A yard inspection path dynamic optimization method based on a device management platform

Abstract

The application discloses a yard inspection path dynamic optimization method based on a device management platform, and relates to the field of yard inspection; by analyzing target environment data collected by fixed monitoring devices and robots and processed by edge computing, the comprehensive risk total value of the corresponding grid node is derived in real time, and the latest value of the risk priority index and the environment fitness evaluation factor is dynamically determined accordingly; for each robot, the latest value of the grid node related to the unexecuted segment of the current inspection path is obtained, the unexecuted segment is evaluated and the current inspection path is updated in real time based on the latest value. This dynamic closed-loop mechanism driven by real-time data breaks the rigid mode of traditional fixed route or static traversal strategy relying only on fixed map information, significantly improving the environmental adaptability of the inspection path and the safety of the operation process.

Description

Technical Field

[0001] This invention relates to the field of yard inspection, and in particular to a method for dynamic optimization of yard inspection paths based on an equipment management platform. Background Technology

[0002] In port hazardous materials storage yards, towering container stacks form complex physical barriers that easily block IoT sensor signals, creating communication blind spots. Attempting to achieve full coverage without blind spots by increasing the density of fixed monitoring equipment would face an exponential increase in equipment and maintenance costs, making it difficult to achieve both economically and engineeringally. Therefore, introducing robots to conduct mobile patrols in areas without fixed monitoring equipment has become an inevitable choice to fill the blind spots of fixed sensing networks and achieve comprehensive risk monitoring. However, this hybrid model of "fixed locations + mobile inspection" also poses a severe challenge to path planning: robots not only need to go deep into areas not covered by fixed monitoring equipment to perform tasks, but also need to ensure the safety and controllability of their own movement in complex dynamic environments.

[0003] In automated inspection scenarios at hazardous materials storage yards, existing path planning technologies suffer from insufficient dynamic risk avoidance capabilities and a lack of safety buffer mechanisms. Traditional solutions often employ preset fixed routes or static global traversal strategies, with path decision-making logic frequently based on fixed map information. They lack mechanisms for perceiving and responding to real-time dynamic risks such as gas leaks, temperature anomalies, fires, smoke, and unauthorized personnel intrusion. When the risk level in a certain area of ​​the storage yard suddenly increases, existing systems struggle to dynamically adjust their path planning strategies, causing robots to continue proceeding according to the original plan and failing to proactively avoid newly emerging high-risk areas.

[0004] More importantly, even if some existing technologies can identify high-risk nodes and mark them as impassable, their algorithmic logic is limited to point-to-point obstacle avoidance. That is, as long as the central grid node along the path is not marked as dangerous, the algorithm determines the path is feasible. This mechanism completely ignores the radiating influence of hazards on surrounding nodes (such as the spread of toxic gases, the heat radiation zone of a fire, or the range of an explosion's shock wave). This results in planned paths often running close to the edge of the hazard, lacking necessary safety buffer distances. Once a dangerous situation suddenly spreads (such as gas dispersion or fire spread), robots in high-risk neighboring areas are highly vulnerable to direct threats or even damage, causing inspection missions to be interrupted. Therefore, due to the lack of a dynamic penalty mechanism for path costs based on neighborhood perception, existing technologies struggle to convert high-risk neighborhoods into economic costs for path selection during the path search phase. They cannot guide the algorithm to proactively avoid potentially risky road sections where the path nodes themselves appear safe but neighboring nodes are high-risk, making it difficult to achieve truly proactive risk avoidance and safe inspection. Summary of the Invention

[0005] To overcome the shortcomings of existing path planning technologies, such as lack of awareness of neighborhood risk radiation and difficulty in converting high-risk neighborhoods into economic costs for path selection to guide robots to actively avoid high-risk areas, this invention proposes a dynamic optimization method for yard inspection paths based on an equipment management platform, including:

[0006] The storage yard area is discretized into a grid node diagram. Initial values ​​are set for the risk priority index of each grid node, and the environmental adaptability assessment factor of each grid node is initialized. The initial value of the risk priority index represents the static risk level, and the environmental adaptability assessment factor represents the feasibility of passage.

[0007] Based on the initial values ​​of the risk priority index and the environmental adaptability assessment factor of each grid node, initial inspection paths are generated for multiple robots deployed in the yard area; each initial inspection path is set as the current inspection path of the corresponding robot.

[0008] By deploying fixed monitoring equipment in the yard area and robots traveling along their respective current inspection paths, multi-dimensional environmental data of grid nodes within their respective monitoring ranges are collected in real time. Edge computing is performed on the multi-dimensional environmental data to obtain the target environmental data, which is then uploaded to the equipment management platform.

[0009] The target environment data is analyzed through the equipment management platform, and the comprehensive risk value of the corresponding grid node is derived based on the analysis results. The latest values ​​of the risk priority index and environmental adaptability assessment factor of the corresponding grid node are determined according to the comprehensive risk value. For each robot, the latest values ​​of the risk priority index and environmental adaptability assessment factor of the grid node related to its current inspection path are obtained. Based on the obtained latest values, the unexecuted segments of its current inspection path are evaluated, and the corresponding current inspection path is updated based on the evaluation results.

[0010] Furthermore, setting initial values ​​for the risk priority index of each grid node specifically includes:

[0011] Obtain the storage distribution information of goods within the yard area, and define the risk level of the corresponding grid node based on the hazardous characteristics of the goods within the coverage area of ​​the grid node; wherein, the risk level includes at least high risk level, medium risk level, low risk level and safe level;

[0012] Based on the risk level of the grid node, an initial value for its risk priority index is set. The higher the risk level, the larger the initial value of the risk priority index corresponding to the grid node.

[0013] Furthermore, the multidimensional environmental data includes:

[0014] Concentrations of various gases, ambient temperature, and visual image data.

[0015] Furthermore, the step of performing edge computing on the multidimensional environmental data to obtain the target environmental data specifically involves:

[0016] The concentrations of various gases and ambient temperature are collected and denoised, and outlier values ​​are removed to generate standardized environmental parameter values.

[0017] Lightweight vision algorithms are used to perform real-time detection and processing of acquired visual image data.

[0018] When at least one target is detected, semantic label data containing the type identifier of the corresponding target and the location coordinates of the grid node where it is located is generated; the type identifier of the target is either a specific category identifier of an obstacle or a category identifier of fireworks;

[0019] When a person is detected, semantic label data containing the location coordinates of the grid node where each person is located and the number of people is generated, and the corresponding face region is simultaneously extracted from the visual image data for each detected person to generate the corresponding face bounding box.

[0020] When no target objects or people are detected, semantic label data representing a normal scene is generated.

[0021] The standardized environmental parameter values, the generated semantic label data, and the face bounding image generated when a person is detected are spatiotemporally labeled and encapsulated to form the target environmental data.

[0022] Furthermore, the step of parsing the target environmental data through the equipment management platform and deriving the comprehensive total risk value of the corresponding grid nodes based on the parsing results specifically involves:

[0023] Analyze target environment data;

[0024] For each grid node, perform the following risk score generation steps:

[0025] If the concentration of any gas or the ambient temperature in the standardized environmental parameter values ​​exceeds its corresponding preset environmental safety threshold, a corresponding environmental risk score is generated.

[0026] If the semantic tag data contains a category identifier for fireworks, a fire risk score is generated.

[0027] If the semantic label data contains a specific category identifier for the obstacle, the degree of traffic obstruction is assessed based on that specific category identifier, and an obstacle risk score is generated based on the assessment result.

[0028] If the semantic tag data contains the location coordinates of a person, and the identity is determined to be an unauthorized intrusion after identity comparison based on the face bounding box, then an intrusion risk score is generated.

[0029] The environmental risk score, fire risk score, obstacle risk score, and intrusion risk score are summed to obtain the total comprehensive risk value of the grid node.

[0030] Furthermore, the latest values ​​for environmental fitness assessment factors are determined, specifically including:

[0031] Obtain the current values ​​of the multidimensional environmental feature parameters for each grid node, wherein the current values ​​of the multidimensional environmental feature parameters include at least:

[0032] The accessibility attribute of the corresponding grid node, wherein the accessibility attribute indicates whether the grid node belongs to a passable area or an impassable area;

[0033] The coverage of the corresponding grid nodes within the field of view of the fixed visual monitoring equipment;

[0034] The risk cost of the corresponding grid node; the risk cost is determined by substituting the total comprehensive risk value into a preset mapping relationship between the total comprehensive risk value and the risk cost;

[0035] In addition, when the passage attribute represents a passable area, it also includes the actual passage width corresponding to the grid node;

[0036] For each grid node, perform the following judgment and calculation steps:

[0037] If the accessibility attribute indicates that a grid node is an impassable area, then the current value of the environmental fitness evaluation factor of that grid node is set to the maximum value indicating impassability.

[0038] If the accessibility attribute indicates that a grid node is a passable area, then the latest value of the environmental fitness assessment factor of that grid node is calculated based on the current value of the multidimensional environmental characteristic parameters.

[0039] Furthermore, the calculation of the latest value of the environmental fitness evaluation factor for the grid node based on the current values ​​of the multidimensional environmental characteristic parameters specifically involves:

[0040] Based on the actual passage width, a width factor is calculated, wherein the larger the actual passage width, the smaller the value of the width factor, so as to characterize the lower passage cost of the grid node;

[0041] Based on the coverage and risk cost, the corresponding visibility factor and risk factor are calculated respectively; wherein, the higher the coverage, the smaller the value of the visibility factor; and the higher the risk cost, the larger the value of the risk factor.

[0042] The width factor, visibility factor, and risk factor are weighted and fused to generate the latest value of the environmental adaptability assessment factor for the grid node.

[0043] Furthermore, determine the latest value of the risk priority index, specifically including:

[0044] The total comprehensive risk value is substituted into the preset mapping relationship between the total comprehensive risk value and the risk priority index to obtain the latest value of the risk priority index; wherein, the larger the total comprehensive risk value, the larger the corresponding value of the risk priority index.

[0045] Furthermore, based on the initial values ​​of the risk priority index and the environmental adaptability assessment factor for each grid node, initial inspection paths are generated for multiple robots deployed within the yard area, specifically including:

[0046] Based on the initial values ​​of the risk priority index and the initial values ​​of the environmental adaptability assessment factor of all grid nodes in the yard area, a global access cost map is constructed.

[0047] For each robot, its dedicated inspection area, preset starting point, and multiple mandatory inspection nodes to be traversed within the area are determined. Taking the preset starting point as the starting node, and ensuring that the path covers all the mandatory inspection nodes, the preset inspection target point within the dedicated inspection area is taken as the ending node. Based on the global access cost map, a preset global path search algorithm is used to search for the path with the minimum cumulative cost within the corresponding dedicated inspection area, which is then used as the initial inspection path for that robot.

[0048] The search space of the preset global path search algorithm is limited to a connected subgraph consisting of all grid nodes with the passability attribute of passable areas within the dedicated inspection area.

[0049] Furthermore, the construction of a global access cost map based on the initial values ​​of the risk priority index and the environmental adaptability assessment factor for all grid nodes within the storage yard area specifically includes:

[0050] Traverse each grid node within the storage area and perform the following cost assignment steps:

[0051] If the initial value of the environmental fitness evaluation factor of the grid node is a maximum value that represents impassability, then the passage cost of the grid node is set to a preset prohibition cost.

[0052] If the initial value of the environmental adaptability assessment factor of the grid node is not a maximum value that represents impassability, then: a weighted calculation is performed based on the initial value of the environmental adaptability assessment factor and the initial value of the risk priority index, and the calculation result is used as the passage cost of the grid node.

[0053] The smaller the initial value of the environmental adaptability assessment factor, the lower the corresponding passage cost;

[0054] The global travel cost map is composed of the travel costs of all grid nodes.

[0055] Furthermore, obtain the latest values ​​of the risk priority index and environmental adaptability assessment factor of the grid nodes related to its current inspection path, specifically including:

[0056] Identify the unexecuted segments of the current inspection path of each robot, and take all the grid nodes included in the unexecuted segments, as well as the adjacent nodes of all the grid nodes, as the grid nodes related to its current inspection path.

[0057] The relevant grid nodes are divided into updated grid nodes and non-updated grid nodes; wherein, the updated grid nodes are grid nodes that correspond to multi-dimensional environmental data at the current time, and the non-updated grid nodes are grid nodes that do not correspond to multi-dimensional environmental data at the current time.

[0058] For grid nodes that have been updated, extract the latest values ​​of the risk priority index and environmental adaptability assessment factor for that grid node;

[0059] For grid nodes that have not been updated, the latest values ​​of the risk priority index and environmental adaptability assessment factor of that grid node are confirmed as the historical values ​​of the previous moment; if there are no historical values ​​of the previous moment, they are confirmed as the corresponding initial values.

[0060] Furthermore, based on the latest acquired value, the unexecuted segments of its current inspection path are evaluated, and the corresponding current inspection path is updated based on the evaluation results, including:

[0061] Based on the latest value of the updated grid node, recalculate the current passage cost of the grid node; for the unupdated grid node, confirm its passage cost at the previous moment as the current passage cost.

[0062] Traverse each grid node included in the unexecuted segment of the current inspection path and determine whether there are high-risk nodes among its adjacent grid nodes. If so, increase the current passage cost of that grid node as a penalty.

[0063] Update the global travel cost map based on the updated current travel cost;

[0064] Determine whether there are high-risk nodes among the grid nodes related to the current inspection path. If so, take the robot's current real-time position as the starting node. Meanwhile, while ensuring coverage of the remaining untraversed mandatory inspection nodes, take the preset inspection target point in the dedicated inspection area as the ending node. Based on the updated global passage cost map, use the preset global path search algorithm to re-search in the corresponding dedicated inspection area to obtain the optimized path with the minimum cumulative cost, and set the optimized path as the new current inspection path.

[0065] Furthermore, the conditions for determining the high-risk nodes are as follows:

[0066] The current passage cost exceeds the preset cost safety threshold, or the latest value of the risk priority index exceeds the preset emergency threshold.

[0067] Furthermore, the penalty increase on the current passage cost of the grid node is specifically as follows:

[0068] Obtain the current passage cost of this grid node as the base passage cost;

[0069] Find the high-risk nodes among the neighboring grid nodes of the given grid node;

[0070] Based on the current passage cost of each high-risk node, a neighborhood risk penalty value is calculated; wherein, the higher the current passage cost of the high-risk node, the higher the neighborhood risk penalty value.

[0071] The basic passage cost is accumulated or weighted and fused with the neighborhood risk penalty value, and the current passage cost of the grid node is updated based on the result of the accumulation or weighted fusion.

[0072] Furthermore, the preset mapping relationship between the total comprehensive risk value and the risk cost is configured as a piecewise function:

[0073] When the total comprehensive risk value is less than a preset threshold, the risk cost increases linearly with the total comprehensive risk value.

[0074] When the total comprehensive risk value is greater than or equal to the preset critical value, the risk cost increases non-linearly with the total comprehensive risk value, and its growth rate increases with the increase of the total comprehensive risk value.

[0075] Furthermore, the dynamic optimization method for the yard inspection path also includes triggering an early warning mechanism when the latest value of the risk priority index of any grid node exceeds a preset early warning threshold.

[0076] Furthermore, the triggering of the early warning mechanism when the latest value of the risk priority index of any grid node exceeds a preset early warning threshold specifically includes:

[0077] When the latest value of the risk priority index exceeds the first-level preset warning threshold but does not reach the second-level preset warning threshold, the corresponding grid node is marked with the first preset warning color on the monitoring interface of the equipment management platform, and the latest value of the risk priority index is displayed as a prompt-level warning.

[0078] When the latest value of the risk priority index reaches or exceeds the second-level preset warning threshold, the corresponding grid node is marked with the second preset warning color, and an alarm message containing the location of the grid node and the latest value of the risk priority index is pushed to the preset safety management personnel terminal as an emergency alarm.

[0079] Compared with the prior art, the present invention has at least the following beneficial effects:

[0080] (1) This invention analyzes the target environment data collected by fixed monitoring equipment and robots and processed by edge computing to derive the comprehensive total risk value of the corresponding grid nodes in real time, and dynamically determines the latest values ​​of the risk priority index and environmental adaptability assessment factor accordingly. For each robot, the latest values ​​of the grid nodes related to the unexecuted segments of its current inspection path are obtained, and the unexecuted segments are evaluated in real time based on the latest values ​​to update the current inspection path. This dynamic closed-loop mechanism based on real-time data breaks the rigid mode of traditional fixed route or static traversal strategies that only rely on fixed map information, ensuring that the robot always plans its route based on the latest risk priority index and environmental adaptability assessment factor reflecting the current environmental state, which significantly improves the environmental adaptability of the inspection path and the safety of the operation process.

[0081] (2) This invention establishes a dynamic penalty mechanism for path costs based on neighborhood perception. By quantifying the radiation impact of a hazard source on the surrounding environment into the economic cost of path search, an active safety buffer barrier is constructed. Specifically, when evaluating the unexecuted segment of the current inspection path, the platform first recalculates the current passage cost of each grid node in the unexecuted segment based on the latest value to accurately reflect its inherent risk. Then, it traverses its adjacent grid nodes to identify whether there are high-risk nodes. If there are, it calculates the neighborhood risk penalty value based on the current passage cost of the adjacent high-risk nodes and accumulates or weights and merges this penalty value into the passage cost of the current grid node. This mechanism successfully transforms the neighborhood radiation impact such as the toxic gas diffusion range and the fire heat radiation zone into high-cost signals that the algorithm can identify. It guides the global path search algorithm to actively avoid potential risky road segments where "the path node itself seems safe but the neighboring nodes are high-risk" in the process of finding the path with the minimum cumulative cost. The inspection path planned in this way has the necessary safety buffer distance, which avoids the robot from driving close to the edge of the danger source. It effectively prevents the robot from being directly threatened or even damaged due to the sudden spread of dangerous situations, and achieves true active risk avoidance and safe inspection.

[0082] (3) This invention constructs a multi-dimensional environmental feature parameter fusion-based passage cost assessment system. By comprehensively considering the actual passage width, the coverage of the fixed visual monitoring equipment's field of view, and the risk cost, it significantly improves the inherent safety and controllability of the inspection process. The platform calculates the width factor based on the actual passage width, the visibility factor based on the coverage, and combines the risk factor for weighted fusion to generate the environmental adaptability assessment factor. This assessment logic guides the robot to prioritize paths with wide passages, effective video surveillance coverage (non-blind spots), and controllable risks during movement. This avoids the risk of the robot scratching or getting stuck in narrow bottleneck areas, and also prevents the risk of loss of control due to the backend being unable to immediately confirm the situation when an anomaly occurs because it enters a video surveillance blind spot.

[0083] (4) This invention realizes a closed-loop management of the entire process of risk priority index from initial setting to real-time dynamic updating, ensuring the timeliness and accuracy of the global access cost map in reflecting the state of the yard environment. The platform first sets the initial value of the risk priority index for each grid node based on the distribution information of goods containing hazardous characteristics within the yard area, providing a basic risk reference for path planning; then, by analyzing the target environment data in real time, it dynamically derives the total comprehensive risk value and determines the latest value of the risk priority index, and uses this latest value to update the historical values ​​of each grid node in real time. This mechanism based on real-time data continuous iterative updates enables the global access cost map to respond instantly to changes in the yard environment (such as gas concentration fluctuations, temperature anomalies, or the appearance of obstacles), ensuring that the robot always makes decisions based on values ​​reflecting the real risk situation at the current moment when searching for paths.

[0084] (5) The early warning mechanism provided by this invention enables the visualization of the status of prompt-level early warnings and the instant push of emergency alarms, ensuring that managers can simultaneously grasp the latest value and precise location of the risk priority index, effectively improving the efficiency of handling emergencies. This hierarchical processing strategy ensures the accurate delivery of early warning information at different levels, providing a reliable basis for the rapid handling of emergencies. Attached Figure Description

[0085] Figure 1 This is a flowchart of a dynamic optimization method for yard inspection paths based on an equipment management platform, according to an embodiment of the present invention. Detailed Implementation

[0086] The following are specific embodiments of the present invention, which are described in conjunction with the accompanying drawings. However, the present invention is not limited to these embodiments.

[0087] To overcome the shortcomings of existing path planning technologies, such as the lack of perception of neighborhood risk radiation and the difficulty in converting high-risk neighborhoods into economic costs for path selection to guide robots to actively avoid high-risk areas, Figure 1 As shown, this invention proposes a dynamic optimization method for yard inspection paths based on an equipment management platform, including:

[0088] The storage yard area is discretized into a grid node diagram. Initial values ​​are set for the risk priority index of each grid node, and the environmental adaptability assessment factor of each grid node is initialized. The initial value of the risk priority index represents the static risk level, and the environmental adaptability assessment factor represents the feasibility of passage.

[0089] The initial value for the risk priority index of each grid node is set, specifically including:

[0090] Obtain the storage distribution information of goods within the yard area, and define the risk level of the corresponding grid node based on the hazardous characteristics of the goods within the coverage area of ​​the grid node; wherein, the risk level includes at least high risk level, medium risk level, low risk level and safe level;

[0091] Based on the risk level of the grid node, an initial value for its risk priority index is set. The higher the risk level, the larger the initial value of the risk priority index corresponding to the grid node.

[0092] Based on the initial values ​​of the risk priority index and the environmental adaptability assessment factor of each grid node, initial inspection paths are generated for multiple robots deployed in the yard area; each initial inspection path is set as the current inspection path of the corresponding robot.

[0093] The initial values ​​of the risk priority index and environmental adaptability assessment factor based on each grid node are used to generate initial inspection paths for multiple robots deployed in the yard area, specifically including:

[0094] Based on the initial values ​​of the risk priority index and the initial values ​​of the environmental adaptability assessment factor of all grid nodes in the yard area, a global access cost map is constructed.

[0095] The construction of a global access cost map based on the initial values ​​of the risk priority index and the environmental adaptability assessment factor for all grid nodes within the storage yard area specifically includes:

[0096] Traverse each grid node within the storage area and perform the following cost assignment steps:

[0097] If the initial value of the environmental fitness evaluation factor of the grid node is a maximum value that represents impassability, then the passage cost of the grid node is set to a preset prohibition cost (the prohibition cost value is a maximum constant that is far greater than the sum of the cumulative costs of normal paths, so as to ensure that the global path search algorithm automatically excludes the grid node when calculating the minimum cumulative cost).

[0098] If the initial value of the environmental adaptability assessment factor of the grid node is not a maximum value that represents impassability, then: a weighted calculation is performed based on the initial value of the environmental adaptability assessment factor and the initial value of the risk priority index, and the calculation result is used as the passage cost of the grid node.

[0099] The smaller the initial value of the environmental adaptability assessment factor, the lower the corresponding passage cost;

[0100] The global travel cost map is composed of the travel costs of all grid nodes.

[0101] For each robot, its dedicated inspection area, preset starting point, and multiple mandatory inspection nodes to be traversed within the area are determined. Taking the preset starting point as the starting node, and ensuring that the path covers all the mandatory inspection nodes, the preset inspection target point within the dedicated inspection area is taken as the ending node. Based on the global access cost map, a preset global path search algorithm is used to search for the path with the minimum cumulative cost within the corresponding dedicated inspection area, which is then used as the initial inspection path for that robot.

[0102] The search space of the preset global path search algorithm is limited to a connected subgraph consisting of all grid nodes with the passability attribute of passable areas within the dedicated inspection area.

[0103] It should be noted that the preset global path search algorithm can employ mature multi-waypoint global path planning algorithms in this field, such as the "Traveling Salesman Problem (TSP) solution strategy based on Algorithm A" or the "Dijkstra's algorithm for multiple objectives". Such algorithms can automatically calculate an optimal traversal path that covers all necessary inspection nodes and minimizes cumulative cost within the search space defined by the "connected subgraph composed of grid nodes of traversable areas". This invention does not limit the specific algorithm implementation; any grid map global planning algorithm that supports multiple constraints (full coverage, minimum cost) can be applied to this solution.

[0104] By deploying fixed monitoring equipment in the yard area and robots traveling along their respective current inspection paths, multi-dimensional environmental data of grid nodes within their respective monitoring ranges are collected in real time. Edge computing is performed on the multi-dimensional environmental data to obtain the target environmental data, which is then uploaded to the equipment management platform.

[0105] The multidimensional environmental data includes:

[0106] Concentrations of various gases, ambient temperature, and visual image data.

[0107] It should be noted that the concentrations of the various gases can be configured according to the characteristics of the goods stored in the storage yard. For example, for areas storing flammable and explosive goods, the concentrations of various gases may include flammable gases such as methane and propane; for areas storing chemicals or in enclosed spaces, the concentrations of toxic or asphyxiating gases such as carbon monoxide, hydrogen sulfide, and oxygen may be included.

[0108] The process of obtaining target environmental data by performing edge computing on multidimensional environmental data specifically involves:

[0109] The concentrations of various gases and ambient temperature are collected and denoised, and outlier values ​​are removed to generate standardized environmental parameter values.

[0110] Lightweight vision algorithms are used to perform real-time detection and processing of acquired visual image data.

[0111] When at least one target is detected, semantic label data containing the type identifier of the corresponding target and the location coordinates of the grid node where it is located is generated; the type identifier of the target is either a specific category identifier of an obstacle or a category identifier of fireworks;

[0112] When a person is detected, semantic label data containing the location coordinates of the grid node where each person is located and the number of people is generated, and the corresponding face region is simultaneously extracted from the visual image data for each detected person to generate the corresponding face bounding box.

[0113] When no target objects or people are detected, semantic label data representing a normal scene is generated.

[0114] The standardized environmental parameter values, the generated semantic label data, and the face bounding image generated when a person is detected are spatiotemporally labeled and encapsulated to form the target environmental data.

[0115] The target environment data is analyzed through the equipment management platform, and the comprehensive risk value of the corresponding grid node is derived based on the analysis results. The latest values ​​of the risk priority index and environmental adaptability assessment factor of the corresponding grid node are determined according to the comprehensive risk value. For each robot, the latest values ​​of the risk priority index and environmental adaptability assessment factor of the grid node related to its current inspection path are obtained. Based on the obtained latest values, the unexecuted segments of its current inspection path are evaluated, and the corresponding current inspection path is updated based on the evaluation results.

[0116] The process of parsing target environmental data through the device management platform and deriving the overall risk value of the corresponding grid nodes based on the parsing results is as follows:

[0117] Analyze target environment data;

[0118] For each grid node, perform the following risk score generation steps:

[0119] If the concentration of any gas or the ambient temperature in the standardized environmental parameter values ​​exceeds its corresponding preset environmental safety threshold, a corresponding environmental risk score is generated.

[0120] If the semantic tag data contains a category identifier for fireworks, a fire risk score is generated.

[0121] If the semantic label data contains a specific category identifier for the obstacle, the degree of traffic obstruction is assessed based on that specific category identifier, and an obstacle risk score is generated based on the assessment result.

[0122] Specifically, assessing the degree of traffic congestion based on specific category identifiers means that the equipment management platform pre-sets a mapping relationship between different obstacle category identifiers and the degree of traffic congestion. For example, when the specific category identifier in the semantic tag data is "small scattered objects," the platform assesses that it only causes minor local congestion, generating a low obstacle risk score; when it is identified as "large goods," the platform assesses that it causes complete blockage or severe congestion in the area, generating an extremely high obstacle risk score. Through this category-based differentiated assessment, the platform can accurately quantify the actual impact of different obstacles on robot passage, thereby making reasonable avoidance or detour decisions in path planning.

[0123] If the semantic tag data contains the location coordinates of a person, and the identity is determined to be an unauthorized intrusion after identity comparison based on the face bounding box, then an intrusion risk score is generated.

[0124] It should be noted that the determination of unauthorized intrusion refers to the device management platform performing real-time comparison of detected facial features with a preset personnel permission database.

[0125] If the comparison results show that the person is an authorized staff member (i.e., exists in the whitelist) and is located in the area where they are allowed to work, no intrusion risk score will be generated (or the score will be 0), and it will be considered normal operation;

[0126] If the comparison results show that the person is not on the authorized list (i.e., a stranger), or if the person is on the list but appears in an unauthorized area (such as a high-risk restricted area), it is determined to be an unauthorized intrusion.

[0127] Once an unauthorized intrusion is identified, an intrusion risk score will be generated.

[0128] The total risk value of the grid node is obtained by summing up the environmental risk scores, fire risk scores, obstacle risk scores, and intrusion risk scores.

[0129] It should be noted that the specific values ​​of various risk scores (such as environmental risk scores, fire risk scores, etc.) involved in this invention, as well as the mapping relationship data between the total comprehensive risk value, risk cost, and risk priority index, are not fixed constants, but can be flexibly configured according to the actual yard safety management specifications, cargo characteristics, and robot performance. However, this configuration must follow the principle of "high-risk features must reach the threshold": that is, for clearly high-risk situations such as semantic tag data containing smoke and fire category identifiers or standardized environmental parameter values ​​that are seriously exceeded, the corresponding risk score and the mapped risk cost must be set to sufficiently high values ​​to ensure that the calculated current passage cost is necessarily greater than the preset cost safety threshold, or that the latest value of its risk priority index necessarily exceeds the preset emergency threshold. Only in this way can it be ensured that such grid nodes are accurately identified as high-risk nodes by the equipment management platform, thereby triggering the penalty increase mechanism and path replanning process.

[0130] In a specific configuration example, if a smoke / fire category identifier is detected, an extremely high fire risk score (e.g., 100 points) can be directly assigned, causing the accumulated total risk value to far exceed the preset threshold. Simultaneously, the mapping relationship between the total risk value and the risk cost is set as a piecewise function: when the total risk value is below the preset threshold, it increases linearly; however, when it reaches or exceeds the preset threshold, the mapping curve turns into a steep, non-linear increase (e.g., quadratic increase), causing the risk cost to instantly spike to tens of times the normal value, thus ensuring that the current passage cost of this node inevitably exceeds the preset cost safety threshold. Administrators can adjust the aforementioned threshold and the curvature parameters of the non-linear increase through the configuration interface of the equipment management platform, based on the actual situation on site. However, it must always be ensured that in any situation that generates a high risk score, the latest value of the final derived current passage cost or risk priority index will exceed the defined corresponding threshold. Through this "flexible configuration under principle constraints" mechanism, this invention can adapt to different scenarios and ensure absolute identification and proactive avoidance of high-risk nodes from the algorithmic level.

[0131] Determine the latest values ​​for environmental fitness assessment factors, specifically including:

[0132] Obtain the current values ​​of the multidimensional environmental feature parameters for each grid node, wherein the current values ​​of the multidimensional environmental feature parameters include at least:

[0133] The accessibility attribute of the corresponding grid node, wherein the accessibility attribute indicates whether the grid node belongs to a passable area or an impassable area;

[0134] The coverage of the corresponding grid nodes within the field of view of the fixed visual monitoring equipment;

[0135] It should be noted that the fixed visual monitoring equipment refers to monitoring cameras that are pre-deployed at fixed locations in the yard (such as light poles, building exterior walls, or dedicated poles). Their field of view is fixed and covers specific grid node areas, which is used to achieve wide-area and continuous visual monitoring.

[0136] The risk cost of the corresponding grid node; the risk cost is determined by substituting the total comprehensive risk value into a preset mapping relationship between the total comprehensive risk value and the risk cost;

[0137] The preset mapping relationship between the total comprehensive risk value and the risk cost is configured as a piecewise function:

[0138] When the total comprehensive risk value is less than a preset threshold, the risk cost increases linearly with the total comprehensive risk value.

[0139] When the total comprehensive risk value is greater than or equal to the preset critical value, the risk cost increases non-linearly with the total comprehensive risk value, and its growth rate increases with the increase of the total comprehensive risk value.

[0140] In addition, when the passage attribute represents a passable area, it also includes the actual passage width corresponding to the grid node;

[0141] In this embodiment, the actual passage width is a geometric parameter determined based on the passage attributes of a grid node and a preset neighborhood range, representing the size of the passable space in the local area where the current grid node is located. Specifically, for any current grid node whose passage attribute represents a passable area, the platform detects the passage attributes of all adjacent grid nodes within its preset neighborhood range; the actual passage width is the effective span of the physical space covered by the set of all consecutive passable grid nodes in the current grid node and its adjacent grid nodes.

[0142] For example, if a current grid node is located within a passage, and its adjacent grid nodes on its left and right sides (or top and bottom sides) are all passable areas, then the actual passage width is equal to the total physical width covered by these consecutive passable nodes in the transverse direction of the passage. This actual passage width directly quantifies the space sufficiency when the robot passes through the node: a larger value indicates a wider passable transverse space, greater room for robot posture adjustment, and a lower collision risk; a smaller value indicates a narrower passage. Therefore, in the calculation step of generating the environmental fitness evaluation factor, a larger actual passage width results in a smaller width factor (i.e., a lower corresponding passage cost), thereby guiding the global path search algorithm to prioritize paths through spacious areas and avoid the robot entering narrow bottleneck areas.

[0143] For each grid node, perform the following judgment and calculation steps:

[0144] If the accessibility attribute indicates that a grid node is an impassable area, then the current value of the environmental fitness evaluation factor of that grid node is set to the maximum value indicating impassability.

[0145] If the accessibility attribute indicates that a grid node is a passable area, then the latest value of the environmental fitness assessment factor of that grid node is calculated based on the current value of the multidimensional environmental characteristic parameters.

[0146] The calculation of the latest value of the environmental fitness evaluation factor for the grid node based on the current value of the multidimensional environmental characteristic parameters is as follows:

[0147] Based on the actual passage width, a width factor is calculated, wherein the larger the actual passage width, the smaller the value of the width factor, so as to characterize the lower passage cost of the grid node;

[0148] Based on the coverage and risk cost, the corresponding visibility factor and risk factor are calculated respectively; wherein, the higher the coverage, the smaller the value of the visibility factor; and the higher the risk cost, the larger the value of the risk factor.

[0149] The width factor, visibility factor, and risk factor are weighted and fused to generate the latest value of the environmental adaptability assessment factor for the grid node.

[0150] The specific calculation of each factor adopts a preset mapping relationship (such as inverse proportional function or direct proportional function) that conforms to the above-mentioned trend. The actual passage width, coverage and risk cost are quantified into dimensionless factor values ​​so as to generate environmental adaptability assessment factors through weighted fusion.

[0151] This invention constructs a multi-dimensional environmental characteristic parameter fusion-based passage cost assessment system. By comprehensively considering the actual passage width, the coverage of the fixed visual monitoring equipment's field of view, and the risk cost, it significantly improves the inherent safety and controllability of the inspection process. The platform calculates a width factor based on the actual passage width, a visibility factor based on the coverage, and combines this with a risk factor for weighted fusion to generate an environmental adaptability assessment factor. This assessment logic guides the robot to prioritize paths with wide passages, effective video surveillance coverage (non-blind spots), and controllable risks during movement. This avoids the risk of the robot scratching or getting stuck in narrow bottleneck areas, and also prevents the risk of loss of control due to the backend being unable to promptly confirm the situation when an anomaly occurs if the robot enters a video surveillance blind spot.

[0152] Determine the latest value of the risk priority index, specifically including:

[0153] The total comprehensive risk value is substituted into the preset mapping relationship between the total comprehensive risk value and the risk priority index to obtain the latest value of the risk priority index; wherein, the larger the total comprehensive risk value, the larger the corresponding value of the risk priority index.

[0154] Obtain the latest values ​​of the risk priority index and environmental adaptability assessment factor of the grid nodes related to its current inspection path, specifically including:

[0155] Identify the unexecuted segments of the current inspection path of each robot, and take all the grid nodes included in the unexecuted segments, as well as the adjacent nodes of all the grid nodes, as the grid nodes related to its current inspection path.

[0156] The relevant grid nodes are divided into updated grid nodes and non-updated grid nodes; wherein, the updated grid nodes are grid nodes that correspond to multi-dimensional environmental data at the current time, and the non-updated grid nodes are grid nodes that do not correspond to multi-dimensional environmental data at the current time.

[0157] For grid nodes that have been updated, extract the latest values ​​of the risk priority index and environmental adaptability assessment factor for that grid node;

[0158] For grid nodes that have not been updated, the latest values ​​of the risk priority index and environmental adaptability assessment factor of that grid node are confirmed as the historical values ​​of the previous moment; if there are no historical values ​​of the previous moment, they are confirmed as the corresponding initial values.

[0159] It should be noted that there is a difference in whether the grid nodes of the currently unexecuted segment of the inspection path and their adjacent grid nodes have corresponding multidimensional environmental data at the current moment.

[0160] Although each robot operates independently within its designated inspection area, at boundary nodes where these areas intersect, the monitoring range of one robot may extend to the edge of the designated inspection area of ​​an adjacent robot. Therefore, when other collaborative robots approach the boundary and collect data, their uploaded data may include target environmental data from segments not yet executed by the robot itself or from adjacent nodes. Furthermore, data from fixed monitoring equipment in areas adjacent to the boundary of some designated inspection areas can also cover parts of the boundary grid nodes.

[0161] Based on this, this embodiment logically divides the relevant grid nodes into updated grid nodes and unupdated grid nodes:

[0162] Updated grid nodes: These refer to grid nodes that correspond to multi-dimensional environmental data at the current moment (i.e., multi-dimensional environmental data within a preset valid time window, which can be, for example, the last 30 minutes). This data typically originates from the robot's own real-time monitoring, cross-monitoring by other collaborative robots in the boundary area, or coverage by nearby fixed monitoring equipment. For these nodes, the equipment management platform performs timeliness verification on the multi-source data, selecting only the latest data whose timestamp falls within the preset valid time window as the node's current moment data. For these nodes, the platform directly extracts the latest values ​​of their risk priority index and environmental adaptability assessment factor to instantly reflect sudden changes in the environment.

[0163] Unupdated grid nodes: These refer to grid nodes that do not have corresponding multidimensional environmental data for the current moment (i.e., they do not have corresponding multidimensional environmental data within the preset valid time window). This is usually because the node is located deep within a dedicated inspection area, neither covered by fixed monitoring equipment nor within the real-time monitoring range of other robots. For such nodes, the platform does not make unfounded estimations, but instead confirms the latest values ​​of its risk priority index and environmental adaptability assessment factor as the historical values ​​of the previous moment; if no historical values ​​exist, the corresponding initial values ​​are confirmed.

[0164] This strategy enables the platform to share data through multi-robot boundary collaborative perception, and also ensures that the path planning algorithm can continue to run based on reliable historical data in areas with no real-time data, thus avoiding path planning interruptions due to data loss.

[0165] Based on the latest acquired value, the unexecuted segments of the current inspection path are evaluated, and the corresponding current inspection path is updated based on the evaluation results, including:

[0166] Based on the latest values ​​of the risk priority index and environmental adaptability assessment factor of the updated grid node, the current passage cost of the grid node is recalculated using the same weighted calculation method as that used to calculate the passage cost based on the initial value; for grid nodes that have not been updated, their passage cost at the previous moment is confirmed as the current passage cost.

[0167] Traverse each grid node included in the unexecuted segment of the current inspection path and determine whether there are high-risk nodes among its adjacent grid nodes. If so, increase the current passage cost of that grid node as a penalty.

[0168] The penalty increase on the current passage cost of the grid node is specifically as follows:

[0169] Obtain the current passage cost of this grid node as the base passage cost;

[0170] Find the high-risk nodes among the neighboring grid nodes of the given grid node;

[0171] Based on the current passage cost of each high-risk node, a neighborhood risk penalty value is calculated; wherein, the higher the current passage cost of the high-risk node, the higher the neighborhood risk penalty value.

[0172] Specifically, the platform uses a preset positive mapping function (such as a linear or non-linear growth function) to directly determine the neighborhood risk penalty value based on the current passage cost of high-risk nodes. This results in a higher current passage cost value for high-risk nodes, leading to a larger calculated neighborhood risk penalty value and thus stronger guidance for avoiding obstacles in their surrounding paths.

[0173] The basic passage cost is accumulated or weighted and fused with the neighborhood risk penalty value, and the current passage cost of the grid node is updated based on the result of the accumulation or weighted fusion.

[0174] Update the global travel cost map based on the updated current travel cost;

[0175] Determine whether there are high-risk nodes among the grid nodes related to the current inspection path. If so, take the robot's current real-time position as the starting node. Meanwhile, while ensuring coverage of the remaining untraversed mandatory inspection nodes, take the preset inspection target point in the dedicated inspection area as the ending node. Based on the updated global passage cost map, use the preset global path search algorithm to re-search in the corresponding dedicated inspection area to obtain the optimized path with the minimum cumulative cost, and set the optimized path as the new current inspection path.

[0176] This invention establishes a dynamic path cost penalty mechanism based on neighborhood awareness. By quantifying the radiation impact of a hazard source on the surrounding environment into the economic cost of path search, it constructs a proactive safety buffer barrier. Specifically, when evaluating the unexecuted segments of the current inspection path, the platform first recalculates the current passage cost of each grid node in the unexecuted segment based on the latest values ​​to accurately reflect its inherent risks. Then, it traverses its neighboring grid nodes to identify whether there are high-risk nodes. If so, it calculates a neighborhood risk penalty value based on the current passage cost of the adjacent high-risk nodes, and accumulates or weights this penalty value into the passage cost of the current grid node. This mechanism successfully transforms the neighborhood radiation impact, such as the toxic gas diffusion range and the fire heat radiation zone, into high-cost signals that the algorithm can identify. This guides the global path search algorithm to proactively avoid potentially risky road segments where "the path node itself seems safe, but the neighboring nodes are high-risk" during the process of finding the path with the minimum cumulative cost. The inspection path planned in this way has the necessary safety buffer distance, which avoids the robot from driving close to the edge of the danger source. It effectively prevents the robot from being directly threatened or even damaged due to the sudden spread of dangerous situations, and achieves true active risk avoidance and safe inspection.

[0177] In this embodiment, for a grid node, when a high-risk node exists among its neighboring grid nodes, the platform will penalize and increase the current passage cost of that grid node. Specifically, when an emergency risk such as a gas leak, fire, or smoke is detected around a grid node, the platform uses a penalty mechanism to increase the passage cost of that grid node to a level far exceeding the normal passage cost (e.g., reaching tens of times or even higher than the normal grid node cost). This high-cost setting aims to create a strong repulsion for robot passage, ensuring that any path attempting to traverse or closely follow the area will have its total cumulative cost skyrocket due to the inclusion of a huge neighborhood risk penalty value, thus being automatically judged as an extremely uneconomical and inferior path in the algorithm evaluation.

[0178] Based on the aforementioned penalty mechanism, the platform naturally develops an intelligent cost-benefit logic during path search: while avoiding high-risk nodes may require the robot to choose a longer detour, increasing the cumulative basic travel cost, the detour successfully avoids high-risk neighborhoods, resulting in zero or extremely low neighborhood risk penalty values. In contrast, directly traversing high-risk areas, although the physical distance is shortest, incurs a huge neighborhood risk penalty, making the total cost far higher than the detour. Therefore, driven by the goal of minimizing cumulative cost, the global path search algorithm automatically prioritizes the longer but safer detour, never choosing to traverse dangerous areas to shorten the journey. This mechanism ensures from the algorithm's core that the robot always adheres to the proactive risk avoidance principle of "better to take a longer route than a dangerous one," maximizing the inherent safety of the operation process while ensuring full coverage of the inspection task.

[0179] The conditions for determining the high-risk nodes are as follows:

[0180] The current passage cost exceeds the preset cost safety threshold, or the latest value of the risk priority index exceeds the preset emergency threshold.

[0181] In this embodiment, the robot is equipped with multi-dimensional sensors and an edge computing module, enabling spatial resolution. The edge computing module can accurately map the multi-dimensional sensors' collected gas concentrations, ambient temperature, and visual image data within their monitoring range, combined with the robot's current positioning information, onto the corresponding grid nodes in the grid node map, forming target environmental data containing multiple independent records (each record corresponds to the coordinates of a grid node). Based on this data, when the robot travels along the current inspection path, if the equipment management platform analyzes the target environmental data and finds that the robot's own grid node is in a safe state, but a certain grid node within the monitoring range that belongs to the unexecuted segment of the current inspection path, or its adjacent nodes, is identified as a high-risk node, the platform immediately triggers a path update: according to the penalty increase mechanism, the platform will increase the current passage cost of adjacent grid nodes around the high-risk node, thereby constructing a high-cost safety buffer zone around the risk source. Subsequently, based on the updated global passage cost map, starting from the robot's current real-time position, and ensuring coverage of the remaining untraversed mandatory inspection nodes, the optimized path with the minimum cumulative cost is re-searched. Because the total cost of traversing this buffer zone is extremely high, the newly planned path will automatically guide the robot to detour from a safe direction away from the risk source. This process enables the robot to proactively avoid risks based on forward-looking local perception data before entering high-risk nodes, ensuring the inherent safety of the inspection task.

[0182] It should be noted that the calculated current passage cost of high-risk nodes in this invention will inevitably far exceed the total accumulated cost of normal inspection paths. The global path search algorithm can already absolutely avoid paths directly traversing these nodes at the basic level; however, the neighborhood risk penalty mechanism further proposed in this invention extends this avoidance scope from the node itself to its surrounding neighborhood. By constructing a high-cost buffer ring around high-risk nodes, the algorithm is forced to not only avoid traversing the danger source when planning paths, but also actively move away from its edge, thereby achieving a safety upgrade from point-to-point avoidance to regional isolation.

[0183] This invention achieves closed-loop management of the risk priority index from initial setting to real-time dynamic updating, ensuring the timeliness and accuracy of the global access cost map in reflecting the yard environment. The platform first sets initial values ​​for the risk priority index for each grid node based on the distribution information of hazardous cargo within the yard area, providing a basic risk reference for path planning. Then, by analyzing target environmental data in real time, it dynamically derives the total comprehensive risk value and determines the latest value of the risk priority index, using this latest value to update the historical values ​​of each grid node in real time. This mechanism, based on continuous iterative updates using real-time data, enables the global access cost map to respond instantly to changes in the yard environment (such as gas concentration fluctuations, temperature anomalies, or the appearance of obstacles), ensuring that the robot always makes decisions based on values ​​reflecting the true risk situation at the current moment during path search.

[0184] The dynamic optimization method for yard inspection paths also includes triggering an early warning mechanism when the latest value of the risk priority index of any grid node exceeds a preset early warning threshold.

[0185] The early warning mechanism provided by this invention enables visual monitoring of alert levels and real-time push notifications of emergency alarms, ensuring that managers can simultaneously grasp the latest value and precise location of the risk priority index, effectively improving the efficiency of handling emergencies. This tiered processing strategy ensures accurate delivery of early warning information at different levels, providing a reliable basis for rapid response to emergencies.

[0186] When the latest value of the risk priority index at any grid node exceeds a preset warning threshold, a warning mechanism is triggered, specifically including:

[0187] When the latest value of the risk priority index exceeds the first-level preset warning threshold but does not reach the second-level preset warning threshold, the corresponding grid node is marked with the first preset warning color on the monitoring interface of the equipment management platform, and the latest value of the risk priority index is displayed as a prompt-level warning.

[0188] When the latest value of the risk priority index reaches or exceeds the second-level preset warning threshold, the corresponding grid node is marked with the second preset warning color, and an alarm message containing the location of the grid node and the latest value of the risk priority index is pushed to the preset safety management personnel terminal as an emergency alarm.

[0189] The visual warning intensity of the second preset warning color is higher than that of the first preset warning color.

[0190] This invention analyzes target environmental data collected by fixed monitoring equipment and robots and processed by edge computing to derive the comprehensive total risk value of corresponding grid nodes in real time, and dynamically determines the latest values ​​of the risk priority index and environmental adaptability assessment factor based on this. For each robot, the latest values ​​of the aforementioned grid nodes related to the unexecuted segments of its current inspection path are obtained, and the unexecuted segments are evaluated in real time based on these latest values ​​to update the current inspection path. This dynamic closed-loop mechanism based on real-time data breaks the rigid mode of traditional fixed route or static traversal strategies that rely solely on fixed map information. It ensures that the robot always plans its route based on the latest risk priority index and environmental adaptability assessment factor reflecting the current environmental state, significantly improving the environmental adaptability of the inspection path and the safety of the operation process.

[0191] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0192] Furthermore, in this invention, descriptions involving terms such as "first," "second," and "a" are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0193] In this invention, unless otherwise explicitly specified and limited, the terms "connection," "fixed," etc., should be interpreted broadly. For example, "fixed" can mean a fixed connection, a detachable connection, or an integral part; it can mean a mechanical connection or an electrical connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0194] Furthermore, the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are feasible for those skilled in the art. If the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

Claims

1. A method for dynamic optimization of yard inspection paths based on an equipment management platform, characterized in that, include: The storage yard area is discretized into a grid node diagram. Initial values ​​are set for the risk priority index of each grid node, and the environmental adaptability assessment factor of each grid node is initialized. The initial value of the risk priority index represents the static risk level, and the environmental adaptability assessment factor represents the feasibility of passage. Based on the initial values ​​of the risk priority index and the environmental adaptability assessment factor of each grid node, initial inspection paths are generated for multiple robots deployed in the yard area; each initial inspection path is set as the current inspection path of the corresponding robot. By deploying fixed monitoring equipment in the yard area and robots traveling along their respective current inspection paths, multi-dimensional environmental data of grid nodes within their respective monitoring ranges are collected in real time. Edge computing is performed on the multi-dimensional environmental data to obtain the target environmental data, which is then uploaded to the equipment management platform. The target environment data is analyzed through the equipment management platform, and the comprehensive risk value of the corresponding grid node is derived based on the analysis results. The latest values ​​of the risk priority index and environmental adaptability assessment factor of the corresponding grid node are determined based on the comprehensive risk value. For each robot, the latest values ​​of the risk priority index and environmental adaptability assessment factor of the grid node related to its current inspection path are obtained. Based on the obtained latest values, the unexecuted segments of its current inspection path are evaluated, and the corresponding current inspection path is updated based on the evaluation results. Determine the latest values ​​for environmental fitness assessment factors, specifically including: Obtain the current values ​​of the multidimensional environmental feature parameters for each grid node, wherein the current values ​​of the multidimensional environmental feature parameters include at least: The accessibility attribute of the corresponding grid node, wherein the accessibility attribute indicates whether the grid node belongs to a passable area or an impassable area; The coverage of the corresponding grid nodes within the field of view of the fixed visual monitoring equipment; The risk cost of the corresponding grid node; the risk cost is determined by substituting the total comprehensive risk value into a preset mapping relationship between the total comprehensive risk value and the risk cost; In addition, when the passage attribute represents a passable area, it also includes the actual passage width corresponding to the grid node; For each grid node, perform the following judgment and calculation steps: If the accessibility attribute indicates that a grid node is an impassable area, then the current value of the environmental fitness evaluation factor of that grid node is set to the maximum value indicating impassability. If the accessibility attribute characterizes a grid node as a passable area, then the latest value of the environmental fitness assessment factor of that grid node is calculated based on the current value of the multidimensional environmental characteristic parameters. Determine the latest value of the risk priority index, specifically including: The total comprehensive risk value is substituted into the preset mapping relationship between the total comprehensive risk value and the risk priority index to obtain the latest value of the risk priority index; wherein, the larger the total comprehensive risk value, the larger the corresponding value of the risk priority index.

2. The method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 1, characterized in that, The initial value for the risk priority index of each grid node is set, specifically including: Obtain the storage distribution information of goods within the yard area, and define the risk level of the corresponding grid node based on the hazardous characteristics of the goods within the coverage area of ​​the grid node; wherein, the risk level includes at least high risk level, medium risk level, low risk level and safe level; Based on the risk level of the grid node, an initial value for its risk priority index is set. The higher the risk level, the larger the initial value of the risk priority index corresponding to the grid node.

3. The method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 2, characterized in that, The multidimensional environmental data includes: Concentrations of various gases, ambient temperature, and visual image data.

4. The method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 3, characterized in that, The process of obtaining target environmental data by performing edge computing on multidimensional environmental data specifically involves: The concentrations of various gases and ambient temperature are collected and denoised, and outlier values ​​are removed to generate standardized environmental parameter values. Lightweight vision algorithms are used to perform real-time detection and processing of acquired visual image data. When at least one target is detected, semantic label data containing the type identifier of the corresponding target and the location coordinates of the grid node where it is located is generated; the type identifier of the target is either a specific category identifier of an obstacle or a category identifier of fireworks; When a person is detected, semantic label data containing the location coordinates of the grid node where each person is located and the number of people is generated, and the corresponding face region is simultaneously extracted from the visual image data for each detected person to generate the corresponding face bounding box. When no target objects or people are detected, semantic label data representing a normal scene is generated. The standardized environmental parameter values, the generated semantic label data, and the face bounding image generated when a person is detected are spatiotemporally labeled and encapsulated to form the target environmental data.

5. The method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 4, characterized in that, The process of parsing target environmental data through the device management platform and deriving the overall risk value of the corresponding grid nodes based on the parsing results is as follows: Analyze target environment data; For each grid node, perform the following risk score generation steps: If the concentration of any gas or the ambient temperature in the standardized environmental parameter values ​​exceeds its corresponding preset environmental safety threshold, a corresponding environmental risk score is generated. If the semantic tag data contains a category identifier for fireworks, a fire risk score is generated. If the semantic label data contains a specific category identifier for the obstacle, the degree of traffic obstruction is assessed based on that specific category identifier, and an obstacle risk score is generated based on the assessment result. If the semantic tag data contains the location coordinates of a person, and the identity is determined to be an unauthorized intrusion after identity comparison based on the face bounding box, then an intrusion risk score is generated. The environmental risk score, fire risk score, obstacle risk score, and intrusion risk score are summed to obtain the total comprehensive risk value of the grid node.

6. The method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 5, characterized in that, The calculation of the latest value of the environmental fitness evaluation factor for the grid node based on the current value of the multidimensional environmental characteristic parameters is as follows: Based on the actual passage width, a width factor is calculated, wherein the larger the actual passage width, the smaller the value of the width factor, so as to characterize the lower passage cost of the grid node; Based on the coverage and risk cost, the corresponding visibility factor and risk factor are calculated respectively; wherein, the higher the coverage, the smaller the value of the visibility factor; and the higher the risk cost, the larger the value of the risk factor. The width factor, visibility factor, and risk factor are weighted and fused to generate the latest value of the environmental adaptability assessment factor for the grid node.

7. The method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 1, characterized in that, Based on the initial values ​​of the risk priority index and environmental adaptability assessment factor for each grid node, initial inspection paths are generated for multiple robots deployed in the yard area, specifically including: Based on the initial values ​​of the risk priority index and the initial values ​​of the environmental adaptability assessment factor of all grid nodes in the yard area, a global access cost map is constructed. For each robot, its dedicated inspection area, preset starting point, and multiple mandatory inspection nodes to be traversed within the area are determined. Taking the preset starting point as the starting node, and ensuring that the path covers all the mandatory inspection nodes, the preset inspection target point within the dedicated inspection area is taken as the ending node. Based on the global access cost map, a preset global path search algorithm is used to search for the path with the minimum cumulative cost within the corresponding dedicated inspection area, which is then used as the initial inspection path for that robot. The search space of the preset global path search algorithm is limited to a connected subgraph consisting of all grid nodes with the passability attribute of passable areas within the dedicated inspection area.

8. The method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 7, characterized in that, The construction of a global access cost map based on the initial values ​​of the risk priority index and the environmental adaptability assessment factor for all grid nodes within the storage yard area specifically includes: Traverse each grid node within the storage area and perform the following cost assignment steps: If the initial value of the environmental fitness evaluation factor of the grid node is a maximum value that represents impassability, then the passage cost of the grid node is set to a preset prohibition cost. If the initial value of the environmental adaptability assessment factor of the grid node is not a maximum value that represents impassability, then: a weighted calculation is performed based on the initial value of the environmental adaptability assessment factor and the initial value of the risk priority index, and the calculation result is used as the passage cost of the grid node. The smaller the initial value of the environmental adaptability assessment factor, the lower the corresponding passage cost; The global travel cost map is composed of the travel costs of all grid nodes.

9. A method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 8, characterized in that, Obtain the latest values ​​of the risk priority index and environmental adaptability assessment factor of the grid nodes related to its current inspection path, specifically including: Identify the unexecuted segments of the current inspection path of each robot, and take all the grid nodes included in the unexecuted segments, as well as the adjacent nodes of all the grid nodes, as the grid nodes related to its current inspection path. The relevant grid nodes are divided into updated grid nodes and non-updated grid nodes; wherein, the updated grid nodes are grid nodes that correspond to multi-dimensional environmental data at the current time, and the non-updated grid nodes are grid nodes that do not correspond to multi-dimensional environmental data at the current time. For grid nodes that have been updated, extract the latest values ​​of the risk priority index and environmental adaptability assessment factor for that grid node; For grid nodes that have not been updated, the latest values ​​of the risk priority index and environmental adaptability assessment factor of that grid node are confirmed as the historical values ​​of the previous moment; if there are no historical values ​​of the previous moment, they are confirmed as the corresponding initial values.

10. A method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 9, characterized in that, Based on the latest acquired value, the unexecuted segments of the current inspection path are evaluated, and the corresponding current inspection path is updated based on the evaluation results, including: Based on the latest value of the updated grid node, recalculate the current passage cost of the grid node; for the unupdated grid node, confirm its passage cost at the previous moment as the current passage cost. Traverse each grid node included in the unexecuted segment of the current inspection path and determine whether there are high-risk nodes among its adjacent grid nodes. If so, increase the current passage cost of that grid node as a penalty. Update the global travel cost map based on the updated current travel cost; Determine whether there are high-risk nodes among the grid nodes related to the current inspection path. If so, take the robot's current real-time position as the starting node. Meanwhile, while ensuring coverage of the remaining untraversed mandatory inspection nodes, take the preset inspection target point in the dedicated inspection area as the ending node. Based on the updated global passage cost map, use the preset global path search algorithm to re-search in the corresponding dedicated inspection area to obtain the optimized path with the minimum cumulative cost, and set the optimized path as the new current inspection path.

11. The method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 10, characterized in that, The conditions for determining the high-risk nodes are as follows: The current passage cost exceeds the preset cost safety threshold, or the latest value of the risk priority index exceeds the preset emergency threshold.

12. The method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 11, characterized in that, The penalty increase on the current passage cost of the grid node is specifically as follows: Obtain the current passage cost of this grid node as the base passage cost; Find the high-risk nodes among the neighboring grid nodes of the given grid node; Based on the current passage cost of each high-risk node, a neighborhood risk penalty value is calculated; wherein, the higher the current passage cost of the high-risk node, the higher the neighborhood risk penalty value. The basic passage cost is accumulated or weighted and fused with the neighborhood risk penalty value, and the current passage cost of the grid node is updated based on the result of the accumulation or weighted fusion.

13. The method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 1, characterized in that, The preset mapping relationship between the total comprehensive risk value and the risk cost is configured as a piecewise function: When the total comprehensive risk value is less than a preset threshold, the risk cost increases linearly with the total comprehensive risk value. When the total comprehensive risk value is greater than or equal to the preset critical value, the risk cost increases non-linearly with the total comprehensive risk value, and its growth rate increases with the increase of the total comprehensive risk value.

14. The method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 1, characterized in that, The dynamic optimization method for yard inspection paths also includes triggering an early warning mechanism when the latest value of the risk priority index of any grid node exceeds a preset early warning threshold.

15. A method for dynamic optimization of yard inspection paths based on an equipment management platform according to claim 14, characterized in that, When the latest value of the risk priority index at any grid node exceeds a preset warning threshold, a warning mechanism is triggered, specifically including: When the latest value of the risk priority index exceeds the first-level preset warning threshold but does not reach the second-level preset warning threshold, the corresponding grid node is marked with the first preset warning color on the monitoring interface of the equipment management platform, and the latest value of the risk priority index is displayed as a prompt-level warning. When the latest value of the risk priority index reaches or exceeds the second-level preset warning threshold, the corresponding grid node is marked with the second preset warning color, and an alarm message containing the location of the grid node and the latest value of the risk priority index is pushed to the preset safety management personnel terminal as an emergency alarm.

Images