolerance range is dynamically adjusted to generate the path correction amount.

8. The method according to claim 4, characterized in that, The step of performing path overlap detection on the optimal path candidate set to generate an overlapping road segment distribution includes: Based on the optimal path candidate set, path overlap detection is performed to determine overlapping road segments; In the overlapping road segments, trace the shared nodes to construct a shared node distribution map; Based on the shared node distribution map, highly overlapping regions and low overlapping regions are identified to generate overlapping classification labels; The overlapping road segment distribution is generated by combining the overlapping classification markers with the shared node distribution map.

9. The method according to claim 6, characterized in that, The step of assessing the buffer capacity of the safety buffer segment in conjunction with the risk exposure duration to generate a risk attenuation coefficient includes: The passage time of the safety buffer section is estimated to generate a buffer time distribution; In the buffer duration distribution, a margin location set is constructed by combining the risk exposure duration to identify locations with sufficient time margin; Risk carrying capacity assessment is performed on the aforementioned margin location set to generate carrying capacity values; The risk overflow threshold is determined based on the carrying capacity value to generate a risk attenuation coefficient.

10. An intelligent inspection system for industrial robots, characterized in that, include: The data acquisition module is used to collect inspection task point distribution data, task point constraints and current pose information of the industrial robot, and to construct an inspection status mapping table based on the inspection task point distribution data and the current pose information. The path planning module is used to perform task coverage requirement analysis on the inspection status mapping table to generate a sequence of necessary nodes, and generate path planning instructions according to the degree of matching between the sequence of necessary nodes and the constraints of the task points. The path search module is used to perform multi-branch path search to obtain a reachable path network based on the path planning instructions, divide the reachable path network into main paths and alternative paths to generate a main and alternative path set, determine the passage difficulty coefficient by decomposing the road segments according to the main and alternative path set and the current pose information, and generate path execution parameters based on the passage difficulty coefficient. The risk scheduling module is used to perform risk level assessment and generate regional risk identifiers according to the path execution parameters, divide high-risk road segments and safe road segments based on the regional risk identifiers, insert and arrange the high-risk road segments and the safe road segments at intervals to generate risk balance paths, and generate inspection scheduling sequences based on the risk balance paths. The instruction generation module is used to generate a path correction amount based on the path deviation detection of the inspection scheduling sequence, and generate an inspection control instruction based on the path correction amount and the primary and backup path sets.

Images