Path Optimization Method and System for Inspection Robot Based on Dynamic Obstacle Avoidance

Abstract

This application provides a path optimization method and system for an inspection robot based on dynamic obstacle avoidance. The method includes: the inspection robot obtains turbulence data of gas leakage sources, wind direction data of weather stations, and environmental reflection feature data in real time, generates a high-precision gas diffusion prediction map through multi-layer convolution processing of gas and wind direction data, and simultaneously extracts the optical contour of transparent obstacles and geometric parameters of unstructured obstacles from the reflection data to establish a three-dimensional space mapping relationship with the gas diffusion map. Based on this mapping relationship and the preset task priority, path control parameters are dynamically generated to optimize the distribution of global path nodes, enabling the robot path to adapt to the complex risk environment of sudden gas concentration changes and obstacle movement in real time. This application improves the dynamic obstacle avoidance accuracy and safety response speed of the inspection robot path in complex and dangerous environments.

Description

Technical Field

[0001] This application relates to the technical field of dynamic obstacle avoidance, and particularly to a path optimization method and system for an inspection robot based on dynamic obstacle avoidance. Background Art

[0002] In dangerous environments such as chemical plants and oil storage tank areas, inspection robots need to continuously sense the trend of gas leakage diffusion and the distribution of dynamic obstacles, and quickly plan safe paths. Such scenarios require robots to simultaneously handle the combined risks of sudden changes in gas concentration, occlusion by transparent obstacles (such as glass pipelines), and interference from unstructured obstacles (such as mobile devices), ensuring the safety and efficiency of the inspection process.

[0003] A typical current solution uses a multi-modal data fusion technology of lidar and gas sensors. It obtains three-dimensional point cloud data of obstacles through lidar scanning, combines the detection of local concentration by gas sensors, and plans the global path based on a static risk assessment model. This solution triggers obstacle avoidance behavior through a preset gas concentration threshold and updates the obstacle position information using a point cloud matching algorithm.

[0004] This solution relies on a static risk assessment model and is difficult to adapt to the dynamic changes in the gas diffusion trend. Especially when the wind direction suddenly changes, the path planning lags significantly. Lidar has a blind spot in detecting transparent obstacles, resulting in an underestimated risk of gas diffusion path occlusion. The multi-modal data is processed independently, lacking a collaborative analysis of the spatial influence of gas diffusion and obstacles, and the path optimization has strong locality. Summary of the Invention

[0005] This application provides a path optimization method and system for an inspection robot based on dynamic obstacle avoidance to solve the problems of low dynamic obstacle avoidance accuracy and slow safety response speed of the inspection robot path in complex and dangerous environments in the prior art.

[0006] In a first aspect, this application provides a path optimization method for an inspection robot based on dynamic obstacle avoidance, including:

[0007] During the movement of the inspection robot along the global path, obtain the real-time turbulence data of the gas leakage source in the target area, the real-time wind direction data provided by the weather station, and the environmental reflection feature data;

[0008] Perform multi-layer convolution processing on the real-time turbulence data and the real-time wind direction data to generate a target gas diffusion prediction map;

[0009] Extract the optical contour of the transparent obstacle and the spatial geometric parameters of the unstructured obstacle from the environmental reflection feature data, and establish a three-dimensional space mapping relationship among the optical contour, the spatial geometric parameters, and the target gas diffusion prediction map;

[0010] Generate path control parameters according to the three-dimensional space mapping relationship and the priority conditions of the preset inspection tasks, and determine multiple path nodes in the target area through the path control parameters to optimize the global path, so that the optimized global path dynamically adapts to the combined changes of gas concentration fluctuations and obstacle position offsets.

[0011] Optionally, the generating path control parameters according to the three-dimensional space mapping relationship and the priority conditions of the preset inspection tasks includes:

[0012] Based on the three-dimensional space mapping relationship, determine the shielding influence coefficient of the optical contour of the transparent obstacle on the gas diffusion direction vector in the target gas diffusion prediction map, and calculate the interference intensity of the spatial geometric parameters of the unstructured obstacle on the global path;

[0013] Weight and superimpose the shielding influence coefficient and the interference intensity according to the priority conditions of the preset inspection tasks to generate a coupling risk field of gas and obstacles, and divide the target area into independent area units with different path avoidance levels according to the comprehensive risk value distribution of the coupling risk field;

[0014] Dynamically screen the initial node sequence of the global path according to the independent area units with different path avoidance levels to generate a candidate avoidance node set;

[0015] Based on the inverse correlation between the gas concentration fluctuation amplitude and the obstacle displacement rate of each node in the candidate avoidance node set, iteratively correct the node spacing and connection direction;

[0016] Generate path control parameters for controlling the moving direction and speed of the inspection robot according to the corrected node spacing and the corrected connection direction.

[0017] Optionally, the weighting and superimposing the shielding influence coefficient and the interference intensity according to the priority conditions of the preset inspection tasks to generate a coupling risk field of gas and obstacles includes:

[0018] Based on the deflection angle range of the gas diffusion direction vector in the shielding influence coefficient, determine the shielding intensity of the transparent obstacle on gas diffusion;

[0019] Based on the angle range between the displacement direction of the unstructured obstacle and the global path in the interference intensity, calculate the blocking intensity of the movement of the unstructured obstacle on the path;

[0020] According to the corresponding relationship between the leakage source hazard level and the obstacle movement speed in the priority conditions of the preset inspection tasks, assign a gas risk weight to the shielding intensity and an obstacle risk weight to the blocking intensity;

[0021] Superimpose the first product value and the second product value of the same independent region unit to generate a comprehensive risk value of the independent region unit, and generate a coupled risk field according to the comprehensive risk values of all independent region units. The first product value is the product value of the occlusion intensity and the gas risk weight, and the second product value is the product value of the blocking intensity and the obstacle risk weight.

[0022] Optionally, the superimposing the first product value and the second product value of the same independent region unit to generate a comprehensive risk value of the independent region unit, and generating a coupled risk field according to the comprehensive risk values of all independent region units includes:

[0023] For each independent region unit, extract the corresponding occlusion parameter from the occlusion intensity according to the deflection angle range value, and extract the corresponding blocking parameter from the blocking intensity according to the included angle range value;

[0024] Match the first mapping value of the gas risk weight for the occlusion parameter in each independent region unit, and match the second mapping value of the obstacle risk weight for the blocking parameter;

[0025] Multiply the occlusion parameter in each independent region unit by the first mapping value to generate the first product value, and multiply the blocking parameter by the second mapping value to generate the second product value;

[0026] Accumulate the first product value and the second product value of each independent region unit unit by unit to generate the comprehensive risk value of each independent region unit;

[0027] Merge the independent region units with the same risk value according to spatial continuity to generate a coupled risk field.

[0028] Optionally, the iteratively correcting the node spacing and the connection direction based on the inverse correlation between the gas concentration fluctuation amplitude and the obstacle displacement rate of each node in the candidate avoidance node set includes:

[0029] For each current node in the candidate avoidance node set, obtain the measured value of the gas concentration fluctuation amplitude of the current node and the measured value of the displacement rate of the inspection robot relative to the obstacle;

[0030] Determine the first adjustment amount required for the current node according to the preset corresponding relationship between the gas concentration and the spacing adjustment;

[0031] Determine the second adjustment amount required for the current node according to the preset corresponding relationship between the displacement rate and the direction adjustment;

[0032] Correct the spacing between the current node and the adjacent node according to the first adjustment parameter, and adjust the connection direction between the current node and the next node according to the second adjustment amount.

[0033] Optionally, the multi-layer convolution processing of the real-time turbulence data and the real-time wind direction data to generate a target gas diffusion prediction map includes:

[0034] Align the gas flow velocity change information in the real-time turbulence data with the wind direction angle change information in the real-time wind direction data according to a time window to generate an input data group;

[0035] Input the input data group into a pre-constructed multi-layer convolution structure. In the first processing layer of the multi-layer convolution structure, extract the first gas diffusion trend feature from the gas flow velocity change information and extract the wind direction guiding feature from the wind direction angle change information;

[0036] In the second processing layer, perform feature fusion on the first gas diffusion trend feature and the wind direction guiding feature to generate a second gas diffusion trend feature;

[0037] In the third processing layer, based on the second gas diffusion trend feature, predict the concentration distribution range of gas diffusion within a future time window to generate an initial diffusion prediction map;

[0038] Perform smoothing correction on the initial diffusion prediction map to generate a target gas diffusion prediction map.

[0039] Optionally, the establishment of the three-dimensional space mapping relationship between the optical profile, the spatial geometric parameters, and the target gas diffusion prediction map includes:

[0040] Match the optical profile boundary with the gas concentration gradient direction at the corresponding position in the target gas diffusion prediction map;

[0041] Associate the spatial geometric parameters with the gas flow velocity at the corresponding position in the target gas diffusion prediction map;

[0042] Based on the matching result and the association result, construct a three-dimensional space mapping relationship.

[0043] In a second aspect, the present application provides a path optimization system for an inspection robot based on dynamic obstacle avoidance, including:

[0044] An acquisition module, configured to acquire real-time turbulence data of a gas leakage source in a target area, real-time wind direction data provided by a weather station, and environmental reflection feature data during the movement of the inspection robot along a global path;

[0045] A first generation module, configured to perform multi-layer convolution processing on the real-time turbulence data and the real-time wind direction data to generate a target gas diffusion prediction map;

[0046] An extraction module, configured to extract the optical profile of the transparent obstacle and the spatial geometric parameters of the unstructured obstacle from the environmental reflection feature data, and establish a three-dimensional space mapping relationship among the optical profile, the spatial geometric parameters, and the target gas diffusion prediction map;

[0047] A second generation module, configured to generate path control parameters according to the three-dimensional space mapping relationship and the priority conditions of the preset inspection tasks, and determine multiple path nodes in the target area through the path control parameters, so as to optimize the global path and make the optimized global path dynamically adapt to the composite changes of gas concentration fluctuations and obstacle position offsets.

[0048] In a third aspect, the present application provides a computing device, including a processor and a memory. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the method for optimizing the path of an inspection robot based on dynamic obstacle avoidance according to any one of the first aspects.

[0049] In a fourth aspect, the present application provides a computer storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method for optimizing the path of an inspection robot based on dynamic obstacle avoidance according to any one of the first aspects is implemented.

[0050] In the present application, a method for optimizing the path of an inspection robot based on dynamic obstacle avoidance is provided. The method includes: during the movement of the inspection robot along the global path, obtaining the real-time turbulence data of the gas leakage source in the target area, the real-time wind direction data provided by the weather station, and the environmental reflection feature data; performing multi-layer convolution processing on the real-time turbulence data and the real-time wind direction data to generate a target gas diffusion prediction map; extracting the optical profile of the transparent obstacle and the spatial geometric parameters of the unstructured obstacle from the environmental reflection feature data, and establishing a three-dimensional space mapping relationship among the optical profile, the spatial geometric parameters, and the target gas diffusion prediction map; generating path control parameters according to the three-dimensional space mapping relationship and the priority conditions of the preset inspection tasks, and determining multiple path nodes in the target area through the path control parameters, so as to optimize the global path and make the optimized global path dynamically adapt to the composite changes of gas concentration fluctuations and obstacle position offsets.

[0051] The technical solution provided by the present application has the following beneficial effects:

[0052] Through multi-source sensing fusion, this application achieves comprehensive perception of gas leakage dynamics, environmental airflow changes, and obstacle characteristics, providing high-timeliness data support for risk modeling; uses multi-layer convolution processing to capture the spatio-temporal correlation characteristics of gas turbulence and wind direction, accurately predicting the diffusion trend and concentration distribution of leaked gas; solves the problem of detecting transparent obstacles through optical contour and geometric parameter identification, and constructs a spatial quantization model of gas-obstacle interaction; based on dynamic risk coupling and task priority, outputs an optimal path node sequence that takes into account both safe obstacle avoidance and inspection efficiency, realizing autonomous decision-making in complex environments.

[0053] Furthermore, by quantifying the shielding effect of transparent obstacles on gas diffusion and the interference intensity of unstructured obstacles on the path, a coupled risk field is generated in combination with task priority, and a node set is dynamically screened according to the risk level; based on the reverse association rule of gas concentration fluctuation and obstacle displacement, the node layout is iteratively optimized, and finally the path control parameters that are adapt to the composite risk in real time are generated.

[0054] Moreover, this solution breaks through the limitations of traditional static planning, enabling the robot to simultaneously improve the accuracy of obstacle avoidance and the coherence of the path in scenarios where gas leakage and moving obstacles coexist.

[0055] These aspects or other aspects of this application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0057] Figure 1 It is a flowchart of a path optimization method for an inspection robot based on dynamic obstacle avoidance provided by an embodiment of this application;

[0058] Figure 2 It is a schematic structural diagram of a path optimization system for an inspection robot based on dynamic obstacle avoidance provided by an embodiment of this application;

[0059] Figure 3 It is a schematic structural diagram of a computing device provided by an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] In order to enable those skilled in the art to better understand the solution of this application, the following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the drawings in the embodiments of this application.

[0061] In some processes described in the specification, claims, and the above-mentioned drawings of this application, multiple operations appear in a specific order. However, it should be clearly understood that these operations can be executed not in the order in which they appear herein or in parallel. The operation numbers such as 101 and 102 are only used to distinguish different operations, and the numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations can be executed sequentially or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit that "first" and "second" are of different types.

[0062] Researchers have found that existing inspection robots are difficult to simultaneously meet the dynamic changes of gas leakage diffusion and the real-time obstacle avoidance requirements for complex obstacles in dangerous environments such as chemical plants. In particular, there are blind spots in the detection of transparent obstacles, and there is a lack of collaborative analysis of gas diffusion trends and the spatial impact of obstacles. Based on this, the embodiment of this application provides a path optimization method for an inspection robot based on dynamic obstacle avoidance. This method can construct a three-dimensional gas-obstacle space mapping model by integrating gas turbulence data, wind direction data, and environmental reflection characteristics, and dynamically generate optimal path control parameters based on task priorities to achieve synchronous avoidance of gas leakage risks and moving obstacles. The technical solution of this application is applicable to industrial inspection scenarios such as chemical plants and oil storage tank areas with risks of toxic gas leakage and complex obstacle environments.

[0063] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of this application.

[0064] Figure 1 It is a flowchart of a path optimization method for an inspection robot based on dynamic obstacle avoidance provided by an embodiment of this application. As Figure 1 shown, this method includes:

[0065] Step 101: During the process of the inspection robot moving along the global path, obtain the real-time turbulence data of the gas leakage source in the target area, the real-time wind direction data provided by the weather station, and the environmental reflection characteristic data.

[0066] In this step, the real-time turbulence data refers to the dynamically changing data such as the air flow velocity and eddy current intensity near the gas leakage source collected by the gas sensor, which is used to reflect the instantaneous diffusion state of the leaked gas. The real-time wind direction data represents the current environmental wind direction and wind speed information provided by the weather station, which is used to predict the gas diffusion direction. The environmental reflection feature data represents the data on the reflection characteristics of the surface of environmental objects obtained by the optical sensor, including polarization light characteristics and reflection intensity distribution.

[0067] In the embodiment of the present application, multiple groups of sensors installed on the inspection robot synchronously collect the gas flow characteristics, meteorological conditions, and object reflection characteristic data in the working environment; the gas sensor array obtains the air flow parameters around the leakage source at a fixed sampling frequency; the weather station updates the wind direction and wind speed data in real time through wireless transmission; the optical sensor obtains the reflection characteristics of environmental objects by means of multi-spectral scanning, and all the data is output to the central processing unit after preprocessing.

[0068] For example, in the leakage monitoring scenario of an ethylene storage tank in a chemical plant, the laser gas sensor carried by the inspection robot collects the methane concentration distribution data around the storage tank at a frequency of 10 times per second (concentration value = sensor output voltage × calibration coefficient 0.25), and at the same time receives the real-time wind direction data (wind direction angle = magnetometer reading + geomagnetic declination correction value 5.3°) wirelessly transmitted by the plant weather station through LoRa (Long Range), and obtains the polarization reflection image of the pipeline surface through the polarization camera. All the data is stored after time stamp alignment.

[0069] Step 102: Perform multi-layer convolution processing on the real-time turbulence data and the real-time wind direction data to generate a target gas diffusion prediction map.

[0070] In this step, the multi-layer convolution processing means adopting a multi-layer neural network structure with the ability to extract spatial features, and layer by layer extracting the spatio-temporal correlation features of the air flow and wind direction through local receptive fields. The target gas diffusion prediction map includes a grid map of the spatial distribution of gas concentration in the future period, and each grid cell stores the predicted concentration value and the diffusion direction vector.

[0071] In the embodiment of the present application, the preprocessed turbulence data and wind direction data are input into a three-dimensional convolutional neural network. The first-layer convolutional kernel extracts the local air flow pattern features, the second-layer convolutional kernel fuses the spatio-temporal correlation of the air flow and wind direction, and the third-layer deconvolution operation reconstructs the complete diffusion field, and finally outputs a gas diffusion prediction map including the concentration gradient and diffusion direction; when training the network, historical leakage event data is used as samples, and the network parameters are optimized through the loss function.

[0072] For example, for the above-mentioned ethylene storage tank leakage scenario, the system inputs the collected turbulent data (10×10×5 cm grid) and wind direction data into the trained prediction model. After three-layer convolutional operations, a gas diffusion prediction map for the next 5 minutes is generated. Each 20×20 cm grid contains concentration values (in the range of 0-100%) and direction vectors (angle accuracy of 1°), and the prediction results are updated every 30 seconds.

[0073] Step 103: Extract the optical contour of the transparent obstacle and the spatial geometric parameters of the unstructured obstacle from the environmental reflection feature data, and establish a three-dimensional space mapping relationship among the optical contour, the spatial geometric parameters, and the target gas diffusion prediction map.

[0074] In this step, the transparent obstacle refers to a physical obstacle in the chemical plant environment composed of glass, plexiglass, or other light-transmitting materials. The optical contour represents the geometric boundary features formed by the reflected light on the surface of the transparent obstacle, which is extracted by polarization imaging technology. The unstructured obstacle refers to a random obstacle in the chemical plant environment that does not conform to the standard geometric shape. The spatial geometric parameters include the external dimensions, surface curvature, and spatial position information of the unstructured obstacle. The three-dimensional space mapping relationship is used to describe the topological structure of the interaction relationship between the obstacle spatial distribution and the gas diffusion field.

[0075] In the embodiment of the present application, a continuous light intensity gradient change region is extracted from the polarization reflection image as the transparent obstacle contour, and a point cloud segmentation algorithm is used to extract the geometric parameters of the unstructured obstacle from the depth image; a unified coordinate system including the gas concentration field, the obstacle position, and the inspection area structure is established, and the occlusion weight of the obstacle to gas diffusion is calculated through spatial interpolation to generate a three-dimensional risk distribution model.

[0076] For example, in the storage tank area scenario, the system identifies the elliptical contour (major axis 2.3 m, minor axis 1.5 m) of the glass observation window (transparent obstacle) and the cubic parameters (2×1.5×1.2 m) of the moving material cart (unstructured obstacle). These obstacles are superimposed on the predicted gas diffusion field, and the occlusion coefficient of the observation window to the downwind area is calculated to be 0.7 (=sin(45° occlusion angle)), and the blocking coefficient of the material cart to the path is 0.8 (= (actual speed 1.2 m / s) / (maximum safe speed 1.5 m / s)).

[0077] Step 104: Generate path control parameters according to the three-dimensional space mapping relationship and the priority conditions of the preset inspection task, and determine multiple path nodes in the target area through the path control parameters to optimize the global path, so that the optimized global path dynamically adapts to the composite changes of gas concentration fluctuations and obstacle position offsets.

[0078] In this step, the priority condition represents the path optimization weight coefficient set according to the risk levels of different leakage sources. The path control parameters include the constraints of path planning elements such as node spacing, steering angle, and speed.

[0079] In the embodiment of the present application, based on the three-dimensional risk model and the preset priorities (such as the weight of the ethylene storage tank is 0.9, and the weight of the pipeline valve is 0.7), the dynamic programming algorithm is used to generate the initial path nodes; the node positions are adjusted by the risk gradient descent method to increase the node spacing in the high-risk area and keep the original path in the low-risk area; finally, an optimized path instruction set including position coordinates, arrival time, and motion parameters is output.

[0080] For example, in the case of a storage tank leakage scenario, the system first plans an initial path passing through 6 key nodes. After detecting that the risk value near the No. 2 node exceeds the standard (0.75 > threshold 0.6), 1 avoidance node is inserted on each side of this node (the spacing increases from 1m to 1.5m). The total length of the adjusted path increases by 12%, but the highest risk contact probability is reduced to the safe range.

[0081] This method constructs a gas-obstacle coupled risk field through multi-source data fusion, realizes the precise perception of transparent obstacles and unstructured obstacles, can dynamically predict the gas diffusion trend and optimize the inspection path in real time, effectively avoids various risks while ensuring the integrity of leakage monitoring, and improves the inspection safety and operation efficiency in dangerous environments.

[0082] To solve the problem of the accuracy of dynamic obstacle avoidance of inspection robots in complex industrial environments, in some embodiments, step 104: generating path control parameters according to the three-dimensional space mapping relationship and the priority conditions of the preset inspection tasks includes:

[0083] Step 201: Based on the three-dimensional space mapping relationship, determine the shielding influence coefficient of the optical contour of the transparent obstacle on the gas diffusion direction vector in the target gas diffusion prediction map, and calculate the interference intensity of the spatial geometric parameters of the unstructured obstacle on the global path.

[0084] In step 201, the shielding influence coefficient represents the degree of occlusion of the transparent obstacle on the gas diffusion direction. The larger the value, the stronger the occlusion effect, which is calculated by the angle between the gas diffusion direction vector and the normal vector of the obstacle surface. The interference intensity represents the degree of obstruction of the unstructured obstacle to the preset path, which is comprehensively determined according to the ratio of the geometric size of the obstacle to the path spacing and the moving speed.

[0085] In the embodiment of the present application, the system first analyzes the geometric relationship between the contour of the transparent obstacle and the gas diffusion direction, and calculates the occlusion weight corresponding to the light deflection angle; at the same time, it measures the spatial position relationship between the shape size of the unstructured obstacle and the robot path, and evaluates the path interference degree in combination with its motion state.

[0086] Step 202: Weightedly superimpose the shielding influence coefficient and the interference intensity according to the priority conditions of the preset inspection task to generate a coupling risk field of gas and obstacles, and divide the target area into independent area units with different path avoidance levels according to the distribution of the comprehensive risk values of the coupling risk field.

[0087] In step 202, a specific embodiment of the weighted superposition process is as follows: In the inspection scenario of a dangerous area in a chemical plant, for the area near a leakage point of a certain ethylene pipeline, the shielding influence coefficient of the transparent observation window (transparent obstacle) on the gas diffusion direction vector is measured to be 0.65 through a three-dimensional space mapping relationship (calculation method: the ratio of the shielding angle of 45° to the maximum possible shielding angle of 90° is sin(45°)=0.71, and the value is 0.65 after being corrected by the gas penetration rate). At the same time, the interference intensity of the moving material vehicle (unstructured obstacle) on the preset path is calculated to be 0.8 (calculation method: the square of the ratio of the vehicle speed of 2 m / s to the maximum allowable speed of 1 m / s is (2 / 1)^2 = 4, and the value is 0.8 after being corrected by the path deviation tolerance coefficient of 0.2); according to the weights of the "ethylene leakage" danger level of 0.7 and the "moving equipment" weight of 0.3 in the preset priority conditions, a weighted superposition calculation is performed: the coupling risk value = 0.65×0.7 + 0.8×0.3 = 0.695. When this value exceeds the division threshold of 0.6, this area unit is marked as a first-level avoidance area, and a new candidate avoidance node needs to be generated to bypass this high-risk area. The weight distribution is determined according to the corresponding relationship between the leakage danger level and the equipment moving speed in the "Code for Safety Design of Petroleum Chemical Enterprises" GB50160. The threshold of 0.6 is obtained through statistical analysis of the risk critical values corresponding to 80% of the dangerous events in historical accident data. The coupling risk field represents a two-dimensional risk distribution map formed by superimposing the gas diffusion risk and the obstacle risk, and each grid unit contains a comprehensive risk value. The avoidance level represents the area classification divided according to the risk value size, and the higher the level, the greater the danger. An independent area unit refers to a grid unit obtained by dividing the target area according to the spatial position, and each unit contains the calculated comprehensive risk value, which is used to identify the risk level of this local area. The unit division is based on the spatial resolution of the coupling risk field to ensure that the risk characteristics within each unit are relatively uniform, and the unit boundary corresponds to the mutation position of the risk value. In the chemical plant scenario, the unit size is usually set to be equivalent to the obstacle characteristic size, such as a 2m×2m unit around the storage tank, which is convenient for accurately identifying high-risk areas.

[0088] In the embodiment of the present application, different weights are assigned to the calculated shielding coefficient and interference intensity according to the risk level of the leakage source, and a full-field risk distribution map is generated through linear superposition. Then, the area is divided into different avoidance levels according to a preset risk threshold.

[0089] Step 203: Dynamically screen the initial node sequence of the global path according to the independent regional units with different path avoidance levels, and generate a candidate avoidance node set.

[0090] In step 203, the path avoidance level includes that in the area corresponding to the first avoidance level, the risk quantification value is greater than the first preset threshold, and the risk level of the leakage source is the highest priority; in the area corresponding to the second avoidance level, the risk quantification value is between the second preset threshold and the first preset threshold, and the moving speed of the obstacle is greater than the preset speed; in the area corresponding to the third avoidance level, the risk quantification value is less than the second preset threshold, and the deviation angle from the initial global path is less than the preset angle. The candidate avoidance node set represents a set of alternative path nodes added for high-risk areas based on the initial path nodes.

[0091] In the embodiment of the present application, the system scans the risk levels of the areas where each node of the initial path is located, and inserts a number of candidate nodes before and after the high-risk level areas to form an alternative path network.

[0092] Step 204: Iteratively correct the node spacing and connection direction based on the inverse correlation between the gas concentration fluctuation amplitude and the obstacle displacement rate of each node in the candidate avoidance node set.

[0093] In step 204, the gas concentration fluctuation amplitude reflects the severity of gas concentration changes at the monitoring point. It is calculated from the standard deviation of continuously sampled data from the gas sensor. A larger value indicates more unstable leakage diffusion. The obstacle displacement rate refers to the speed of movement of unstructured obstacles. The displacement is obtained by feature matching between two consecutive frames of point cloud data and then divided by the time interval. Both data are from the real-time monitoring system and updated every second. They are used to dynamically assess the risk trend of path nodes. Inverse correlation refers to an optimization rule in which the node spacing adjustment is opposite to the gas concentration change trend, and the connection direction adjustment is opposite to the obstacle movement trend. Node spacing refers to the physical distance between two adjacent path nodes in the inspection robot's optimized path. It is dynamically adjusted based on the gas concentration fluctuation amplitude in the area. The more severe the concentration change, the larger the spacing. The specific distance is determined by querying a preset concentration-spacing comparison table. The connection direction refers to the angle of change in movement direction between adjacent path nodes, that is, the steering angle required when the robot moves from one node to the next. Correction is made based on the inverse of the obstacle's displacement direction. When the obstacle moves from left to right, the connection direction deflects to the right by a corresponding angle. The deflection is calculated by multiplying the displacement rate by the direction adjustment coefficient. The two work together to ensure that the path avoids both high-risk gas areas and moving obstacles.

[0094] In an embodiment of the present application, the distribution density and connection angle of candidate nodes are dynamically adjusted according to real-time monitoring data. When the gas concentration at a node rises, the distance between adjacent nodes is automatically increased. When an obstacle approaches, the direction of the node connection is adjusted.

[0095] Step 205: Generate path control parameters for controlling the moving direction and speed of the inspection robot according to the corrected node spacing and the corrected connection direction.

[0096] In step 205 , the path control parameters include an optimized node coordinate sequence, a recommended speed to reach each node, a steering angle, and other control instruction sets.

[0097] In an embodiment of the present application, the finally determined node position information is converted into path instructions executable by the robot motion control system, including specific parameters such as straight segment cruising speed and turning radius.

[0098] Here's a specific example:

[0099] In the scenario of ethylene storage tank leakage, the system calculates that the shielding influence coefficient of the observation window on the gas diffusion in the downwind direction is 0.7 (calculation formula: sin(45° shielding angle)) through the superposition analysis of the elliptical contour of the glass observation window (major axis 2.3 meters, minor axis 1.5 meters) and the gas diffusion prediction map. At the same time, based on the cube parameters (2×1.5×1.2 meters) and the moving speed of 1.2 meters per second of the moving material truck, the interference intensity of the truck on the preset path is calculated to be 0.8 (calculation formula: actual speed 1.2 meters per second divided by the maximum allowable speed 1.5 meters per second). Weighted superposition is carried out according to the high-risk level weight of 0.9 for ethylene leakage and the weight of 0.3 for the moving equipment (calculation process: 0.7×0.9 + 0.8×0.3 = 0.87), generating a coupled risk field showing the highest risk area (risk value 0.87) on the northeast side of the storage tank, and dividing this area into a first-level avoidance unit. The system detects that the risk value exceeds the standard (0.87 > threshold 0.6) near the No. 2 node (coordinates X = 35.2, Y = 18.7) of the original path, so two new candidate avoidance nodes are added 1.5 meters on both sides of the node (spacing calculation basis: basic spacing 1 meter × risk coefficient 1.5). At the same time, according to the fact that the northward displacement speed of the material truck detected in real time increases to 1.5 meters per second, the connection direction of the new nodes is deflected 15 degrees eastward (deflection amount = speed increment 0.3 meters per second × deflection coefficient 50 degrees·second / meter). The finally generated path control parameters include 8 nodes (original 6 + newly added 2), their corresponding moving speeds (the speed is reduced to 0.8 meters per second in the high-risk area), and steering instructions.

[0100] In the embodiment of the present application, this method quantifies the interaction relationship between various risk factors, realizes the collaborative obstacle avoidance of gas leakage and moving obstacles, enables the inspection robot to autonomously avoid compound risks on the premise of ensuring the integrity of monitoring, and improves the operation safety in dangerous environments.

[0101] In order to further improve the accuracy of risk field generation in dangerous environments, in some embodiments, step 202: The weighted superposition of the shielding influence coefficient and the interference intensity according to the priority conditions of the preset inspection task to generate a coupled risk field of gas and obstacles includes:

[0102] Step 301: Based on the deflection angle range of the gas diffusion direction vector in the shielding influence coefficient, determine the shielding intensity of the transparent obstacle on gas diffusion.

[0103] In step 301, the deflection angle range of the gas diffusion direction vector refers to the angular interval in which the flow direction of the gas changes when it flows through the transparent obstacle, and is calculated by comparing the gas diffusion direction vectors upstream and downstream of the obstacle. The specific acquisition method is as follows: extract the average diffusion direction of the undisturbed area upstream of the obstacle contour line in the gas diffusion prediction diagram as the reference direction, and then measure the angle between the diffusion direction of the affected area downstream of the contour line and the reference direction. The value range of this angle (such as 0° - 90°) is the deflection angle range. In the chemical plant scenario, this parameter reflects the degree of distortion of the diffusion path of the leaked gas by transparent obstacles such as glass observation windows. The occlusion intensity represents the degree to which the transparent obstacle causes the change of the gas diffusion direction, and is calculated by the ratio of the actual deflection angle of the gas diffusion direction vector at the obstacle contour to the maximum possible deflection angle.

[0104] In the embodiment of the present application, the system first identifies the contour line of the transparent obstacle, calculates the angle change range of the direction vector when the gas flows through the contour, and then determines the occlusion intensity according to the proportional relationship between the angle change amplitude and the geometric size of the obstacle.

[0105] Step 302: Calculate the blocking intensity of the movement of the unstructured obstacle on the path based on the angle range between the displacement direction of the unstructured obstacle in the interference intensity and the global path.

[0106] In step 302, the blocking intensity represents the degree of obstruction of the unstructured obstacle to the preset path, which is jointly determined by the angle between the displacement direction of the obstacle and the path direction and the moving speed.

[0107] In the embodiments of the present application, the angle between the moving direction of the obstacle and the planned path of the robot is measured. Considering the shape size and moving speed of the obstacle, the degree of path obstruction is calculated through a spatial position relationship model. Specifically: The calculation of the degree of path obstruction comprehensively considers three key factors: First, the angle between the moving direction of the obstacle and the planned path. The smaller the angle, the closer the obstacle is to directly blocking the path; Second, the shape size of the obstacle. The larger the size, the larger the blocking range; Third, the moving speed. The faster the speed, the higher the threat level to the path. The specific calculation process is as follows: First, calculate the cosine value of the angle as the direction influence factor (the smaller the angle, the larger the cosine value). Then, take the ratio of the maximum shape size of the obstacle to the standard size as the size influence factor. Finally, take the ratio of the moving speed to the safety speed threshold as the speed influence factor. Multiply the three to obtain the path obstruction degree coefficient. The calculation formula is: Degree of path obstruction = cos(angle) × (shape size / reference size) × (moving speed / safety speed), where the reference size is taken as 1 meter and the safety speed is taken as 1.5 m / s. Exemplarily, in an inspection scenario of a chemical plant, the system detects a moving material truck (shape size 2.5×1.8×1.2 m) moving at a speed of 1.2 m / s, and the angle between its moving direction and the planned path of the robot is 45 degrees. The calculation process is as follows: The direction influence factor is taken as cos(45°)=0.707; The size influence factor is taken as the maximum size of 2.5 meters divided by the reference size of 1 meter to get 2.5; The speed influence factor is taken as 1.2 m / s divided by the safety speed of 1.5 m / s to get 0.8; The final degree of path obstruction = 0.707×2.5×0.8 = 1.414.

[0108] Step 303: According to the correspondence between the leakage source risk level and the moving speed of the obstacle in the priority conditions of the preset inspection task, assign a gas risk weight to the occlusion intensity and an obstacle risk weight to the blocking intensity.

[0109] In step 303, the correspondence is a pre-set risk assessment rule library. Different risk levels (such as high, medium, low) are divided according to the properties of the leaked substance (such as toxicity, flammability), and a basic weight coefficient is assigned to each level; At the same time, a mapping table between the moving speed interval of the obstacle and the additional weight is established. The faster the speed, the higher the additional weight. For example, in the scenario of an ethylene storage tank, ethylene leakage is classified as the highest risk level (basic weight 0.9), and when the speed of the moving equipment exceeds 1 m / s, a speed additional weight (0.1 - 0.3) is triggered. The system realizes dynamic weight assignment by querying this rule library to ensure that high-risk leakage sources and fast-moving obstacles obtain a higher risk assessment priority. The gas risk weight represents a coefficient set according to the dangerous characteristics such as the toxicity and concentration of the leaked substance. The higher the risk level, the greater the weight. The obstacle risk weight represents a coefficient set based on the moving speed and volume of the obstacle. The faster the speed and the larger the volume, the higher the weight.

[0110] In the embodiments of the present application, the system queries a preset hazardous substance classification table to obtain the leakage source weight coefficient, and at the same time matches the corresponding risk weight according to the obstacle dynamic parameter table.

[0111] Step 304: Superimpose the first product value and the second product value of the same independent area unit to generate the comprehensive risk value of the independent area unit. According to the comprehensive risk values of all independent area units, a coupled risk field is generated. The first product value is the product value of the shielding intensity and the gas risk weight, and the second product value is the product value of the blocking intensity and the obstacle risk weight.

[0112] In step 304, the comprehensive risk value represents the risk level of the combined action of gas leakage and obstacles within a single area unit. The higher the value, the higher the danger.

[0113] In the embodiments of the present application, for each divided independent area unit, the weighted risk values of gas shielding and obstacle blocking are calculated separately, and the sum of the two is the comprehensive risk value of the unit. The comprehensive risk values of all units constitute a complete coupled risk field.

[0114] The following is a specific example:

[0115] In the scenario of ethylene storage tank leakage, the system detects that the glass observation window causes the gas diffusion direction to deflect by 45 degrees (deflection angle range 0 - 90 degrees), and calculates the shielding intensity to be 0.7 (calculation formula: actual deflection angle 45 degrees divided by the maximum possible deflection angle 90 degrees); at the same time, it monitors that the moving material truck moves at a speed of 1.2 m / s at an angle of 30 degrees to the preset path, and calculates the blocking intensity to be 0.8 (calculation formula: speed component 1.04 m / s of 1.2 m / s divided by the maximum allowable speed 1.5 m / s). According to the high - risk level of ethylene leakage (gas risk weight 0.9) and the moving speed of the material truck (obstacle risk weight 0.3), within a 2 m × 2 m independent area unit, the first product value 0.63 (0.7 × 0.9) and the second product value 0.24 (0.8 × 0.3) are calculated, and the comprehensive risk value 0.87 is obtained by superimposing them.

[0116] In the embodiments of the present application, this method realizes the hierarchical and accurate identification of dangerous areas by quantifying the composite risk of gas leakage and moving obstacles, provides a reliable risk distribution basis for subsequent path optimization, and effectively improves the decision - making scientificity of the inspection robot in complex environments.

[0117] To further improve the spatial accuracy of risk field generation, in some embodiments, step 304: The superimposing the first product value and the second product value of the same independent area unit to generate the comprehensive risk value of the independent area unit, and generating a coupled risk field according to the comprehensive risk values of all independent area units, includes:

[0118] Step 401: For each independent region unit, extract the corresponding occlusion parameter from the occlusion intensity according to the deflection angle range value, and extract the corresponding blocking parameter from the blocking intensity according to the included angle range value.

[0119] In step 401, the deflection angle range value refers to the angle interval in which the gas diffusion direction is deflected by the transparent obstacle within a single independent region unit. This parameter is determined by comparing the original gas diffusion direction at the upstream boundary of the unit with the actual diffusion direction at the downstream boundary, and reflects the degree of distortion of the gas flow path by the transparent obstacle at this unit position. In the chemical plant scenario, the larger this value is, the stronger the occlusion effect of the transparent obstacle (such as a glass observation window) on gas leakage diffusion. The blocking parameter characterizes the spatial position relationship between the movement direction of the unstructured obstacle and the preset global path direction within the independent region unit, and is obtained by calculating the minimum included angle between the obstacle movement direction vector and the path direction vector. The larger the included angle range value is, the more serious the conflict between the obstacle movement direction and the robot's predetermined path. In the inspection scenario, this parameter is used to quantify the potential interference intensity of mobile devices (such as forklifts, transport vehicles) on the inspection path. When the included angle is close to 90 degrees, it means the obstacle is crossing the path laterally, with the highest risk. The occlusion parameter characterizes the actual influence degree of the transparent obstacle on gas diffusion within a single unit, and is calculated by the average value of the gas direction deflection angle within the unit. The blocking parameter represents the actual obstruction degree of the unstructured obstacle on the path within a single unit, and is calculated by the weighted value of the included angle between the obstacle movement direction and the path within the unit.

[0120] In the embodiment of the present application, for each unit, the maximum deflection angle of the gas direction vector and the minimum included angle between the obstacle movement direction and the path are extracted, and after normalization, unit-level parameters are obtained.

[0121] Step 402: Match the first mapping value of the gas risk weight for the occlusion parameter in each independent region unit, and match the second mapping value of the obstacle risk weight for the blocking parameter.

[0122] In step 402, the first mapping value represents the unit-level gas risk weight determined according to the leakage source type and location. The second mapping value represents the unit-level obstacle risk weight determined according to the obstacle type and speed.

[0123] In the embodiment of the present application, the system queries the preset weight mapping table to match the corresponding weight value for each unit, and the weight assignment considers the distance attenuation effect between the unit and the leakage source / obstacle.

[0124] Step 403: Multiply the occlusion parameter in each independent region unit by the first mapping value to generate a first product value, and multiply the blocking parameter by the second mapping value to generate a second product value.

[0125] In step 403, the product value calculation means multiplying the unit-level risk parameter by the corresponding weight to obtain the independent risk contribution values of the gas and the obstacle.

[0126] In the embodiment of the present application, a per-unit parallel calculation method is adopted to synchronously generate the first product value and the second product value of all units, ensuring the calculation efficiency.

[0127] Step 404: Accumulate the first product value and the second product value of each independent region unit unit by unit to generate the comprehensive risk value of each independent region unit.

[0128] In step 404, the comprehensive risk value represents the superposition result of the gas and obstacle risks within a single unit, reflecting the composite hazard degree at that position.

[0129] In the embodiment of the present application, the two product values of each unit are linearly superimposed, and the result is constrained within the standard risk value range.

[0130] Step 405: Merge the independent region units with the same risk value according to spatial continuity to generate a coupled risk field.

[0131] In step 405, the region merging means merging adjacent units with similar risk characteristics into a continuous risk region to reduce redundant calculations.

[0132] In the embodiment of the present application, a region growing algorithm is adopted to start from the seed unit and merge adjacent units with a risk value difference less than the threshold to form a complete risk field partition.

[0133] The following is a specific example:

[0134] In the scenario of ethylene storage tank leakage, the system divides the area of 20 meters × 20 meters around the storage tank into 100 independent area units of 2 meters × 2 meters. For the 6 units (coordinates X30 - 32, Y20 - 22) covered by the glass observation window, the average deflection of the gas diffusion direction is measured to be 50 degrees (deflection angle range 0 - 90 degrees), and the shielding parameter is calculated to be 0.56 (50 / 90); for the 4 units (X34 - 36, Y18 - 20) passed by the moving material vehicle track, the angle between the average movement direction and the path is measured to be 40 degrees, and the blocking parameter is calculated to be 0.83 (the component of the vehicle speed of 1.2 m / s in the direction perpendicular to the path, which is 0.77 m / s, divided by the safety threshold of 0.93 m / s). According to the ethylene leakage hazard level (gas weight 0.9) and the vehicle risk level (obstacle weight 0.4), the first product value 0.504 (0.56×0.9) and the second product value 0.332 (0.83×0.4) are calculated in the X32Y20 unit, and the combined risk value is 0.836 after superposition. The system combines 8 adjacent units (X30 - 32, Y18 - 22) with a risk value ≥ 0.8 into a first-level risk area, and combines 12 units with a risk value of 0.7 - 0.8 into a second-level risk area. The finally generated coupled risk field shows that the northeast side of the storage tank is a high-risk red area (risk value 0.83), and the southwest side is a medium-risk yellow area (risk value 0.75).

[0135] In the embodiment of the present application, through refined unit division and dynamic weight mapping, the method realizes the spatial precise quantification of compound risks. The generated coupled risk field can accurately reflect the differences in hazard levels of different regions, providing a high-precision environmental situation awareness basis for path optimization.

[0136] In order to further improve the real-time performance and safety of path dynamic optimization, in some embodiments, step 204: based on the inverse correlation between the gas concentration fluctuation amplitude of each node in the candidate avoidance node set and the displacement rate of the obstacle, iteratively correct the node spacing and connection direction, including:

[0137] Step 501: For each current node in the candidate avoidance node set, obtain the measured value of the gas concentration fluctuation amplitude of the current node and the measured value of the displacement rate of the inspection robot relative to the obstacle.

[0138] In step 501, the current node refers to the path node that is currently being processed during the iterative correction process (i.e., a certain node to be optimized in the candidate avoidance node set). The determination method is as follows: according to the traversal order of the candidate avoidance node set (such as processing sequentially from the starting point to the ending point), each polled node is the current node. The measured value of the gas concentration fluctuation amplitude refers to the degree of severity of the gas concentration change with time at the position of the current node, which is calculated by the standard deviation of the continuously sampled data of the gas sensor. The measured value of the displacement rate represents the relative movement speed of the inspection robot with respect to the nearest obstacle, which is obtained by comparing the position changes of the robot and the obstacle at adjacent times.

[0139] In the embodiment of the present application, the system collects the gas concentration data around each node in real time through a gas sensor array, and at the same time uses a lidar to track the movement trajectory of the obstacle to calculate the relative velocity vector between the robot and the obstacle.

[0140] Step 502: Determine the first adjustment amount required for the current node according to the preset correspondence between gas concentration and spacing adjustment.

[0141] In step 502, the first adjustment amount represents the node spacing correction amount determined according to the gas concentration change situation. The greater the concentration fluctuation, the greater the adjustment amount.

[0142] In the embodiment of the present application, the system queries the preset concentration-spacing adjustment comparison table, which is formulated based on the analysis of historical accident data, divides the concentration fluctuation amplitude into several intervals and corresponds to different spacing adjustment coefficients.

[0143] Step 503: Determine the second adjustment amount required for the current node according to the preset correspondence between displacement rate and direction adjustment.

[0144] In step 503, the second adjustment amount represents the direction correction amount determined based on the relative movement speed of the obstacle. The faster the speed, the greater the adjustment amplitude.

[0145] In the embodiment of the present application, a direction adjustment model based on motion dynamics is adopted, and factors such as the movement direction of the obstacle, speed, and braking performance of the robot are considered to calculate the optimal avoidance direction.

[0146] Step 504: Correct the spacing between the current node and the adjacent node according to the first adjustment parameter, and adjust the connection direction between the current node and the next node according to the second adjustment amount.

[0147] In step 504, an adjacent node refers to the previous or next path node directly connected to the current node (spatially adjacent). It is used to calculate the node spacing correction amount (the distance between the current node and the adjacent node needs to be adjusted). The spacing correction means dynamically adjusting the distances between the current node and the adjacent nodes before and after to ensure sufficient avoidance space in the dangerous area. The direction adjustment means changing the movement direction of the current node pointing to the next node to avoid approaching obstacles quickly. The next node refers to the next path node to be visited in the moving direction of the current node (logically adjacent). It is used to correct the connection direction (the direction of the current node pointing to the next node needs to be adjusted).

[0148] In the embodiment of the present application, the system first recalculates the node coordinates according to the first adjustment amount, then rotates the connection vector between nodes according to the second adjustment amount, and finally verifies whether the new path meets all safety constraint conditions.

[0149] The following is a specific example:

[0150] In the scenario of ethylene storage tank leakage, the system dynamically adjusts the newly added candidate avoidance node No. 2 (coordinates X = 36.7, Y = 18.7): First, obtain the gas concentration fluctuation amplitude of 0.25 at this node in the past 30 seconds (calculation formula: standard deviation of 10 sampling values 0.2 multiplied by time decay coefficient 1.25), and determine the first adjustment amount as an increase in spacing of 0.4 meters according to the concentration-spacing adjustment table (the corresponding coefficient in the 0.2 - 0.3 interval is 1.2) (basic spacing 1.5 meters × coefficient 1.2 - 1.5 meters); at the same time, it is monitored that the material truck approaches from the northwest direction at a speed of 1.8 m / s (the included angle with the node connection line is 50°), and the second adjustment amount calculated by the direction adjustment model is a deflection of 22° (calculation process: 1.8 m / s × sin(50°) × 15°·s / m). After the adjustment is executed, the distance between candidate node No. 2 and node No. 1 increases from 1.5 meters to 1.9 meters, and the connection direction with node No. 3 is adjusted from the original due east direction to 22° north of east.

[0151] In the embodiment of the present application, this method enables the inspection robot to intelligently respond to sudden risks by real-time sensing environmental changes and dynamically adjusting path parameters, effectively avoiding the dual threats of gas leakage and moving obstacles while ensuring the continuity of the monitoring task, and improving the operation safety in complex environments.

[0152] In order to further improve the accuracy of gas diffusion prediction, in some embodiments, step 102: The multi-layer convolution processing of the real-time turbulence data and the real-time wind direction data to generate the target gas diffusion prediction map includes:

[0153] Step 601: Align the gas flow velocity change information in the real-time turbulence data and the wind direction angle change information in the real-time wind direction data according to a time window to generate an input data group.

[0154] In step 601, the input data group represents a three-dimensional data structure formed by aligning gas flow rate and wind direction data according to the same time stamp, including three dimensions: spatial position, time series, and eigenvalue.

[0155] In the embodiment of the present application, the system synchronously collects the gas flow rate matrix and the wind direction angle matrix at a fixed sampling period, and forms a spatio-temporal unified data cube after alignment on the time axis.

[0156] Step 602: Input the input data group into a pre-constructed multi-layer convolution structure. In the first processing layer of the multi-layer convolution structure, extract the first gas diffusion trend feature from the gas flow rate change information, and extract the wind direction guiding feature from the wind direction angle change information.

[0157] In step 602, the first gas diffusion trend feature represents the underlying feature reflecting the spatial distribution law of the gas flow rate, including flow rate gradient and eddy current intensity information. The wind direction guiding feature characterizes the basic feature of the guiding effect of the wind direction spatio-temporal change on gas diffusion, including wind direction stability and angle change rate.

[0158] In the embodiment of the present application, the first processing layer uses a large-size convolution kernel to extract the macroscopic distribution pattern of the gas flow rate, and at the same time uses a direction-sensitive convolution kernel to capture the main guiding feature of the wind direction.

[0159] Step 603: In the second processing layer, fuse the first gas diffusion trend feature and the wind direction guiding feature to generate a second gas diffusion trend feature.

[0160] In step 603, the second gas diffusion trend feature represents the middle-level feature of the interaction between the fused air flow and the wind direction, reflecting the comprehensive influence of environmental factors on diffusion.

[0161] In the embodiment of the present application, the air flow feature and the wind direction feature are weighted and fused in the channel dimension through feature cross-operation, and the feature combinations with strong correlation between the two are retained.

[0162] Step 604: In the third processing layer, based on the second gas diffusion trend feature, predict the concentration distribution range of gas diffusion within the future time window, and generate an initial diffusion prediction map.

[0163] In step 604, the initial diffusion prediction map contains the preliminary prediction results of the spatial distribution of gas concentration in the future period, and each pixel stores the predicted concentration value.

[0164] In the embodiment of the present application, the third processing layer upsamples the feature map to the original image size through deconvolution operation, and uses a regression loss function to optimize the prediction accuracy of the concentration value.

[0165] Step 605: Smoothly correct the initial gas diffusion prediction map to generate a target gas diffusion prediction map.

[0166] In step 605, smooth correction means eliminating noise and mutation points in the prediction results and improving the physical rationality of the prediction map.

[0167] In the embodiment of the present application, a smoothing algorithm based on hydrodynamics constraints is adopted to correct abnormal prediction values that do not conform to the gas movement law on the premise of maintaining the main diffusion trend.

[0168] The following is a specific example:

[0169] In the scenario of ethylene storage tank leakage monitoring, the system aligns the 10×10 cm grid turbulent data (flow velocity range 0 - 3 m / s) collected by the laser gas sensor with the wind direction data (accuracy 0.5 degrees) provided by the weather station in a 30-second time window to form an input data group of 40×40×30 (40×40 spatial grids, 30 time points). After this data group is input into the three-layer convolution structure: the first layer extracts the characteristics of the strong diffusion area with a maximum flow velocity of 2.8 m / s (the first gas diffusion trend characteristic) from the turbulent data, and at the same time extracts the stable northwest wind characteristic (wind direction angle 300±5 degrees) from the wind direction data; the second layer fuses to generate the second gas diffusion trend characteristic showing northeastward diffusion (diffusion angle 45±10 degrees); the third layer predicts the area where the ethylene concentration exceeds the warning value after 5 minutes (a fan-shaped area accounting for 15% of the plant area, with a radius of 7.5 m). After the smooth correction of the three abnormal high-concentration points (concentration sudden increase exceeding 30%) in the initial prediction map by the fluid continuity constraint, the generated target prediction map shows that the ethylene vapor mainly diffuses in the northeast direction (diffusion angle 40 - 50 degrees).

[0170] In the embodiment of the present application, this method realizes the accurate prediction of the diffusion trend of hazardous gases through multi-level feature extraction and fusion. The generated prediction map can accurately reflect the physical law of gas movement and provides the inspection robot with a highly reliable environmental situation perception ability.

[0171] To further improve the accuracy of environmental modeling, in some embodiments, step 103: Establishing the three-dimensional space mapping relationship between the optical contour, the spatial geometric parameters, and the target gas diffusion prediction map includes:

[0172] Step 701: Match the optical contour boundary with the gas concentration gradient direction at the corresponding position in the target gas diffusion prediction map.

[0173] In step 701, matching means spatially aligning the contour geometric features of the transparent obstacle with the gas diffusion direction field and analyzing the influence of the contour on the diffusion path. Specifically, when the angle between the optical contour boundary and the gas concentration gradient direction is less than the first set angle, it is determined that the transparent obstacle has an obstructive effect on gas diffusion; when the angle between the optical contour boundary and the gas concentration gradient direction is greater than or equal to the first set angle, it is determined that the transparent obstacle has no influence on gas diffusion;

[0174] In the embodiment of the present application, the system determines the actual obstructive area of the transparent obstacle on gas diffusion by calculating the angle between the normal vector of each point on the contour and the gas concentration gradient direction.

[0175] Step 702: Associate the spatial geometric parameters with the gas flow velocity at the corresponding position in the target gas diffusion prediction map.

[0176] In step 702, association means establishing a quantitative relationship between the external dimensions and position information of the unstructured obstacle and the local gas flow velocity distribution. Specifically, when the maximum geometric dimension of the unstructured obstacle is greater than the influence radius corresponding to the local flow velocity in the gas diffusion prediction map, it is determined that the obstacle has a blocking effect on gas flow; when the maximum geometric dimension of the unstructured obstacle is less than or equal to the influence radius, it is determined that the obstacle has no block on gas flow.

[0177] In the embodiment of the present application, the bounding box collision detection algorithm is used to calculate the intersection area between the space occupied by the obstacle and the gas flow field, and evaluate the attenuation degree of the obstacle on the air flow velocity.

[0178] Step 703: Based on the matching result and the association result, construct a three-dimensional space mapping relationship.

[0179] In the embodiments of the present application, a three-dimensional grid model containing multi-layer information such as gas concentration, flow rate, and obstacle distribution is generated through a spatial interpolation algorithm, and each grid cell stores complete interaction parameters. Specifically, first, the matching result of the transparent obstacle contour (the angle distribution between the contour points and the gas diffusion direction) is converted into a spatial occlusion weight matrix, and the value of each cell in the matrix represents the degree of obstruction of the transparent obstacle to gas diffusion at the corresponding position; at the same time, the geometric correlation result of the obstacle (the corresponding relationship between the obstacle size and the gas flow rate) is generated into a flow rate attenuation coefficient field. Then, a spatial superposition algorithm is used to perform three-dimensional grid fusion on the occlusion weight matrix, the flow rate attenuation coefficient field, and the original gas diffusion prediction map: for each grid cell, its final attribute value = gas concentration × (1 - occlusion weight) × flow rate attenuation coefficient, thereby forming a three-dimensional space mapping relationship that includes both the influence of obstacles and the characteristics of gas diffusion. Exemplarily, in the ethylene storage tank scenario, the system generates a 40×40 occlusion weight matrix (such as the weight in the area of coordinates X30 - 32, Y20 - 22 is 0.7 - 0.8) according to the contour matching result of the glass observation window, combines it with the flow rate attenuation field (the coefficient in the area of X34 - 36, Y18 - 20 is 0.6) generated by the associated result of the charging vehicle, and fuses it with the gas diffusion prediction map (the concentration in the northeast direction is 20 - 30%): calculate the final attribute value for the X32Y20 cell = 25% concentration × (1 - 0.75 occlusion weight) × 0.65 flow rate coefficient = 4.06%, and this value is stored in the corresponding grid of the three-dimensional model; the finally constructed mapping relationship clearly shows that there is a fan-shaped low-concentration area (4 - 8%) behind the observation window, and the movement path of the charging vehicle forms a strip-shaped turbulent area (the concentration fluctuates ±5%).

[0180] The following is a specific example:

[0181] In the scenario of ethylene storage tank leakage, the system first matches the elliptical contour of the glass observation window (major axis 2.3 meters, minor axis 1.5 meters) with the gas diffusion prediction map, calculates the angle between the normal vector of each point on the contour line and the gas concentration gradient direction, where the maximum angle at the contour vertex is 55 degrees, and calculates the occlusion coefficient at this position to be 0.82 (calculation formula: sin(55°)); at the same time, analyze the spatial relationship between the moving charging vehicle (size 2×1.5×1.2 meters) and the local gas flow rate field, and measure that the flow rate in front of the charging vehicle drops from 1.8 m / s to 1.2 m / s (attenuation coefficient 0.67, calculation formula: 1.2 / 1.8). The three-dimensional space model constructed based on these data shows that a low-speed diffusion area with a length of 3.2 meters and a width of 1.8 meters (average flow rate 0.9 m / s) is formed behind the observation window, and a turbulent area with a diameter of 2.2 meters (flow rate fluctuation range ±0.4 m / s) is generated on the movement path of the charging vehicle.

[0182] In the embodiments of the present application, the method constructs a high-precision environmental dynamic model by quantifying the spatial interaction effect between obstacles and the gas diffusion field, enabling the robot to accurately predict the change trend of the risk area and improving the quality of obstacle avoidance decisions in complex environments.

[0183] Figure 2 FIG. is a schematic structural diagram of a path optimization system for an inspection robot based on dynamic obstacle avoidance provided by an embodiment of the present application. As Figure 2 shown, the system includes:

[0184] An acquisition module 21, configured to acquire real-time turbulence data of a gas leakage source in a target area, real-time wind direction data provided by a weather station, and environmental reflection feature data during the movement of the inspection robot along the global path.

[0185] A first generation module 22, configured to perform multi-layer convolution processing on the real-time turbulence data and the real-time wind direction data to generate a target gas diffusion prediction map.

[0186] An extraction module 23, configured to extract the optical contour of a transparent obstacle and the spatial geometric parameters of an unstructured obstacle from the environmental reflection feature data, and establish a three-dimensional spatial mapping relationship among the optical contour, the spatial geometric parameters, and the target gas diffusion prediction map.

[0187] A second generation module 24, configured to generate path control parameters according to the three-dimensional spatial mapping relationship and the priority conditions of a preset inspection task, determine multiple path nodes in the target area through the path control parameters, so as to optimize the global path, and make the optimized global path dynamically adapt to the composite changes of gas concentration fluctuations and obstacle position offsets.

[0188] Figure 2 The path optimization system for an inspection robot based on dynamic obstacle avoidance can execute Figure 1 the path optimization method for an inspection robot based on dynamic obstacle avoidance described in the embodiments shown. Its implementation principle and technical effects will not be elaborated further. For the path optimization system for an inspection robot based on dynamic obstacle avoidance in the above embodiments, the specific manners in which each module and unit perform operations have been described in detail in the embodiments related to the method, and will not be elaborated in detail here.

[0189] In a possible design, Figure 2 the path optimization system for an inspection robot based on dynamic obstacle avoidance in the embodiments shown can be implemented as a computing device. As Figure 3 shown, the computing device may include a storage component 31 and a processing component 32;

[0190] The storage component 31 stores one or more computer instructions, and the one or more computer instructions are called and executed by the processing component 32.

[0191] The above-mentioned processing component 32 Figure 1 A path optimization method for an inspection robot based on dynamic obstacle avoidance according to the embodiment.

[0192] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0193] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0194] Of course, the computing device may also necessarily include other components, such as input / output interfaces, display components, communication components, etc.

[0195] The input / output interface provides an interface between the processing component and the peripheral interface module, and the peripheral interface module may be an output device, an input device, etc.

[0196] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0197] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device can refer to a cloud server. The above processing component, storage component, etc. can be basic server resources rented or purchased from a cloud computing platform.

[0198] The embodiment of the present application also provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above Figure 1 A path optimization method for an inspection robot based on dynamic obstacle avoidance shown in the embodiment.

[0199] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0200] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement without creative efforts.

[0201] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0202] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A path optimization method for an inspection robot based on dynamic obstacle avoidance, characterized in that: include: As the inspection robot moves along the global path, it obtains real-time turbulence data of gas leak sources in the target area, real-time wind direction data provided by the weather station, and environmental reflection characteristic data; Performing multi-layer convolution processing on the real-time turbulence data and the real-time wind direction data to generate a target gas diffusion prediction map; Extracting the optical profile of the transparent obstacle and the spatial geometric parameters of the unstructured obstacle from the environmental reflection feature data, and establishing a three-dimensional spatial mapping relationship between the optical profile, the spatial geometric parameters, and the target gas diffusion prediction map; Generate path control parameters based on the three-dimensional spatial mapping relationship and the priority conditions of the preset inspection tasks, and determine multiple path nodes within the target area using the path control parameters to optimize the global path, so that the optimized global path dynamically adapts to the combined changes of gas concentration fluctuations and obstacle position offsets; Generating the path control parameters according to the three-dimensional space mapping relationship and the priority conditions of the preset inspection tasks includes: Based on the three-dimensional spatial mapping relationship, determining the shielding influence coefficient of the optical profile of the transparent obstacle on the gas diffusion direction vector in the target gas diffusion prediction map, and calculating the interference intensity of the spatial geometric parameters of the unstructured obstacle on the global path; The shielding influence coefficient and the interference intensity are weighted and superimposed according to the priority conditions of the preset inspection tasks to generate a coupled risk field of gas and obstacles, and the target area is divided into independent area units with different path avoidance levels according to the comprehensive risk value distribution of the coupled risk field; Dynamically screening the initial node sequence of the global path according to the independent area units of the different path avoidance levels to generate a candidate avoidance node set; Iteratively correcting the node spacing and connection direction based on the inverse correlation between the gas concentration fluctuation amplitude of each node in the candidate avoidance node set and the obstacle displacement rate; According to the corrected node spacing and the corrected connection direction, path control parameters for controlling the moving direction and speed of the inspection robot are generated.

2. The method according to claim 1, characterized in that The weighted superposition of the shielding influence coefficient and the interference intensity according to the priority condition of the preset inspection task to generate a coupling risk field of gas and obstacles includes: Determining the shielding intensity of the transparent obstacle on gas diffusion based on the deflection angle range of the gas diffusion direction vector in the shielding influence coefficient; Calculating the blocking strength of the unstructured obstacle movement on the path based on the angle range between the displacement direction of the unstructured obstacle and the global path in the interference intensity; According to the correspondence between the leakage source hazard level and the obstacle movement speed in the priority conditions of the preset inspection task, a gas risk weight is assigned to the shielding intensity, and an obstacle risk weight is assigned to the blocking intensity; The first product value and the second product value of the same independent area unit are superimposed to generate a comprehensive risk value of the independent area unit. According to the comprehensive risk values of all independent area units, a coupled risk field is generated. The first product value is the product value of the blocking intensity and the gas risk weight, and the second product value is the product value of the blocking intensity and the obstacle risk weight.

3. The method according to claim 2, characterized in that Each independent area unit includes a deflection angle range value of the gas diffusion direction vector and an angle range value between the obstacle displacement direction and the global path; The first product value and the second product value of the same independent regional unit are superimposed to generate the comprehensive risk value of the independent regional unit. According to the comprehensive risk values of all independent regional units, a coupled risk field is generated, including: For each independent area unit, extract a corresponding shielding parameter from the shielding intensity according to the deflection angle range value, and extract a corresponding blocking parameter from the blocking intensity according to the angle range value; Matching the first mapping value of the gas risk weight to the occlusion parameter in each independent area unit, and matching the second mapping value of the obstacle risk weight to the blocking parameter; multiplying the occlusion parameter in each independent area unit by the first mapping value to generate a first product value, and multiplying the occlusion parameter by the second mapping value to generate a second product value; The first product value and the second product value of each independent regional unit are accumulated unit by unit to generate a comprehensive risk value of each independent regional unit; Independent regional units with the same risk value are merged according to spatial continuity to generate a coupled risk field.

4. The method according to claim 1, wherein The iterative correction of the node spacing and connection direction based on the inverse correlation between the gas concentration fluctuation amplitude of each node in the candidate avoidance node set and the obstacle displacement rate includes: For each current node in the candidate avoidance node set, obtaining a gas concentration fluctuation amplitude measurement value and a displacement rate measurement value of the inspection robot relative to the obstacle at the current node; Determining a first adjustment amount required for the current node according to a preset correspondence between gas concentration and spacing adjustment; Determining a second adjustment amount required for the current node based on a preset correspondence between displacement rate and direction adjustment; The distance between the current node and the adjacent node is corrected according to the first adjustment parameter, and the connection direction between the current node and the next node is adjusted according to the second adjustment amount.

5. The method according to claim 1, wherein The performing multi-layer convolution processing on the real-time turbulence data and the real-time wind direction data to generate a target gas diffusion prediction map includes: Aligning the gas flow velocity change information in the real-time turbulence data with the wind direction angle change information in the real-time wind direction data according to a time window to generate an input data set; Inputting the input data set into a pre-built multi-layer convolution structure, and extracting, in a first processing layer of the multi-layer convolution structure, a first gas diffusion trend feature from the gas flow rate change information, and a wind direction guidance feature from the wind direction angle change information; In the second processing layer, the first gas diffusion trend feature and the wind direction guidance feature are fused to generate a second gas diffusion trend feature; In the third processing layer, based on the second gas diffusion trend characteristics, the concentration distribution range of gas diffusion in the future time window is predicted to generate an initial diffusion prediction map; The initial diffusion prediction map is smoothed and corrected to generate a target gas diffusion prediction map.

6. The method according to claim 1, characterized in that The establishing of a three-dimensional spatial mapping relationship between the optical profile, the spatial geometric parameters, and the target gas diffusion prediction map includes: Matching the optical contour boundary with the gas concentration gradient direction at a corresponding position in the target gas diffusion prediction map; Associating the spatial geometric parameters with the gas flow rate at a corresponding position in the target gas diffusion prediction map; Based on the matching results and association results, a three-dimensional spatial mapping relationship is constructed.

7. A patrol robot path optimization system based on dynamic obstacle avoidance, characterized in that: include: The acquisition module is used to obtain real-time turbulence data of gas leakage sources in the target area, real-time wind direction data provided by the weather station, and environmental reflection characteristic data while the inspection robot moves along the global path; A first generating module is configured to perform multi-layer convolution processing on the real-time turbulence data and the real-time wind direction data to generate a target gas diffusion prediction map; an extraction module, configured to extract the optical profile of the transparent obstacle and the spatial geometric parameters of the unstructured obstacle from the environmental reflection feature data, and establish a three-dimensional spatial mapping relationship between the optical profile, the spatial geometric parameters, and the target gas diffusion prediction map; A second generation module is configured to generate path control parameters based on the three-dimensional spatial mapping relationship and the priority conditions of the preset inspection tasks, and to determine multiple path nodes within the target area using the path control parameters to optimize the global path, so that the optimized global path dynamically adapts to the combined changes of gas concentration fluctuations and obstacle position offsets; Generating the path control parameters according to the three-dimensional space mapping relationship and the priority conditions of the preset inspection tasks includes: Based on the three-dimensional spatial mapping relationship, determining the shielding influence coefficient of the optical profile of the transparent obstacle on the gas diffusion direction vector in the target gas diffusion prediction map, and calculating the interference intensity of the spatial geometric parameters of the unstructured obstacle on the global path; The shielding influence coefficient and the interference intensity are weighted and superimposed according to the priority conditions of the preset inspection tasks to generate a coupled risk field of gas and obstacles, and the target area is divided into independent area units with different path avoidance levels according to the comprehensive risk value distribution of the coupled risk field; Dynamically screening the initial node sequence of the global path according to the independent area units of the different path avoidance levels to generate a candidate avoidance node set; Iteratively correcting the node spacing and connection direction based on the inverse correlation between the gas concentration fluctuation amplitude of each node in the candidate avoidance node set and the obstacle displacement rate; According to the corrected node spacing and the corrected connection direction, path control parameters for controlling the moving direction and speed of the inspection robot are generated.

8. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a patrol robot path optimization method based on dynamic obstacle avoidance as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, a path optimization method for an inspection robot based on dynamic obstacle avoidance as described in any one of claims 1 to 6 is implemented.

Images