Path planning method and device for pipeline inspection robots in oil refineries

Abstract

This invention provides a path planning method and device for a pipeline inspection robot in an oil refinery, relating to the field of robot path planning technology. The invention simultaneously acquires an infrared temperature matrix, a 3D point cloud, background ambient temperature, a global gravity vector, and a scalar velocity. It locates the coordinates of abnormal heat sources through cross-modal mapping of infrared anomaly features and the 3D point cloud, and calculates the physical distance to the chassis. It generates a spatial thermal risk matrix using relative temperature difference features and environmental thermal attenuation properties. It determines a safe yaw direction based on the gradient descent of the matching mesh and generates a thermal repulsion vector by combining repulsion gain modulation determined based on the distance deficit. It spatially fuses the thermal repulsion with the global gravity vector to calculate the obstacle avoidance guidance angle and uses the velocity damping exponential decay characteristic to perform nonlinear amplitude suppression, generating a comprehensive obstacle avoidance yaw angle to drive the mobile chassis to perform obstacle avoidance. This achieves accurate obstacle avoidance and efficient path planning for the robot in complex thermal environments.

Description

Technical Field

[0001] This invention relates to the field of robot path planning technology, specifically to a method and apparatus for path planning of a robot for pipeline inspection in an oil refinery. Background Technology

[0002] With the continuous development of modern oil refining technology, dense pipeline areas present dual safety hazards due to their complex physical spatial layout and dynamic evolution of high-temperature thermal radiation fields. Traditional inspection methods rely excessively on manual labor or fixed monitoring facilities, resulting in shortcomings such as scattered sampling points, numerous blind spots, and low detection efficiency. In recent years, mobile inspection equipment has become an important means of monitoring pipeline conditions due to its autonomous operation advantages. However, the thermal environment inside oil refineries is extremely complex, and traditional path planning methods relying on single-vision or simple radar ranging are insufficient to ensure the safe passage of equipment in complex pipeline networks.

[0003] Existing robot path planning methods typically involve pre-installing radar or single-point thermometers on the chassis to assess environmental accessibility by collecting geometric distance or instantaneous surface temperature in a single dimension. These methods often rely on converting sensor signals into discrete commands to trigger alarms after simple threshold comparisons. During obstacle avoidance, current technologies generally employ empirically set fixed safety distance limits, triggering mechanical steering by determining whether physical distances exceed these limits. When processing multimodal data, such solutions often treat thermal radiation temperature and physical pose as independent parameters, focusing on monitoring the absolute values ​​of geometric coordinates and lacking a global assessment of radiation risks.

[0004] In actual oil refinery environments, there is a complex nonlinear time-delay coupling relationship between the spatial evolution of abnormal thermal radiation and the dynamic displacement of the robot. The linear accumulation of a single physical distance is insufficient to accurately reflect the deep thermal anomalies caused by the chassis being forced to approach. On the one hand, existing obstacle avoidance planning methods struggle to establish an effective mapping between the thermal field gradient propagation law and the robot's physical pose, easily leading to path oscillations or delayed obstacle avoidance response when approaching the boundary of a high-risk heat source. On the other hand, existing technologies often neglect the dynamic constraints brought about by the chassis's own inertial motion and physical velocity when performing obstacle avoidance vector fusion, failing to fully consider the engineering hazards such as trajectory instability that can easily be caused by sharp turns. This makes it difficult to accurately quantify the true safety boundary of the obstacle avoidance path under multi-source thermal disturbances.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a path planning method and apparatus for a pipeline inspection robot in an oil refinery, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A path planning method for a pipeline inspection robot in an oil refinery, comprising the following steps: Step 1: Obtain the infrared temperature matrix, three-dimensional point cloud, background ambient temperature, global gravity vector, and scalar velocity of the center of the moving chassis of the preset forward detection field according to the preset sampling frequency. Perform heat value extreme value retrieval to extract infrared anomaly features and the highest anomaly temperature. Combine the three-dimensional point cloud cross-modal mapping to locate the coordinates of the anomaly heat source and calculate its physical distance from the center of the moving chassis. Step 2: Obtain the relative temperature difference feature based on the difference between the highest abnormal temperature and the background ambient temperature. Combined with the environmental heat attenuation attribute, and using the mapping logic that the relative temperature difference feature is nonlinearly negatively correlated with the distance from each point to the abnormal heat source coordinates, calculate the radiation risk index of each spatial location in the preset forward detection field and encapsulate it to generate a spatial thermal risk matrix. Step 3: Determine the safe yaw direction vector based on the gradient descent of the grid matching the center of the mobile chassis in the spatial thermal risk matrix, determine the repulsion gain based on the degree of loss of the chassis physical distance relative to the preset critical distance, and perform vector amplitude modulation on the safe yaw direction vector to generate a thermal repulsion vector. Step 4: Perform spatial vector fusion using thermal repulsion vector and global gravity vector to solve obstacle avoidance guidance angle. Use the exponential decay characteristics of scalar velocity and preset velocity damping coefficient to perform nonlinear amplitude suppression on obstacle avoidance guidance angle, and generate comprehensive obstacle avoidance yaw angle to drive the mobile chassis to perform obstacle avoidance.

[0008] Furthermore, the data acquired at the preset sampling frequency also includes real-time angular velocity and real-time wheel speed; The preset forward detection field is a fan-shaped area extending from the center of the mobile chassis as the origin, along the current heading angle toward the preset detection depth. The current heading angle is the azimuth angle relative to the global reference coordinate system obtained by inputting the real-time angular velocity and real-time wheel speed into a preset filtering algorithm and calculating through dead reckoning. The logic for obtaining the background ambient temperature is as follows: The temperature values ​​of each pixel in the infrared temperature matrix at the current sampling time are statistically analyzed and a temperature distribution histogram is constructed. High-temperature pixel clusters that are in the preset high-temperature probability range are identified and blocked. The average value operation is performed on the remaining pixel set after blocking to obtain the background ambient temperature.

[0009] Furthermore, the specific steps for extracting the infrared anomaly features and the highest anomaly temperature are as follows: The temperature values ​​of each pixel in the infrared temperature matrix are traversed. Pixels with temperatures higher than the background ambient temperature are identified as infrared anomalous features. The highest temperature value among the infrared anomalous features is obtained by numerical comparison as the highest anomalous temperature, and the row and column index of the highest temperature value is recorded as the target pixel coordinates. The specific steps for locating the coordinates of the abnormal heat source and calculating the physical distance to the chassis are as follows: The target pixel coordinates are projected into a three-dimensional direction vector using a preset infrared intrinsic parameter. The three-dimensional direction vector is then transformed into the coordinate space of the three-dimensional point cloud using a preset extrinsic parameter matrix. In the three-dimensional point cloud, a nearest neighbor search is performed along the transformed three-dimensional direction vector to lock the coordinates of the abnormal heat source in the three-dimensional space. Extract the coordinate deviation values ​​between the abnormal heat source coordinates and the center of the mobile chassis in the three-axis direction, calculate the sum of squares of the coordinate deviation values ​​and perform a square root operation to obtain the physical distance of the chassis.

[0010] Furthermore, the specific steps for obtaining the relative temperature difference feature quantity and determining the distance between grid points corresponding to each spatial location within the preset forward detection field are as follows: Based on the physical dimensions and preset resolution of the preset forward detection field, spatial discretization is performed to construct a discrete grid point set composed of multiple spatial locations; Calculate the difference between the highest abnormal temperature and the background ambient temperature to obtain the relative temperature difference characteristic quantity; For each spatial location in the discrete grid point set, the spatial Euclidean distance between it and the coordinates of the abnormal heat source is calculated, and used as the grid point distance corresponding to each spatial location.

[0011] Furthermore, the steps for constructing the space thermal risk matrix are as follows: The distance between the grid points is summed with a preset anti-singularity constant. The ratio of the relative temperature difference characteristic to the sum is used as the base, and a preset thermal dissipation power exponent is used as the exponent to perform a power-law decay term. The exponential damping term is constructed by using the negative of the product of the preset environmental thermal damping coefficient and the distance to the grid points as the exponent of the natural constant. The power-law attenuation term and the exponential damping term are multiplied to calculate the radiation risk index corresponding to each spatial location. Each radiation risk index is then stored in a discrete grid point set according to its corresponding spatial coordinate index to generate a spatial thermal risk matrix. The specific calculation formula for the radiation risk index is as follows: In the formula, Spatial location The corresponding radiation risk index, It is a characteristic quantity of relative temperature difference. Spatial location The corresponding grid point distance, To preset the anti-singularity constant, To preset the power exponent of thermal divergence, This is the preset environmental thermal damping coefficient.

[0012] Furthermore, the specific steps for generating the thermal repulsion vector are as follows: The spatial gradient of the local grid and its neighborhood that matches the center of the mobile chassis in the spatial thermal risk matrix is ​​extracted using a preset spatial difference operator, and the spatial gradient is reversed by a negative operation to obtain a negative gradient vector as a safe yaw direction vector. Calculate the difference between the reciprocal of the physical distance of the chassis and the reciprocal of the preset critical distance to obtain the distance reciprocal difference value; The maximum value is obtained by taking the difference between the constant zero and the reciprocal of the distance, and the cube of the obtained maximum value is calibrated as the repulsive force gain. The thermal repulsion vector is calculated by performing vector amplitude modulation on the safe yaw direction vector using the repulsion gain, and the specific calculation formula is as follows: In the formula, Let this be the vector of the thermal repulsion force experienced at the center of the mobile chassis. To extract the spatial gradient at the current coordinates for the spatial thermal risk matrix, This refers to the physical distance of the chassis. This is a preset critical distance.

[0013] Furthermore, the specific steps for generating the integrated obstacle avoidance yaw angle are as follows: Calculate the spatial cross product of the thermal repulsion vector and the global gravity vector and extract its magnitude as an orthogonal component; Calculate the spatial dot product of the thermal repulsion vector and the global gravity vector, and sum it with the square of the magnitude of the global gravity vector to obtain the projection reference component; The deflection angle of the orthogonal component relative to the projection reference component is calculated using the standard arctangent function to obtain the obstacle avoidance guidance angle; Calculate the power value with the natural constant as the base and the negative of the product of the preset speed damping coefficient and the scalar speed at the center of the moving chassis as the exponent, and use it as the damping attenuation factor; The obstacle avoidance guidance angle is multiplied by the damping attenuation factor to calculate the comprehensive obstacle avoidance yaw angle. The specific calculation formula is as follows: In the formula, To calculate the overall obstacle avoidance yaw angle, The thermal repulsion vector, The global gravity vector. The preset velocity damping coefficient, The scalar velocity is the velocity at the center of the moving chassis.

[0014] Furthermore, the specific steps for generating the integrated obstacle avoidance yaw angle to drive the mobile chassis to perform obstacle avoidance are as follows: Obtain the preset command execution cycle and the preset physical wheel track, divide the comprehensive obstacle avoidance yaw angle by the preset command execution cycle, and calculate the target angular velocity of the mobile chassis within the current preset command execution cycle; Calculate the product of the target angular velocity and the preset physical wheel track, and divide the resulting product by a constant two to obtain the single-wheel linear velocity compensation amount used for chassis steering; Using the scalar velocity as the reference for the center of mass linear velocity, the single wheel linear velocity compensation is subtracted from the scalar velocity to calculate the expected execution speed of the left drive wheel of the mobile chassis. The scalar speed is summed with the single-wheel linear speed compensation to calculate the desired execution speed of the right drive wheel of the mobile chassis, so as to drive the mobile chassis to perform deflection obstacle avoidance.

[0015] The present invention also provides a path planning device for a refinery pipeline inspection robot, which is used to implement the above-mentioned path planning method for a refinery pipeline inspection robot, comprising: The heat source localization module is used to acquire the infrared temperature matrix, three-dimensional point cloud, background ambient temperature, global gravity vector and scalar velocity of the center of the moving chassis of the preset forward detection field according to the preset sampling frequency, perform heat value extreme value retrieval to extract infrared anomaly features and the highest anomaly temperature, and combine the three-dimensional point cloud cross-modal mapping to locate the coordinates of the abnormal heat source and calculate its physical distance from the center of the moving chassis. The risk matrix construction module is used to obtain the relative temperature difference feature based on the difference between the highest abnormal temperature and the background ambient temperature. Combined with the environmental heat attenuation attribute, it uses the mapping logic that the relative temperature difference feature is nonlinearly negatively correlated with the distance from each point to the abnormal heat source coordinates to calculate the radiation risk index of each spatial location in the preset forward detection field and encapsulate it to generate a spatial thermal risk matrix. The thermal repulsion generation module is used to determine the safe yaw direction vector based on the gradient descent of the grid that matches the center of the mobile chassis in the spatial thermal risk matrix, determine the repulsion gain based on the degree of loss of the chassis physical distance relative to the preset critical distance, and perform vector amplitude modulation on the safe yaw direction vector to generate the thermal repulsion vector. The obstacle avoidance yaw calculation module is used to perform spatial vector fusion using thermal repulsion vector and global gravity vector to calculate obstacle avoidance guidance angle. It uses the exponential decay characteristics of scalar velocity and preset velocity damping coefficient to perform nonlinear amplitude suppression on obstacle avoidance guidance angle, and generates comprehensive obstacle avoidance yaw angle to drive the mobile chassis to perform obstacle avoidance.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention accurately locates abnormal heat sources by cross-modal mapping of infrared anomaly features and 3D point clouds, and constructs a spatial thermal risk matrix by combining relative temperature difference and nonlinear environmental attenuation properties. Subsequently, a thermal repulsion vector is dynamically generated based on the matrix gradient descent and the physical distance to the chassis. This invention overcomes the shortcomings of traditional methods in perceiving the evolution of spatial temperature gradients. By deeply fusing dynamic high-temperature radiation fields with 3D geometric pose, it effectively quantifies the hidden high-thermal radiation distortion inside complex utility tunnels, significantly improving the robot's risk perception sensitivity in multi-source thermal environments. This invention further calculates the comprehensive obstacle avoidance yaw angle by fusing the thermal repulsion vector with the global gravity vector space and utilizing the exponential decay characteristic of the chassis scalar velocity to perform nonlinear amplitude suppression. This optimizes the robot's steering response with motion inertia, solving the obstacle avoidance lag problem caused by the inability of traditional fixed thresholds to adapt to dynamic thermal field offsets. This mechanism achieves early locking of the real safety boundary and smooth obstacle avoidance in extreme environments. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a mapping diagram showing the relationship between the chassis physical distance, obstacle avoidance yaw angle, and the expected execution speed of the two wheels; Figure 3 This is a schematic diagram of the overall device structure of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] Example: Please see Figures 1-2 The present invention provides a technical solution: A path planning method for a pipeline inspection robot in an oil refinery, comprising the following steps: Step 1: Obtain the infrared temperature matrix, three-dimensional point cloud, background ambient temperature, global gravity vector, and scalar velocity of the center of the moving chassis from the preset forward detection field at the preset sampling frequency. Perform thermal extreme value retrieval to extract infrared anomaly features and the highest anomaly temperature. Combine the three-dimensional point cloud cross-modal mapping to locate the coordinates of the anomaly heat source and calculate its physical distance from the center of the moving chassis.

[0021] In this embodiment, to prevent excessively long observation intervals from causing a delay in the perception of sudden thermal events, environmental data is continuously acquired at a preset sampling frequency to ensure that transient changes in background ambient temperature and the highest abnormal temperature can be captured. The preset sampling frequency is set to... When moving the chassis to During normal scalar speed cruising, this preset sampling frequency means that the device can acquire the latest thermal and spatial data every 0.1 seconds, limiting the displacement of the mobile chassis to 0.1 meters within a single sampling cycle. This setting ensures that cross-modal mapping operations keep up with minor environmental changes and prevents the chassis's physical distance from exceeding the safety limit due to data update delays.

[0022] Meanwhile, due to the omnidirectional thermal scattering within the tunnel, this embodiment sets up a preset forward detection field to avoid introducing irrelevant backward thermal interference. The preset forward detection field is a fan-shaped area extending from the center of the mobile chassis as the origin, along the current heading angle towards a preset detection depth. The current heading angle is the azimuth angle relative to the global reference coordinate system, obtained by inputting real-time angular velocity and real-time wheel speed into a preset filtering algorithm and calculating dead reckoning. That is, during the mobile chassis calibration phase, its initial absolute orientation in the global reference coordinate system is extracted first and used as the initial zero point and accumulation benchmark for subsequent dead reckoning. Since random high-frequency noise is unavoidable under complex operating conditions, this embodiment uses an extended Kalman filter algorithm as the preset filtering algorithm to forcibly suppress long-term divergence errors in the dead reckoning process, ensuring that the calculated current heading angle is always accurately aligned with the actual physical trajectory of the chassis. By extracting the mechanical structural parameters of the mobile chassis, the difference in real-time wheel speed is divided by the physical wheelbase, multiplied by a preset command execution cycle, and then added to the angle from the previous moment to construct a nonlinear kinematic equation for preliminary calculation of the uncorrected predicted heading angle. To quantify the error data in the calculation process and physical measurement, the static variance of the sensors is extracted to fill the observation noise matrix, and a process noise matrix is ​​set based on the physical slip rate. Finally, the Kalman gain constructed by the observation noise matrix and the process noise matrix is ​​used to perform weighted compensation for the predicted heading angle. Since the entire process is always anchored to the global initial physical reference, and the single-step kinematic error is effectively filtered out by the filtering gain, the current heading angle is calculated. By combining the actual physical spacing of the pipe gallery and the chassis braking requirements, this embodiment sets the angle of the sector region to 120 degrees and the preset detection depth to 3.0 meters. This size just matches the maximum physical safety boundary required for the mobile chassis to perform yaw obstacle avoidance and emergency braking at normal cruising speed.

[0023] Multidimensional sensing parameters are acquired in a preset forward detection field. The infrared temperature matrix provides the thermal radiation field strength distribution characteristics of the two-dimensional plane; the three-dimensional point cloud depicts the physical geometric contours of the intricate pipes; the global gravity vector anchors the target tendency at the macroscopic level of the chassis; and the scalar velocity characterizes the current kinematic inertia of the chassis. In this embodiment, a pure directional landmark unaffected by distance is extracted for the chassis by acquiring the global gravity vector. First, the three-dimensional coordinates of the next target node that the moving chassis is about to reach, as well as the real-time three-dimensional coordinates of the moving chassis, are extracted. Subtracting the three-dimensional coordinates of the target node from the current three-dimensional coordinates generates an original direction vector pointing directly to the destination in space. Since the length of this original direction vector directly depends on the actual physical span between the chassis and the target, if the target is extremely far away, its large value can easily interfere with near-field obstacle avoidance actions in subsequent calculations. Therefore, by calculating the absolute physical length of the original direction vector, and then dividing each of its coordinate components by this length, a pure geometric pointing unit vector with a constant magnitude of one is extracted and calibrated as the global gravity vector. To accurately isolate the truly hazardous thermal areas from a complex infrared background, this embodiment constructs a temperature distribution histogram by statistically analyzing the temperature values ​​of each pixel in the infrared temperature matrix at the current sampling time. High-temperature pixel clusters within a preset high-temperature probability range are identified and masked. The remaining set of masked pixels is then averaged to obtain the background ambient temperature. This averaging process effectively prevents localized intense heat sources from contaminating the overall environmental baseline. Specifically, the nominal thermal radiation temperature frequency band under normal operating conditions is extracted, and under a safe baseline condition without abnormal leaks, the area is scanned multiple times using an infrared sensor to construct a baseline temperature distribution histogram. The high-temperature side of this baseline temperature distribution histogram is ranked first... to The statistical interval is defined as the current preset high temperature probability interval.

[0024] In this embodiment, by traversing the temperature values ​​of each pixel in the infrared temperature matrix and performing a thermal extreme value search, pixels with temperatures higher than the background ambient temperature are identified as infrared anomaly features. The highest temperature value among the infrared anomaly features is obtained through numerical comparison as the highest anomaly temperature, and the row and column index of the highest temperature value is recorded as the target pixel coordinates. The target pixel coordinates are projected into a three-dimensional direction vector using preset infrared intrinsic parameters, and the three-dimensional direction vector is transformed into the coordinate space of the three-dimensional point cloud using a preset extrinsic parameter matrix. In the three-dimensional point cloud, a nearest neighbor search is performed along the transformed three-dimensional direction vector to locate the coordinates of the abnormal heat source in the three-dimensional space. The coordinate deviation values ​​between the abnormal heat source coordinates and the center of the mobile chassis in the three-axis directions are extracted, the sum of the squares of the coordinate deviation values ​​is calculated, and the square root operation is performed to obtain the physical distance of the chassis, further quantifying the absolute spatial distance between the mobile chassis and the core hazard source. The preset infrared intrinsic parameters are obtained by acquiring multiple sets of calibration images at different spatial angles and distances using a feature calibration plate with a constant-temperature heat source; extracting the two-dimensional pixel coordinates of feature corner points in each image; and then solving for the intrinsic parameter data characterizing the lens focal length component and optical principal point component using Zhang's calibration method. The preset infrared intrinsic parameters are a... The matrix, in this embodiment, sets the preset infrared intrinsic parameters to Where 600 represents the estimated focal length parameter, and 320 and 256 represent the pixel coordinates of the optical center on the image plane. A joint calibration target with significant thermal radiation differences and a physically geometric convex profile is placed in front of the moving chassis; the corresponding coordinates of the same feature points in the two-dimensional image plane and three-dimensional physical space are extracted by acquiring the three-dimensional point cloud and infrared temperature matrix, respectively; finally, the rigid body transformation relationship between these two sets of spatial coordinates is obtained through singular value decomposition, thereby determining the relative rotation angle and translation distance to construct a preset extrinsic parameter matrix. The preset extrinsic parameter matrix is ​​a... The homogeneous transformation matrix (which internally consists of a) The rotation submatrix, a The translation column vector and the constant row at the bottom (Assembled from pieces).

[0025] Step 2: Obtain the relative temperature difference feature based on the difference between the highest abnormal temperature and the background ambient temperature. Combined with the environmental thermal attenuation attribute, and using the mapping logic that the relative temperature difference feature is nonlinearly negatively correlated with the distance from each point to the abnormal heat source coordinates, calculate the radiation risk index of each spatial location in the preset forward detection field and encapsulate it to generate a spatial thermal risk matrix.

[0026] In this embodiment, spatial discretization is performed based on the physical dimensions and preset resolution of the preset forward detection field to construct a discrete grid point set composed of multiple spatial locations. Specifically, the vertical row number of the array is obtained by dividing the vertical physical depth of the preset forward detection field by the preset resolution, and the horizontal column number of the array is obtained by dividing the maximum horizontal width by the preset resolution. In this embodiment, the discrete grid point set is set to 60 rows. A 104-column two-dimensional spatial topology matrix. Within this two-dimensional spatial topology matrix, each valid spatial location independently encapsulates its row and column topology indices, actual physical relative coordinates, grid point distances, and the calculated radiation risk index.

[0027] Since continuous physical space contains an infinite number of coordinate micro-elements, this embodiment sets an absolute upper limit threshold for the grid size by extracting the physical envelope feature scale of the mobile chassis, and sets a lower limit constraint for the grid size by extracting the computing power threshold of the single-cycle operation of the underlying controller. Within the above upper and lower limit constraint range, the corresponding spatial size that can be divided evenly by the minimum detection angular resolution of the front-end environmental sensor is selected and established as the current preset resolution. In this embodiment, the preset resolution is set to 0.05 meters to establish a quantitative benchmark for crossing from the continuous physical thermal field to the discrete topological matrix. The physical scale of 5 centimeters is slightly larger than the initial diameter of radiation dispersion when there is a small leak in a conventional pressure pipeline, and much smaller than the physical width of the chassis drive tires. This ensures that the generated spatial thermal risk matrix has sufficiently delicate spatial gradient guiding features and prevents the chassis from experiencing local mechanical collisions within a single grid. In this embodiment, based on the angle of the fan-shaped region extending towards the preset detection depth along the current heading angle and the preset detection depth, the physical size of the preset forward detection field is set to a longitudinal physical depth of 3.0 meters and a maximum lateral width of 5.2 meters. Used to define the maximum physical safety boundary required for emergency braking and yaw of a mobile chassis under normal cruise conditions.

[0028] For each spatial location within the discrete grid point set, the spatial Euclidean distance between that location and the coordinates of the anomalous heat source is calculated, serving as the grid point distance for each spatial location. This quantifies the propagation span of thermal radiation in the spatial dimension and establishes the spatial independent variable for calculating the energy gradient. Simultaneously, the difference between the highest anomalous temperature and the background ambient temperature is calculated to obtain a relative temperature difference characteristic. By using this difference calculation to eliminate interference from the background noise of the conventional physical environment, the net increase in dangerous heat energy caused by pipeline leaks is accurately identified.

[0029] After obtaining the initial thermal boundary, the radiation risk index of each spatial location is calculated using a mapping logic that shows a nonlinear negative correlation between the relative temperature difference characteristic and the distance from each point to the coordinates of the anomalous heat source. In this embodiment, the distances between the grid points are summed with a preset anti-singularity constant. Using the ratio of the relative temperature difference characteristic to the sum as the base and a preset thermal divergence exponent as the exponent, a power-law attenuation term is obtained, aiming to simulate the cliff-like energy attenuation characteristic of thermal radiation in the near-field region due to geometric divergence. The specific calculation formula is as follows: In the formula, Spatial location The corresponding value of the power-law decay term, It is a characteristic quantity of relative temperature difference. Spatial location The corresponding grid point distance, To preset the anti-singularity constant, The preset thermal dissipation power index.

[0030] The power-law attenuation term characterizes the basic thermal radiation hazard intensity of spatial grid nodes under the influence of spatial geometric diffusion effects alone. A large value indicates that the target spatial point is located in the near-field core region of the anomalous heat source; a small value indicates that the physical spatial span is sufficient to completely dilute the initial hazardous thermal energy. This power-law attenuation term uses the grid point distance as the denominator variable, mapping the objective attenuation law of drastic dilution of heat flux per unit area as the physical span extends; it uses the relative temperature difference characteristic as the numerator variable, ensuring a positive physical mapping between the intensity of the risk source and the overall radiation field characteristics. A preset anti-singularity constant is introduced and superimposed on the distance variable to further avoid overflow collapse caused by the physical distance approaching zero when the chassis infinitely approaches the center of the heat source. By extracting the shortest physical radial distance from the center of the moving chassis to its outermost mechanical protective shell, and simultaneously extracting the absolute physical detection blind zone distance of the front-end sensor, these two physical dimensions are compared, and the maximum value is selected as the current preset anti-singularity constant. This embodiment sets the preset anti-singularity constant to 0.1, eliminating chassis collapse caused by numerical overflow at extremely close distances. Finally, by performing a global power operation through the preset thermal divergence exponent of the outer layer, the nonlinear constraint of the heat flow diffusion rate on the specific tunnel topology dimension is quantified. In this embodiment, as heat radiation diffuses outward in three-dimensional space, its energy density inevitably decreases geometrically as the area of ​​the spherical or fan-shaped geometric diffusion surface increases. This embodiment applies the spherical wave geometric diffusion principle from classical physics, assigning the preset thermal divergence exponent to 2.0 to quantify the steepness of the risk core area boundary.

[0031] This embodiment utilizes the negative of the product of the preset environmental thermal damping coefficient and the distance to the grid points as the exponent of the natural constant to perform exponential calculations, constructing an exponential damping term to quantify the physical absorption and scattering of thermal radiation by the air medium in the far-field region. The specific calculation formula is as follows: In the formula, Spatial location The corresponding value of the exponential damping term, Spatial location The corresponding grid point distance, This is the preset environmental thermal damping coefficient.

[0032] The exponential damping term characterizes the absolute physical loss rate of thermal radiation at the spatial grid nodes as it penetrates the solid medium within the pipe gallery. A large value indicates a very short physical span for thermal radiation or an extremely thin spatial medium; a small value indicates a very long physical span for thermal radiation or an extremely dense spatial medium, meaning that a large amount of the initial dangerous thermal energy has already been absorbed. This exponential damping term takes the negative of the product of the preset environmental thermal damping coefficient and the grid point distance, and uses it as the exponent of the natural constant, accurately mapping the nonlinear absorption characteristics of thermal energy as the penetration depth increases. The grid point distance directly determines the cumulative frequency of collisions between thermal radiation photons and air particles; the greater the distance, the more exponentially the cumulative absorption loss increases. The preset environmental thermal damping coefficient objectively anchors the density and absorption properties of the medium under the current operating conditions; a larger value results in stronger thermal energy attenuation, effectively ensuring that the chassis avoids being misled by ineffective background heat sources at a distance. This embodiment extracts macroscopic medium attenuation characteristics within the current operating space. Based on the inherent physical absorption characteristics of this type of medium for infrared radiation bands, it deduces and calibrates the corresponding equivalent attenuation rate as the preset environmental thermal damping coefficient. In this embodiment, the preset environmental thermal damping coefficient is set to 0.05, effectively filtering out irrelevant background thermal disturbances at a distance, significantly improving the chassis's macroscopic cruising efficiency within the safe zone.

[0033] Finally, the power-law attenuation term and the exponential damping term are multiplied to calculate the radiation risk index corresponding to each spatial location, which is used to quantify the hazard intensity of each independent node. Each radiation risk index is stored in a discrete grid point set according to its corresponding spatial coordinate index to generate a spatial thermal risk matrix. The specific calculation formula for the radiation risk index is as follows: In the formula, Spatial location The corresponding radiation risk index, For power-law decay terms, It is a characteristic quantity of relative temperature difference. Spatial location The corresponding grid point distance, To preset the anti-singularity constant, To preset the power exponent of thermal divergence, For exponential damping, This is the preset environmental thermal damping coefficient.

[0034] The radiation risk index characterizes the absolute danger intensity of a specific spatial grid node subjected to the combined thermal radiation from an abnormal heat source. A large value indicates an extremely high risk of direct thermal damage to the chassis; a small value indicates that the thermal energy has been fully dissipated in space, and the area is within safe physical limits. This radiation risk index accurately recreates the true physical evolution of thermal radiation in the intricate utility tunnel by multiplying a power-law decay term with an exponential damping term. The power-law decay term anchors the initial energy base of the radiation burst and maps the near-field divergence effect caused by the sudden drop in heat due to spatial geometric expansion; while the exponential damping term quantifies the continuous absorption and loss of thermal energy by the far-field physical medium. Through multiplicative coupling, this formula reconstructs the true dissipation law: the near field is controlled by a power-law decay term triggered by the minimum distance between grid points, maintaining extremely high sensitivity to displacement to trigger obstacle avoidance; the far field utilizes the sudden decrease characteristic of the exponential damping term to directly suppress ineffective weak thermal disturbances.

[0035] Step 3: Determine the safe yaw direction vector based on the gradient descent of the grid matching the center of the mobile chassis in the spatial thermal risk matrix. Determine the repulsion gain based on the degree of loss of the chassis physical distance relative to the preset critical distance. Then, perform vector amplitude modulation on the safe yaw direction vector to generate a thermal repulsion vector.

[0036] In this embodiment, since the spatial thermal risk matrix is ​​essentially used to depict the intensity of local thermal energy hazard, and continuous partial differentials cannot be directly applied to discrete digital arrays, a preset spatial difference operator must be introduced as a mathematical probe to extract the rate of environmental change. This embodiment introduces a preset spatial difference operator to establish a quantitative evaluation benchmark for the absolute difference between local grids, transforming the static scalar risk field into a spatially oriented dynamic vector field. Specifically, this embodiment extracts the matrix convolution processing limit of the underlying computational unit and the characteristic thermal-to-noise ratio of the front-end infrared sensor. Within this dual constraint boundary, it matches the convolution kernel template size and corresponding direction weight matrix that maximizes the smoothing of high-frequency noise without causing computational lag. In this embodiment, the preset spatial difference operator is set as... The Sobel operator, with its specific dimensions, assigns higher weights to adjacent grids in the orthogonal direction than to grids in the diagonal direction when calculating the difference, effectively suppressing spurious extremum interference caused by single sensor pixel jumps or micro-dust occlusion. By utilizing a preset spatial difference operator to extract the spatial gradient of the local grid and its neighborhood that matches the center of the mobile chassis in the spatial thermal risk matrix, the steepness of the risk on the current physical slice can be accurately quantified. The spatial gradient always points to the core region where the risk value rises the fastest, i.e., where thermal radiation is most concentrated. Therefore, a negative inversion operation is performed on this spatial gradient to obtain a geometric ray that runs completely in the opposite direction, resulting in a negative gradient vector as the safe yaw direction vector.

[0037] The difference between the reciprocal of the physical distance to the chassis and the reciprocal of the preset critical distance is calculated to obtain the distance reciprocal difference value. Using the reciprocal form to calculate the difference ensures that as the chassis gradually approaches the hazard source, the reciprocal of its physical distance will increase rapidly in a hyperbolic manner. A maximum value operation is performed on the difference between the constant zero and the distance reciprocal to ensure that the chassis is outside the preset critical distance (at which point the difference is negative). The maximum value operation forces a zero output; and the cube of the obtained maximum value is calibrated as the repulsive force gain, used to utilize a nonlinear amplification mechanism. The specific calculation formula is as follows: In the formula, For repulsive force gain, This refers to the physical distance of the chassis. This is a preset critical distance.

[0038] The repulsive force gain is used to characterize the degree of extreme nonlinear dynamic intervention of the mobile chassis when it takes over the underlying drive components. A large repulsive force gain indicates an extremely high risk of thermal deformation; a zero repulsive force gain indicates that the chassis is outside the preset defense zone. The calculation process obtains the difference between the reciprocal of the chassis physical distance and the reciprocal of the preset critical distance to derive the distance reciprocal difference value. The preset critical distance constructs a physical defense red line, establishing the absolute spatial boundary that distinguishes between normal cruise and emergency obstacle avoidance, and anchoring the trigger threshold of the underlying nonlinear intervention mechanism. In this embodiment, the maximum braking physical distance of the mobile chassis under full load and normal scalar speed is extracted and summed with the shortest safe space distance required to reach the thermal deformation critical temperature to calibrate the preset critical distance. Since the physical depth required to envelop the chassis and eliminate extreme forward inertia needs to be considered, this embodiment sets the preset critical distance to 1.5 meters, strictly quantifying the minimum absolute spatial bottom line for the hardware to be protected from irreversible thermal damage, ensuring that the mobile chassis has sufficient physical buffer span to complete a complete stop or large-angle deflection after triggering the extreme repulsive force. After establishing this boundary, the chassis physical distance serves as a dynamic parameter, determining the intensity of spatial loss. When the chassis physical distance exceeds the preset critical distance, the maximum value extraction operation is performed, and this difference is negative, thus being forcibly covered by the constant zero, completely severing the dynamic pull of unrelated heat sources at a distance. Once the chassis crosses the boundary and approaches the danger source, the chassis physical distance continuously decreases, and the generated positive distance reciprocal difference increases sharply, accurately fitting the singularity of the instantaneous repulsion trend towards infinity before the collision. Furthermore, a cubic operation is performed on the extracted maximum value, and through a brute-force amplification mechanism of geometric progression, a multiplicative coefficient with absolute crushing authority is forcibly generated in the extremely near-field high-risk zone, driving the chassis to instantly unleash its ultimate escape kinetic energy, resolutely suppressing forward inertia.

[0039] The thermal repulsion vector is calculated by performing vector amplitude modulation on the safe yaw direction vector using the repulsion gain, and the specific calculation formula is as follows: In the formula, Let this be the vector of the thermal repulsion force experienced at the center of the mobile chassis. To extract the spatial gradient at the current coordinates for the spatial thermal risk matrix, For repulsive force gain, This refers to the physical distance of the chassis. This is a preset critical distance.

[0040] The thermal repulsion vector is used to characterize the integrated dynamic intervention command of the mobile chassis taking over the underlying components to execute emergency obstacle avoidance in an extremely dangerous thermal field. When the amplitude of this vector is large, it means that the chassis is in a high-risk area in the near field; when the amplitude is zero, it means that the chassis is outside the preset defense zone. This formula is deeply coupled with the direction calibration and intensity weighting logic. The safe yaw direction vector is obtained by performing a negative reversal operation on the spatial gradient extracted from the spatial thermal risk matrix. Since the spatial gradient always points to the core area where the radiation rises most steeply, the reversal accurately anchors the safe escape ray with the least geometric resistance. The repulsion gain intensity is generated by calculating the difference between the reciprocal of the chassis physical distance and the reciprocal of the preset critical distance, extracting the maximum value and performing a cubic operation. The closer the chassis is to the heat source, the more severe the spatial loss, and the more explosively the multiplier coefficient increases. Finally, the thermal repulsion vector is calculated by performing vector amplitude modulation on the safe yaw direction vector using the repulsion gain intensity. Once the boundary is exceeded, the near field will be forced to erupt along the safety gradient direction by the nonlinear extreme repulsive force to resolutely curb the inertial forward momentum; the far field will be forcibly truncated by the constant zero to ensure that the action is efficient and objective.

[0041] Step 4: Perform spatial vector fusion using thermal repulsion vector and global gravity vector to solve obstacle avoidance guidance angle. Use the exponential decay characteristics of scalar velocity and preset velocity damping coefficient to perform nonlinear amplitude suppression on obstacle avoidance guidance angle, and generate comprehensive obstacle avoidance yaw angle to drive the mobile chassis to perform obstacle avoidance.

[0042] In this embodiment, the global gravity vector and the thermal repulsion vector are uniformly mapped to the same dimensionless geometric coordinate domain. Spatial vector fusion is performed using the thermal repulsion vector and the global gravity vector. The spatial cross product of the thermal repulsion vector and the global gravity vector is calculated, and its magnitude is extracted as the orthogonal component. The spatial dot product of the thermal repulsion vector and the global gravity vector is calculated, and summed with the square of the magnitude of the global gravity vector to obtain the projection reference component. The deflection angle of the orthogonal component relative to the projection reference component is calculated using the standard arctangent function to obtain the obstacle avoidance guidance angle. This completely eliminates the logical conflict between obstacle avoidance actions and the main navigation task, ensuring that the chassis does not spin out or lose its course during obstacle avoidance. The specific calculation formula is as follows: In the formula, For obstacle avoidance guidance angle, The thermal repulsion vector, This is the global gravity vector.

[0043] The obstacle avoidance guidance angle is used to characterize the theoretically optimal deflection heading of a mobile chassis when avoiding local high-risk heat sources, balancing safe escape and macroscopic mission guidance. A large angle indicates strong lateral thermal repulsion, necessitating a large-angle maneuver to escape the danger source; a small angle indicates weak thermal radiation or that its direction aligns with global gravity. This angle calculation is jointly dominated by the thermal repulsion vector and the global gravity vector. The orthogonal component is obtained by calculating the spatial cross product of the thermal repulsion vector and the global gravity vector and extracting their magnitudes, quantifying the lateral tensile intensity of the repulsion force perpendicular to the gravity direction. The projected reference component is obtained by calculating the spatial dot product of the two and summing it with the square of the magnitude of the global gravity vector, thus evaluating the anti-deviation feedforward basis along the original target direction. The obstacle avoidance guidance angle is derived by calculating the deflection angle of the orthogonal component relative to the projection reference component using the standard arctangent function. During the calculation, when the projection reference component approaches zero, the deflection angle is forcibly anchored to an extreme value of ±90 degrees to avoid division-by-zero overflow anomalies in the underlying computation. This formula performs spatial vector fusion of the thermal repulsion vector representing avoidance and the global gravitational vector representing benefit, completely eliminating the heading disorientation caused by a single repulsion force.

[0044] Since large-angle yaws can easily cause tire sideslip and loss of control when the chassis is in high-speed cruising mode, this embodiment extracts the scalar velocity at the center of the moving chassis and calculates a power value with a natural constant as the base and the negative of the product of a preset speed damping coefficient and the scalar velocity at the center of the moving chassis as the exponent, which is then used as the damping attenuation factor. The obstacle avoidance guide angle is multiplied by the damping attenuation factor to establish a negative correlation hardware protection boundary between the steering amplitude and the instantaneous kinetic energy of the chassis. The larger the scalar velocity, the closer the attenuation factor is to zero; the smaller the scalar velocity, the more stable the chassis attitude, and the closer the attenuation factor is to one. After the above nonlinear modulation, the comprehensive obstacle avoidance yaw angle is calculated, and the specific calculation formula is as follows: In the formula, To calculate the overall obstacle avoidance yaw angle, For obstacle avoidance guidance angle, The thermal repulsion vector, The global gravity vector. The preset velocity damping coefficient, The scalar velocity is the velocity at the center of the moving chassis.

[0045] The integrated obstacle avoidance yaw angle is used to characterize the absolute corrective steering command ultimately executed by the underlying hardware of the mobile chassis. A larger value means that the chassis is authorized to perform violent lateral maneuvers to escape the hot zone under low kinetic energy or extremely dangerous conditions; a smaller value means that the thermal threat is weaker. This calculation process is constructed by multiplying the obstacle avoidance guide angle and the damping attenuation factor. The obstacle avoidance guide angle is obtained by fusing the thermal repulsion vector and the global gravity vector. As a basic spatial geometric parameter, it determines the original bypass trajectory that takes into account both escape and target guidance. The larger its initial amplitude, the stronger the deflection base given to the chassis. The damping attenuation factor is calculated using the scalar velocity at the center of the mobile chassis and the preset velocity damping coefficient, reflecting the current motion inertia limit of the chassis. To establish the anti-skid friction limit of the environmental ground surface and anchor the downward slope of the exponential decay, this embodiment extracts the chassis's center of gravity physical height, preset physical wheel track, and the ultimate static friction coefficient between the tires and the current utility tunnel surface. Based on the anti-rollover moment balance law, a preset speed damping coefficient is derived. This embodiment sets the preset speed damping coefficient to 0.8. The reason for adopting this recommended value is that it closely matches the actual physical contact properties between conventional rubber tires and industrial steel grating, ensuring that even if the chassis triggers extreme avoidance at its highest scalar speed, its command amplitude can be forcibly reduced to a safe range allowed by lateral grip. On this basis, since the scalar speed and this coefficient together constitute a negative exponential operation, the higher the scalar speed, the closer this factor is to zero, and the stronger the weakening effect on the steering angle. This multiplicative logic establishes a rigid defense boundary between the steering amplitude and the instantaneous forward impulse energy, and ensures that the chassis fully releases its maneuverability at low speeds.

[0046] In this embodiment, to address the mathematical limitations of the standard arctangent function in cross-quadrant mapping, a phase compensation branch is added for extremely critical conditions. In conventional obstacle avoidance, the projection reference component is positive, and the calculated obstacle avoidance guidance angle is a conventional acute angle in the first quadrant. However, when the moving chassis encounters a lethal heat flow directly in front, the spatial dot product of the thermal repulsion vector and the global gravity vector becomes negative, and its negative absolute value forcibly exceeds the square of the magnitude of the global gravity vector, causing the projection reference component to turn negative. In this case, the standard arctangent function, limited by its mathematical domain, will incorrectly output a negative angle in the fourth quadrant. At this point, the phase compensation branch is immediately triggered, forcibly adding a constant to this negative output. (i.e., 180 degrees) phase offset operation. This logic effectively avoids quadrant folding errors in high-risk oncoming conditions, enabling the mobile chassis to obtain positive obtuse angle commands between 90 degrees and 180 degrees.

[0047] In this embodiment, the specific steps for generating the comprehensive obstacle avoidance yaw angle to drive the mobile chassis to perform obstacle avoidance are as follows: Obtain the preset command execution cycle and the preset physical wheelbase. Since the underlying motor cannot directly execute a static yaw angle, it must be combined with specific time constraints to convert it into a rotational speed that the motor can actually output. In this embodiment, by obtaining the highest physical response frequency of the hardware, within the hardware constraint boundary that ensures stable data flow and does not cause mechanical vibration, the corresponding time interval is selected and defined as the preset command execution cycle. In this embodiment, the preset command execution cycle is set to 0.05 seconds, a value that closely matches the data throughput rhythm of a conventional industrial control bus, completely avoiding communication congestion and motor deadlock caused by excessively frequent command issuance. The comprehensive obstacle avoidance yaw angle is divided by the preset command execution cycle to calculate the target angular velocity of the mobile chassis within the current preset command execution cycle. The product of the target angular velocity and the preset physical wheelbase is calculated, and the resulting product is divided by a constant two to obtain the single-wheel linear velocity compensation amount used for chassis steering. Using the scalar velocity as the center-of-gravity linear velocity reference, the single-wheel linear velocity compensation is subtracted from the scalar velocity to calculate the desired execution speed of the left drive wheel of the mobile chassis. The scalar velocity and the single-wheel linear velocity compensation are then summed to calculate the desired execution speed of the right drive wheel of the mobile chassis, thereby driving the mobile chassis to perform deflection obstacle avoidance. The specific calculation formula is as follows: In the formula, and These are the expected execution speeds of the left and right drive wheels, respectively. The scalar velocity at the center of the moving chassis. To calculate the overall obstacle avoidance yaw angle, To preset the physical wheelbase, This is the preset instruction execution cycle.

[0048] The expected execution speeds of the left and right drive wheels are used to characterize the absolute linear velocity command ultimately tracked by the underlying servo motor. A large difference between the two values ​​indicates that the chassis is undergoing a violent spatial yaw; when the values ​​are equal, it means the chassis is maintaining a straight trajectory. This calculation process uses scalar velocity as a common benchmark for both wheels, determining the physical kinetic energy of the chassis's overall forward translation, ensuring the chassis can quickly traverse dangerous areas during obstacle avoidance. To break the symmetry of the two wheels, the linear velocity compensation of a single wheel needs to be calculated. This is achieved by dividing the comprehensive obstacle avoidance yaw angle by a preset command execution cycle to obtain the target angular velocity. The shorter the time cycle, the greater the instantaneous burst of speed difference is required. This is then multiplied by a preset physical wheelbase and divided by a constant two to calculate the linear velocity compensation of a single wheel. Since the underlying motor only recognizes absolute linear velocity, the center-of-gravity angular velocity must be converted into the tangential velocity of the wheel contact surface using the wheel span. In this embodiment, the straight spatial span between the center lines of the ground imprints of the left and right drive wheels of the moving chassis is directly extracted using a ranging gauge. In this embodiment, the preset physical wheelbase is set to 0.45 meters. This size satisfies the space requirements for mounting the internal core explosion-proof components while ensuring the chassis can smoothly pass through the narrow steel grating passages inside the refinery's pipe gallery. While providing a highly resistant and stable physical base, it ensures that the compensation calculated in a single step is within a reasonable range, completely preventing single-wheel speed command overflow due to excessive lever arm length. Finally, based on the physical laws of rigid body differential kinematics, the desired execution speed of the left drive wheel is calculated by subtracting the single-wheel linear velocity compensation from the scalar velocity, and the desired execution speed of the right drive wheel is calculated by summing the scalar velocity and the single-wheel linear velocity compensation.

[0049] Table 1 is a comprehensive data example of the nonlinear mapping and constraint relationship between the physical distance of the mobile chassis and the comprehensive obstacle avoidance yaw angle and the expected execution speed of the two wheels under the condition of approaching an extremely dangerous heat source. Based on the extreme nonlinear repulsive force intervention and speed damping suppression mechanism, the mobile chassis is shown in this embodiment.

[0050] Table 1: Mapping Relationship between Chassis Physical Distance and Multidimensional Obstacle Avoidance Control Parameters Table 1 shows the parameter mapping for robot path planning in complex, high-risk thermal environments, achieved by sensing the chassis physical distance, global gravitational tendency, and scalar velocity. When the chassis physical distance is within a relatively safe range (items 1-10), the thermal repulsion vector remains at an extremely low level, and is minimally affected by extreme nonlinear dynamic intervention. As the chassis physical distance gradually decreases and approaches the danger source (items 11-17), even though the scalar velocity maintains a high forward momentum, the amplitude of the thermal repulsion vector begins to surge explosively due to the influence of the distance loss cubic mechanism. This causes the integrated obstacle avoidance yaw angle to no longer passively maintain fine-tuning of the heading, but instead to actively and significantly increase. When the chassis physical distance further deviates from the boundary (items 18-25), even if the bottom obstacle avoidance guidance angle is infinitely amplified, the integrated obstacle avoidance yaw angle is subject to strong nonlinear constraints from the velocity damping attenuation factor; simultaneously, the expected execution speed of the left drive wheel is pushed into the negative range (triggering reverse braking), ensuring that fatal thermal damage caused by the chassis's forward inertia is prevented. This table demonstrates that traditional fixed threshold control easily overlooks the physical inertia and sideslip limits of the chassis during high-speed movement. By deeply coupling the extreme space obstacle avoidance requirements with the exponential decay characteristics derived from chassis kinetic energy, path planning can be effectively performed, further reducing the risk of tire loss of control or vehicle rollover.

[0051] Please see Figure 3 The present invention also provides a path planning device for a refinery pipeline inspection robot, which is used to implement the above-mentioned path planning method for a refinery pipeline inspection robot, including: The heat source localization module is used to acquire the infrared temperature matrix, three-dimensional point cloud, background ambient temperature, global gravity vector and scalar velocity of the center of the moving chassis of the preset forward detection field according to the preset sampling frequency, perform heat value extreme value retrieval to extract infrared anomaly features and the highest anomaly temperature, and combine the three-dimensional point cloud cross-modal mapping to locate the coordinates of the abnormal heat source and calculate its physical distance from the center of the moving chassis. The risk matrix construction module is used to obtain the relative temperature difference feature based on the difference between the highest abnormal temperature and the background ambient temperature. Combined with the environmental heat attenuation attribute, it uses the mapping logic that the relative temperature difference feature is nonlinearly negatively correlated with the distance from each point to the abnormal heat source coordinates to calculate the radiation risk index of each spatial location in the preset forward detection field and encapsulate it to generate a spatial thermal risk matrix. The thermal repulsion generation module is used to determine the safe yaw direction vector based on the gradient descent of the grid that matches the center of the mobile chassis in the spatial thermal risk matrix, determine the repulsion gain based on the degree of loss of the chassis physical distance relative to the preset critical distance, and perform vector amplitude modulation on the safe yaw direction vector to generate the thermal repulsion vector. The obstacle avoidance yaw calculation module is used to perform spatial vector fusion using thermal repulsion vector and global gravity vector to calculate obstacle avoidance guidance angle. It uses the exponential decay characteristics of scalar velocity and preset velocity damping coefficient to perform nonlinear amplitude suppression on obstacle avoidance guidance angle, and generates comprehensive obstacle avoidance yaw angle to drive the mobile chassis to perform obstacle avoidance.

[0052] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0053] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0054] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0055] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A path planning method for a pipeline inspection robot in an oil refinery, characterized in that, The specific steps include: The infrared temperature matrix, three-dimensional point cloud, background ambient temperature, global gravity vector, and scalar velocity of the center of the moving chassis are obtained according to the preset sampling frequency. The extreme value retrieval of the heat value is performed to extract the infrared anomaly features and the highest anomaly temperature. The coordinates of the anomaly heat source are located by cross-modal mapping of the three-dimensional point cloud, and the physical distance between the heat source and the center of the moving chassis is calculated. The relative temperature difference feature is obtained by the difference between the highest abnormal temperature and the background ambient temperature. Combined with the environmental thermal attenuation attribute, the relative temperature difference feature is nonlinearly negatively correlated with the distance from each point to the abnormal heat source coordinates. The radiation risk index of each spatial location in the preset forward detection field is calculated and encapsulated to generate a spatial thermal risk matrix. The safe yaw direction vector is determined based on the gradient descent of the grid matching the center of the mobile chassis in the spatial thermal risk matrix. The repulsion gain is determined based on the degree of loss of the chassis physical distance relative to the preset critical distance. The vector amplitude is then modulated on the safe yaw direction vector to generate a thermal repulsion vector. The obstacle avoidance guidance angle is calculated by performing spatial vector fusion using thermal repulsion vector and global gravity vector. The nonlinear amplitude suppression of the obstacle avoidance guidance angle is performed by utilizing the exponential decay characteristics of scalar velocity and preset velocity damping coefficient, and a comprehensive obstacle avoidance yaw angle is generated to drive the mobile chassis to perform obstacle avoidance.

2. The path planning method for a pipeline inspection robot in an oil refinery according to claim 1, characterized in that: The data acquired at the preset sampling frequency also includes real-time angular velocity and real-time wheel speed; The preset forward detection field is a fan-shaped area extending from the center of the mobile chassis as the origin, along the current heading angle toward the preset detection depth. The current heading angle is the azimuth angle relative to the global reference coordinate system obtained by inputting the real-time angular velocity and real-time wheel speed into a preset filtering algorithm and calculating through dead reckoning. The logic for obtaining the background ambient temperature is as follows: The temperature values ​​of each pixel in the infrared temperature matrix at the current sampling time are statistically analyzed and a temperature distribution histogram is constructed. High-temperature pixel clusters that are in the preset high-temperature probability range are identified and blocked. The average value operation is performed on the remaining pixel set after blocking to obtain the background ambient temperature.

3. The path planning method for a pipeline inspection robot in an oil refinery according to claim 2, characterized in that: The specific steps for extracting infrared anomaly features and the highest anomaly temperature are as follows: The temperature values ​​of each pixel in the infrared temperature matrix are traversed. Pixels with temperatures higher than the background ambient temperature are identified as infrared anomalous features. The highest temperature value among the infrared anomalous features is obtained by numerical comparison as the highest anomalous temperature, and the row and column index of the highest temperature value is recorded as the target pixel coordinates. The specific steps for locating the coordinates of the abnormal heat source and calculating the physical distance to the chassis are as follows: The target pixel coordinates are projected into a three-dimensional direction vector using a preset infrared intrinsic parameter. The three-dimensional direction vector is then transformed into the coordinate space of the three-dimensional point cloud using a preset extrinsic parameter matrix. In the three-dimensional point cloud, a nearest neighbor search is performed along the transformed three-dimensional direction vector to lock the coordinates of the abnormal heat source in the three-dimensional space. Extract the coordinate deviation values ​​between the abnormal heat source coordinates and the center of the mobile chassis in the three-axis direction, calculate the sum of squares of the coordinate deviation values ​​and perform a square root operation to obtain the physical distance of the chassis.

4. The path planning method for a pipeline inspection robot in an oil refinery according to claim 1, characterized in that: The specific steps for obtaining the relative temperature difference feature and determining the distance between grid points corresponding to each spatial location within the preset forward detection field are as follows: Based on the physical dimensions and preset resolution of the preset forward detection field, spatial discretization is performed to construct a discrete grid point set composed of multiple spatial locations; Calculate the difference between the highest abnormal temperature and the background ambient temperature to obtain the relative temperature difference characteristic quantity; For each spatial location in the discrete grid point set, the spatial Euclidean distance between it and the coordinates of the abnormal heat source is calculated, and used as the grid point distance corresponding to each spatial location.

5. The path planning method for a pipeline inspection robot in an oil refinery according to claim 4, characterized in that: The steps for constructing the space thermal risk matrix are as follows: The distance between the grid points is summed with a preset anti-singularity constant. The ratio of the relative temperature difference characteristic to the sum is used as the base, and a preset thermal dissipation power exponent is used as the exponent to perform a power-law decay term. The exponential damping term is constructed by using the negative of the product of the preset environmental thermal damping coefficient and the distance to the grid points as the exponent of the natural constant. The power-law attenuation term and the exponential damping term are multiplied to calculate the radiation risk index corresponding to each spatial location. Each radiation risk index is then stored in a discrete grid point set according to its corresponding spatial coordinate index to generate a spatial thermal risk matrix. The specific calculation formula for the radiation risk index is as follows: In the formula, For spatial location The corresponding radiation risk index, It is a characteristic quantity of relative temperature difference. For spatial location The corresponding grid point distance, To preset the anti-singularity constant, To preset the power exponent of thermal divergence, This is the preset environmental thermal damping coefficient.

6. The path planning method for a pipeline inspection robot in an oil refinery according to claim 1, characterized in that: The specific steps for generating the thermal repulsion vector are as follows: The spatial gradient of the local grid and its neighborhood that matches the center of the mobile chassis in the spatial thermal risk matrix is ​​extracted using a preset spatial difference operator, and the spatial gradient is reversed by a negative operation to obtain a negative gradient vector as a safe yaw direction vector. Calculate the difference between the reciprocal of the physical distance of the chassis and the reciprocal of the preset critical distance to obtain the distance reciprocal difference value; The maximum value is obtained by taking the difference between the constant zero and the reciprocal of the distance, and the cube of the obtained maximum value is calibrated as the repulsive force gain. The thermal repulsion vector is calculated by performing vector amplitude modulation on the safe yaw direction vector using the repulsion gain, and the specific calculation formula is as follows: In the formula, Let this be the vector of the thermal repulsion force experienced at the center of the mobile chassis. To extract the spatial gradient at the current coordinates for the spatial thermal risk matrix, This refers to the physical distance of the chassis. This is a preset critical distance.

7. The path planning method for a pipeline inspection robot in an oil refinery according to claim 1, characterized in that: The specific steps for generating the integrated obstacle avoidance yaw angle are as follows: Calculate the spatial cross product of the thermal repulsion vector and the global gravity vector and extract its magnitude as an orthogonal component; Calculate the spatial dot product of the thermal repulsion vector and the global gravity vector, and sum it with the square of the magnitude of the global gravity vector to obtain the projection reference component; The deflection angle of the orthogonal component relative to the projection reference component is calculated using the standard arctangent function to obtain the obstacle avoidance guidance angle; Calculate the power value with the natural constant as the base and the negative of the product of the preset speed damping coefficient and the scalar speed at the center of the moving chassis as the exponent, and use it as the damping attenuation factor; The obstacle avoidance guidance angle is multiplied by the damping attenuation factor to calculate the comprehensive obstacle avoidance yaw angle. The specific calculation formula is as follows: In the formula, To calculate the overall obstacle avoidance yaw angle, The thermal repulsion vector, The global gravity vector. The preset velocity damping coefficient, The scalar velocity is the velocity at the center of the moving chassis.

8. The path planning method for a pipeline inspection robot in an oil refinery according to claim 1, characterized in that: The specific steps for generating the integrated obstacle avoidance yaw angle to drive the mobile chassis to perform obstacle avoidance are as follows: Obtain the preset command execution cycle and the preset physical wheel track, divide the comprehensive obstacle avoidance yaw angle by the preset command execution cycle, and calculate the target angular velocity of the mobile chassis within the current preset command execution cycle; Calculate the product of the target angular velocity and the preset physical wheel track, and divide the resulting product by a constant two to obtain the single-wheel linear velocity compensation amount used for chassis steering; Using the scalar velocity as the reference for the center of mass linear velocity, the single wheel linear velocity compensation is subtracted from the scalar velocity to calculate the expected execution speed of the left drive wheel of the mobile chassis. The scalar speed is summed with the single-wheel linear speed compensation to calculate the desired execution speed of the right drive wheel of the mobile chassis, so as to drive the mobile chassis to perform deflection obstacle avoidance.

9. A path planning device for a pipeline inspection robot in an oil refinery, characterized in that: The refinery pipeline inspection robot path planning device is used to execute the refinery pipeline inspection robot path planning method according to any one of claims 1-8, including: The heat source localization module is used to acquire the infrared temperature matrix, three-dimensional point cloud, background ambient temperature, global gravity vector and scalar velocity of the center of the moving chassis of the preset forward detection field according to the preset sampling frequency, perform heat value extreme value retrieval to extract infrared anomaly features and the highest anomaly temperature, and combine the three-dimensional point cloud cross-modal mapping to locate the coordinates of the abnormal heat source and calculate its physical distance from the center of the moving chassis. The risk matrix construction module is used to obtain the relative temperature difference feature based on the difference between the highest abnormal temperature and the background ambient temperature. Combined with the environmental heat attenuation attribute, it uses the mapping logic that the relative temperature difference feature is nonlinearly negatively correlated with the distance from each point to the abnormal heat source coordinates to calculate the radiation risk index of each spatial location in the preset forward detection field and encapsulate it to generate a spatial thermal risk matrix. The thermal repulsion generation module is used to determine the safe yaw direction vector based on the gradient descent of the grid that matches the center of the mobile chassis in the spatial thermal risk matrix, determine the repulsion gain based on the degree of loss of the chassis physical distance relative to the preset critical distance, and perform vector amplitude modulation on the safe yaw direction vector to generate the thermal repulsion vector. The obstacle avoidance yaw calculation module is used to perform spatial vector fusion using thermal repulsion vector and global gravity vector to calculate obstacle avoidance guidance angle. It uses the exponential decay characteristics of scalar velocity and preset velocity damping coefficient to perform nonlinear amplitude suppression on obstacle avoidance guidance angle, and generates comprehensive obstacle avoidance yaw angle to drive the mobile chassis to perform obstacle avoidance.

Images