A substation robot autonomous path planning and fire extinguishing method based on fire source positioning

Abstract

The application discloses a transformer substation robot autonomous path planning and fire extinguishing method based on a fire source position, relates to the technical field of intelligent control of transformer substation fire fighting, and comprises the following steps: acquiring real-time positioning information of a transformer substation robot and a three-dimensional point cloud map of the transformer substation, mapping to obtain the real-time position of the robot, receiving a full-station fire situation situation map generated by a multi-sensor fusion detection device, planning a global navigation path by relying on an improved potential field method, and the path can bypass static obstacles and avoid high-risk fire spreading areas. The robot is controlled to move along the global path, local obstacle information is collected in real time, a real-time speed control instruction is generated by fusing the global path guidance, the local obstacle information and the motion state of the robot through a dynamic window method. The method can adapt to the three-dimensional complex spatial structure of the transformer substation and the dynamic spreading scene of the fire, realizes accurate navigation and dynamic obstacle avoidance of the robot, and guarantees that the robot safely and stably reaches the fire source position to complete the fire extinguishing operation.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for fire protection in substations, specifically a method for autonomous path planning and fire extinguishing of substation robots based on fire source location. Background Technology

[0002] Currently, most substation firefighting robots rely on two-dimensional maps for path planning. Fire detection devices can only locate the fire source in two dimensions or determine its presence, but cannot collect information on the fire source's three-dimensional coordinates, fire intensity level, and flame spread direction, thus failing to generate complete fire situation data for the entire substation. Conventional path planning uses a single algorithm, and global path planning only avoids static obstacles such as fixed equipment and building structures within the substation, failing to include high-risk fire spread areas in the path planning constraints. Local obstacle avoidance only responds to real-time obstacles and has no coordinated relationship with the global navigation path.

[0003] Existing technologies cannot match the three-dimensional spatial layout of substations. Insufficient fire perception dimensions make path planning unable to avoid dangerous areas where flames are dynamically spreading. A single algorithm cannot simultaneously meet the needs of global path planning and obstacle avoidance in dynamic environments. The local obstacle avoidance process cannot integrate global navigation paths, real-time local obstacle information, and robot motion status, making it difficult to generate speed control commands adapted to dynamic environments. As a result, the robot's movement path and motion control in fire scenarios have significant limitations.

[0004] This invention aims to achieve the fusion detection of multi-dimensional fire information, construct a full-station fire situation map including the three-dimensional coordinates of the fire source, fire intensity, and spread direction, combine the three-dimensional point cloud map of the substation with the real-time positioning of the robot, use an improved potential field method to complete global path planning, and achieve the coordination of local obstacle avoidance and global path through the dynamic window method, forming an autonomous path planning and motion control mode adapted to the dynamic fire environment. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes a substation robot autonomous path planning and fire extinguishing method based on fire source localization, comprising: The real-time positioning information of the substation robot and the three-dimensional point cloud map of the substation are obtained, and the real-time positioning information is mapped onto the three-dimensional point cloud map to obtain the real-time position of the substation robot in the substation coordinate system. Receive a fire situation map of the entire station generated by a multi-sensor fusion detection device. The fire situation map of the entire station includes the three-dimensional coordinates of at least one fire source, the fire intensity level, and the direction of flame spread. Based on the real-time location of the substation robot, the three-dimensional point cloud map, and the three-dimensional coordinates of the fire source in the overall fire situation map, a global navigation path from the real-time location to the target fire source is calculated using a global path planning algorithm based on the improved potential field method. The global navigation path bypasses obstacles in the three-dimensional point cloud map and avoids high-risk spread areas in the overall fire situation map. Based on the global navigation path, the substation robot is controlled to move along the global navigation path, and local obstacle information is generated in real time using the onboard environmental perception sensors. A local obstacle avoidance algorithm based on the dynamic window method is adopted, which integrates the guidance of the global navigation path, the local obstacle information and the current motion state of the substation robot to generate real-time speed control commands, enabling the substation robot to travel along the global navigation path in a dynamic environment.

[0006] Furthermore, the calculation of a global navigation path from the real-time location to the target fire source using a global path planning algorithm based on the improved potential field method includes: In the three-dimensional point cloud map, the three-dimensional coordinates of the target fire source are set as the target point of the gravitational field to generate a global gravitational field; In the three-dimensional point cloud map, the positions of all static obstacles are set as the core of the repulsive field to generate a global repulsive field, and the static obstacles are identified from the three-dimensional point cloud map. In the fire situation map of the entire station, the direction of flame spread and the nearby fire source points are extracted. A repulsive gradient field is generated along the direction of flame spread, and an additional repulsive field is generated for the nearby fire source points. The repulsive gradient field and the additional repulsive field together constitute the fire repulsive field. The global gravitational field, the global repulsive field, and the fire repulsive field are vector-superimposed to generate a comprehensive potential field. In the comprehensive potential field, starting from the real-time position of the substation robot, an iterative search is performed along the gradient descent direction of the potential field until the three-dimensional coordinates of the target fire source are reached, forming a series of path points; The global navigation path is generated by performing smoothing optimization on the series of path points.

[0007] Furthermore, in the fire situation map of the entire station, extracting the flame spread direction and adjacent fire sources, generating a repulsive gradient field along the flame spread direction, and generating an additional repulsive field for adjacent fire sources includes: The flame spread direction vector and fire intensity level are analyzed from the overall fire situation map. A fan-shaped virtual obstacle region is constructed in the opposite direction of the flame spread direction vector. The magnitude of the repulsive force of this fan-shaped region is proportional to the fire intensity level. Calculate the distance between the target fire source and the three-dimensional coordinates of other fire sources in the overall fire situation map, and mark other fire sources whose distance is less than the safe distance threshold as the neighboring fire source points; A repulsive field is generated for each of the adjacent fire sources, and the radius of action of the repulsive field is related to the fire intensity level of the corresponding fire source. The repulsive field corresponding to the fan-shaped virtual obstacle area is fused with the repulsive fields corresponding to all the adjacent fire sources to obtain the fire repulsive field.

[0008] Furthermore, the local obstacle avoidance algorithm based on the dynamic window method generates real-time speed control commands, including: Based on the dynamic constraints of the substation robot, a series of candidate velocity pairs are generated in the velocity space, each candidate velocity pair containing a linear velocity value and an angular velocity value; For each candidate speed pair, predict the trajectory that the substation robot will generate when it moves according to that candidate speed pair in a future time interval. Each of the predicted trajectories is evaluated, and the evaluation criteria include: the proximity of the predicted trajectory to the global navigation path, the distance of the predicted trajectory's endpoint to the nearest local obstacle, and the substation robot's moving speed. Using the local obstacle information acquired in real time by the environmental perception sensor, it is determined whether the predicted trajectory will collide with the obstacle, and all candidate velocity pairs that may cause a collision are eliminated from the predicted trajectory. Among the remaining predicted trajectories, the candidate velocity pair corresponding to the predicted trajectory with the highest evaluation score is selected as the final adopted linear velocity and angular velocity. The final adopted linear velocity and angular velocity are converted into speed control commands and sent to the underlying motion controller of the substation robot.

[0009] Furthermore, the method also includes: When the substation robot moves to a distance less than the predetermined fire extinguishing distance from the target fire source, the fire source precise alignment process is triggered, and the position of the substation robot is adjusted so that the fire extinguishing device it carries is aligned with the target fire source. The fire extinguishing device is activated to extinguish the target fire source, and during the fire extinguishing operation, the updated fire situation map of the entire station is continuously received to determine whether the fire intensity level of the target fire source has dropped below the safety threshold. The process for precisely aligning the ignition source includes: The substation robot uses a visible light camera and an infrared thermal imager to image the environment in front of it, obtaining visible light images and infrared thermal images. The visible light image and the infrared thermal image are fused to enhance the features of the target fire source in the image. Using a feature-matching-based fire source identification algorithm, the image pixel coordinates of the target fire source are accurately identified in the fused image; Based on the camera imaging model and calibration parameters, the image pixel coordinates of the target fire source are transformed into three-dimensional coordinates in the substation robot's own coordinate system; Calculate the deviation angle between the aiming axis of the fire extinguishing device of the substation robot in its current pose and the direction vector required to point to the three-dimensional coordinates of the target fire source in its own coordinate system. Generate a pose adjustment command to control the substation robot to rotate until the deviation angle approaches zero, thus completing the precise alignment of the fire source.

[0010] Furthermore, the fusion processing of the visible light image and the infrared thermal image includes: Gaussian pyramid decomposition was performed on the visible light image and the infrared thermal image respectively to obtain their respective multi-scale image sequences; On the same scale layer, extract the texture features of the visible light image layer and the temperature features of the infrared thermal image layer; The texture features and temperature features are weighted and fused to generate a fused feature map for this scale layer, wherein the weights are dynamically adjusted according to the feature saliency of the fire source area; The highest-scale fused feature map is reconstructed using a Laplacian pyramid, and then fused upwards level by level to finally obtain a fused image containing rich texture information and significant heat source information.

[0011] Furthermore, activating the fire extinguishing device to perform fire extinguishing operations on the target fire source includes: Based on the fire intensity level of the target fire source in the fire situation map of the whole station, the corresponding spray parameters are matched from the preset fire extinguishing strategy library. The spray parameters include the fire extinguishing agent spray flow rate, spray sweeping mode and spray duration. The gimbal of the fire extinguishing device is controlled to move according to the spray sweeping mode so that the coverage of the fire extinguishing agent can cover the target fire source area. During the spraying duration, the extinguishing agent valve is opened at the extinguishing agent spray flow rate to perform the spraying operation; During the spraying operation, the temperature changes of the target fire source area are continuously monitored using an infrared thermal imager.

[0012] Furthermore, the continuous receipt of updated station-wide fire situation maps during firefighting operations to determine whether the fire intensity level of the target fire source has decreased below a safe threshold includes: After the spraying duration ends, acquire the latest fire situation map of the entire station generated by the multi-sensor fusion detection device; From the latest generated fire situation map of the entire site, check whether the fire intensity level corresponding to the target fire source has been updated; If the fire intensity level of the target fire source has dropped below the predefined safety threshold, it is determined that the target fire source has been extinguished and a fire extinguishing success signal is generated. If the fire intensity level of the target fire source does not drop below the safety threshold, the global navigation path will be recalculated or the spray parameters will be adjusted based on the latest three-dimensional coordinates of the fire source and the fire intensity level, and a new round of fire extinguishing operations will be initiated.

[0013] Furthermore, the method also includes a post-fire extinguishing autonomous inspection and reignition monitoring process: After determining that the current target fire source has been extinguished, check whether there are any other untreated fire sources in the overall fire situation map of the station. If there are other untreated fire sources, the nearest untreated fire source is set as the new target fire source, and the process returns to the step of calculating the global navigation path based on the real-time position of the substation robot. If there are no other untreated fire sources, the substation robot will be controlled to move and patrol within the substation according to the predetermined patrol route. During the patrol, the data from the environmental perception sensor and the multi-sensor fusion detection device are continuously analyzed to monitor whether new fires have occurred or whether extinguished fires have shown signs of reignition.

[0014] Furthermore, the substation robot is controlled to conduct mobile inspections within the substation according to a predetermined inspection path, including: Load the pre-generated routine substation inspection path point sequence from the substation control center or the robot's local storage; Each point in the inspection path point sequence is used as a temporary navigation target, and the substation robot is controlled to move to that point using local path planning and obstacle avoidance algorithms; Upon arriving at each inspection route point, the substation robot is controlled to make a brief stop and rotate its onboard visible light camera and infrared thermal imager to perform a 360-degree scan of the surrounding environment. The environmental data obtained by scanning is compared and analyzed with preset normal state baseline data to detect abnormal temperature, smoke or open flame characteristics.

[0015] Compared with the prior art, the beneficial effects of the present invention are: The multi-sensor fusion detection device can generate a comprehensive fire situation map of the entire substation, integrating the three-dimensional coordinates of the fire source, the fire intensity level, and the direction of flame spread. This fire situation map can be directly used as input data for substation robot path planning. The fire information is expanded from a simple existence determination or two-dimensional planar positioning to a comprehensive representation of three-dimensional spatial coordinates and dynamic fire attributes. The fire data used for path planning can fully match the spatial distribution and development trend of the actual fire in the substation. The path planning process can directly rely on complete fire situation information for calculation, avoiding the lack of planning reference data due to insufficient dimensions of fire information, and ensuring that the robot's movement path is consistent with the actual hazard distribution at the fire site.

[0016] The improved potential field method-based global path planning algorithm can simultaneously combine the real-time position of the substation robot, the 3D point cloud map of the substation, and the 3D coordinates of the fire source to complete path calculation. The planned global navigation path can simultaneously bypass static obstacles in the 3D point cloud map and avoid high-risk fire spread areas in the overall fire situation map. The dynamic window method-based local obstacle avoidance algorithm can fuse the global navigation path guidance, real-time collected local obstacle information, and the current motion state of the substation robot to directly generate corresponding real-time speed control commands. The local obstacle avoidance behavior and the global navigation path form a linkage constraint. The generation of speed control commands can take into account path guidance, environmental obstacles, and the robot's own motion state, enabling the robot to continuously and stably move along the predetermined global navigation path in a dynamically changing environment. The motion control process and the global planning logic maintain a high degree of coordination, adapting to the movement needs of the complex spatial structure of the substation and dynamic fire scenarios. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the steps of a substation robot's autonomous path planning and fire extinguishing method based on fire source location, as described in this invention.

[0018] Figure 2 A flowchart for generating a repulsive field in a fire.

[0019] Figure 3 A flowchart generated for speed control commands. Detailed Implementation

[0020] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] See Figure 1This invention provides a substation robot autonomous path planning and fire suppression method based on fire source localization. The method includes: mapping the real-time positioning information of the substation robot onto the 3D point cloud map of the substation to obtain the real-time position of the substation robot in the substation coordinate system. The method receives a station-wide fire situation map periodically generated by a multi-sensor fusion detection device deployed within the substation. The station-wide fire situation map includes the 3D coordinates of at least one fire source, the fire intensity level, and the direction of flame spread. Based on the real-time position of the substation robot, the 3D point cloud map, and the 3D coordinates of the fire source in the station-wide fire situation map, the method uses a global path planning algorithm based on an improved potential field method to calculate a global navigation path from the real-time position to the target fire source. This global navigation path bypasses obstacles in the 3D point cloud map and avoids high-risk spread areas in the station-wide fire situation map. Based on the global navigation path, the method controls the substation robot to move along the global navigation path and uses environmental perception sensors on the robot to generate local obstacle information in real time. This method employs a local obstacle avoidance algorithm based on the dynamic window method, which integrates the guidance of the global navigation path, the local obstacle information, and the current motion state of the substation robot to generate real-time speed control commands, enabling the substation robot to travel along the global navigation path in a dynamic environment.

[0022] In one embodiment of the present invention, in the three-dimensional point cloud map, the three-dimensional coordinates of the target fire source are set as the target point of the gravitational field, generating a global gravitational field. In the three-dimensional point cloud map, the positions of all static obstacles are set as the core of the repulsive field, generating a global repulsive field, and the static obstacles are identified from the three-dimensional point cloud map. See also... Figure 2In the overall fire situation map, the flame spread direction and adjacent fire sources are extracted. A repulsive gradient field is generated along the flame spread direction, and an additional repulsive field is generated for the adjacent fire sources. The repulsive gradient field and the additional repulsive field together constitute the fire repulsive field. The flame spread direction vector and fire intensity level are analyzed from the overall fire situation map. A fan-shaped virtual obstacle area is constructed along the opposite direction of the flame spread direction vector. The magnitude of the repulsive force in this fan-shaped area is proportional to the fire intensity level. The distance between the target fire source and the three-dimensional coordinates of other fire sources in the overall fire situation map is calculated. Other fire sources with a distance less than the safe distance threshold are marked as adjacent fire sources. A repulsive field is generated for each adjacent fire source. The radius of action of this repulsive field is related to the fire intensity level of the corresponding fire source. The repulsive field corresponding to the fan-shaped virtual obstacle area is fused with the repulsive fields corresponding to all adjacent fire sources to obtain the fire repulsive field. The global gravitational field, the global repulsive field, and the fire repulsive field are vector-superimposed to generate a comprehensive potential field. Within this comprehensive potential field, starting from the real-time position of the substation robot, an iterative search is performed along the gradient descent direction until the three-dimensional coordinates of the target fire source are reached, forming a series of path points. These path points are then smoothed and optimized to generate the global navigation path.

[0023] In practical implementation, the global path planning algorithm of the improved potential field method calculates the global navigation path based on the three-dimensional point cloud map of the substation and the overall fire situation map of the substation. In a specific implementation, taking a substation scenario containing multiple equipment areas and spreading fire as an example, the locations of static obstacles such as transformers and distribution cabinets are marked in the three-dimensional point cloud map of the substation. The overall fire situation map of the substation provides the three-dimensional coordinates of the target fire source, the fire intensity level, and the direction vector of flame spread, and marks the adjacent fire source points. Data comparison shows that when the traditional potential field method only considers the repulsive field of static obstacles, the planned path may pass through high-risk areas of flame spread. However, the improved potential field method, by introducing the fire repulsive field, enables the path to effectively avoid these areas and improve safety. In practice, when generating a global gravitational field, the three-dimensional coordinates of the target fire source are set as the target point of the gravitational field. The gravitational field function makes the substation robot attractive to the target fire source throughout the map. When generating a global repulsive field, the positions of all static obstacles are identified from the three-dimensional point cloud map, and each static obstacle position is set as the core of the repulsive field. A repulsive force field covering the area around the obstacle is generated to prevent the path from colliding with static objects. In practice, generating a fire repulsion field involves analyzing the flame spread direction vector and fire intensity level from the overall fire situation map. A fan-shaped virtual obstacle area is constructed in the opposite direction of the flame spread direction vector. The magnitude of the repulsion force in this fan-shaped area is proportional to the fire intensity level; the higher the fire intensity level, the greater the repulsion force. Simultaneously, the distance between the target fire source and the three-dimensional coordinates of other fire sources in the overall fire situation map is calculated. Other fire sources with a distance less than the safe distance threshold are marked as neighboring fire sources. A repulsion field is generated for each neighboring fire source. The radius of action of this repulsion field is related to the fire intensity level of the corresponding fire source; the higher the fire intensity level, the larger the radius of action. The repulsion field corresponding to the fan-shaped virtual obstacle area is merged with the repulsion fields corresponding to all neighboring fire sources to obtain the fire repulsion field.

[0024] In some embodiments, a global gravitational field, a global repulsive field, and a fire repulsive field are vector-superimposed to generate a combined potential field, the combined potential field function of which is expressed as: in: Indicates position The overall potential field value at that location, Indicates position The global gravitational field value at that location. Indicates position The global repulsive field value at that location, Indicates position The repulsive field value of the fire at the location, These are three-dimensional coordinate points in the substation coordinate system. In practice, starting from the real-time position of the robot in the substation, it iteratively searches along the gradient descent direction of the potential field in the integrated potential field. Each iteration calculates the potential field gradient direction at the current position and moves the robot by a step size until it reaches the three-dimensional coordinates of the target fire source, forming a series of path points. These path points are then smoothed and optimized using a curve fitting algorithm to reduce path turns and generate a global navigation path. Optionally, in the data comparison of the example scenario, without the introduction of a fire repulsion field, path points may be distributed in areas with high fire spread, causing the robot to approach danger. However, with the introduction of a fire repulsion field, the path points automatically avoid the fan-shaped virtual obstacle area and adjacent fire source areas during the gradient descent process, resulting in a safer path point sequence. In practice, the safety distance threshold is preset based on the substation equipment layout and historical fire data, for example, set to 5 meters, to ensure that the robot stays away from nearby fire sources. The fire intensity level is divided into three levels: low, medium and high, corresponding to different repulsive force magnitude coefficients and action radius coefficients. In practice, the smoothing optimization process adopts the B-spline curve method to interpolate and smooth the original path points to generate a continuous and differentiable global navigation path.

[0025] In one embodiment of the present invention, see [reference] Figure 3 Based on the dynamic constraints of the substation robot, a series of candidate velocity pairs are generated in the velocity space. Each candidate velocity pair contains a linear velocity value and an angular velocity value. For each candidate velocity pair, a predicted trajectory is predicted, indicating the substation robot's movement along that candidate velocity pair within a future time interval. Each predicted trajectory is evaluated using criteria including: the proximity of the predicted trajectory to the global navigation path, the distance of the predicted trajectory's endpoint to the nearest local obstacle, and the substation robot's moving speed. Using local obstacle information acquired in real-time by the environmental perception sensors, it is determined whether the predicted trajectory will collide with an obstacle, and all predicted trajectories corresponding to candidate velocity pairs that may lead to collisions are eliminated. Among the remaining predicted trajectories, the candidate velocity pair corresponding to the predicted trajectory with the highest evaluation score is selected as the final adopted linear velocity and angular velocity. The final adopted linear velocity and angular velocity are converted into speed control commands and sent to the substation robot's underlying motion controller.

[0026] In practical implementation, taking a substation passage scenario containing mobile maintenance vehicles and temporary storage as an example, the substation robot is moving along a global navigation path. Environmental perception sensors detect the moving maintenance vehicle ahead and the temporary equipment box on the side of the passage in real time. Data comparison shows that without the dynamic window method, the robot may stall or collide with dynamic obstacles due to fixed speed commands. However, with the dynamic window method, the robot can generate speed control commands that adapt to the dynamic environment in real time, safely passing through complex areas. In the specific implementation, a series of candidate velocity pairs are generated based on the dynamic constraints of the substation robot, including maximum linear velocity, maximum angular velocity, maximum linear acceleration, and maximum angular acceleration. In the velocity space composed of linear velocity and angular velocity, starting from the robot's current motion state, sampling is performed within the range of the maximum acceleration and deceleration capability to generate a series of candidate velocity pairs. Each candidate velocity pair contains a linear velocity value and an angular velocity value. For each candidate velocity pair, the predicted trajectory of the substation robot moving according to the candidate velocity pair in a future time interval is predicted. The predicted trajectory is calculated by integrating the kinematic model, representing the robot's position and attitude sequence in several simulated time steps in the future.

[0027] In some embodiments, each predicted trajectory is evaluated using three criteria: first, the proximity of the predicted trajectory to the global navigation path, measured by calculating the sum of the shortest distances from sampling points on the predicted trajectory to the global navigation path; second, the distance from the endpoint of the predicted trajectory to the nearest local obstacle, calculated from real-time obstacle information acquired by environmental perception sensors; and third, the substation robot's moving speed, to encourage the robot to maintain a reasonable forward speed. In a specific implementation, the evaluation process calculates an evaluation score for each predicted trajectory, with the scoring function designed as follows: in: Represents linear velocity and angular velocity The candidate velocities are evaluated against the corresponding predicted trajectories. This measures the degree of similarity between the predicted trajectory and the global navigation path. This represents the distance metric between the predicted trajectory endpoint and the nearest obstacle. Indicates the magnitude of speed. , , These are the weighting coefficients for each item.

[0028] In practical implementation, local obstacle information acquired in real time by environmental perception sensors is used to determine whether the predicted trajectory will collide with obstacles. Collision detection is performed along the predicted trajectory using the robot's chassis outline. If the robot's outline overlaps with an occupied area in the obstacle map at any simulated time step, the predicted trajectory is determined to cause a collision, and all predicted trajectories corresponding to candidate velocity pairs that may lead to collision are eliminated. Among the remaining predicted trajectories, the candidate velocity pair corresponding to the predicted trajectory with the highest evaluation score is selected as the final adopted linear and angular velocities. The final adopted linear and angular velocities are converted into speed control commands, typically a combination of linear and angular velocity values, and sent to the substation robot's underlying motion controller to drive the robot's movement. Optionally, in the data comparison of the example scenario, when the mobile maintenance vehicle suddenly cuts into the global navigation path from the side, the dynamic window method can quickly eliminate all candidate velocity pairs that could lead to a collision with the vehicle and select the one with the highest evaluation score from the remaining safe candidate velocity pairs, causing the robot to decelerate or make a slight turn to avoid the collision. In some embodiments, the length of the future prediction time interval is set based on the robot's maximum speed and environmental complexity, for example, set to 1 to 3 seconds, to balance computational efficiency and foresight. It can be understood that the weighting coefficients in the evaluation scoring function... , , The adjustment needs to be made according to the task focus; when focusing on path tracing, increase the [efforts / requirements]. When focusing on safe obstacle avoidance, increase It focuses on increasing speed during rapid movement. .

[0029] In practical implementation, data comparison also shows that when encountering temporary obstacles in narrow passages, the predicted trajectory evaluation generated by the dynamic window method prioritizes trajectories with endpoints farther from obstacles, thus allowing the robot to pass through with a larger gap between itself and the temporary obstacles. Optionally, the collision detection process considers not only static obstacle maps but also integrates dynamic obstacle prediction trajectories detected in real time by environmental perception sensors to achieve avoidance of moving targets. In practical implementation, the sampling resolution of candidate velocity pairs affects planning quality and computational load. Linear velocity and angular velocity are typically uniformly or non-uniformly discretized within their respective allowable ranges to generate candidate velocity pair sets. It can be understood that the local obstacle avoidance algorithm is repeatedly executed in each control cycle, recalculating the velocity control command using the latest robot pose, local obstacle information, and global navigation path to adapt to continuous environmental changes.

[0030] In one embodiment of the present invention, when the substation robot moves to a distance less than a predetermined extinguishing distance from the target fire source, a precise fire source alignment process is triggered. The substation robot, equipped with a visible light camera and an infrared thermal imager, images the surrounding environment, obtaining visible light and infrared thermal images. The visible light and infrared thermal images are then fused, with Gaussian pyramid decomposition performed on both images to obtain their respective multi-scale image sequences. At the same scale level, texture features from the visible light image layer and temperature features from the infrared thermal image layer are extracted. The texture and temperature features are then weighted and fused to generate a fused feature map for that scale level, where the weights are dynamically adjusted based on the saliency of the fire source region. Laplacian pyramid reconstruction is performed on the highest-scale fused feature map, and fusion is performed level by level upwards, ultimately resulting in a fused image containing rich texture information and significant heat source information. Using a feature-matching-based fire source identification algorithm, the image pixel coordinates of the target fire source are accurately identified in the fused image. Based on the camera imaging model and calibration parameters, the image pixel coordinates of the target fire source are transformed into three-dimensional coordinates in the substation robot's own coordinate system. The deviation angle between the aiming axis of the fire extinguishing device and the required direction vector pointing to the three-dimensional coordinates of the target fire source in the substation robot's current pose is calculated. A pose adjustment command is generated to control the substation robot's rotation until the deviation angle approaches zero, thus achieving precise alignment of the fire source.

[0031] In practical implementation, taking the substation robot reaching the predetermined fire extinguishing distance from the target fire source as an example, the target fire source is located behind a power distribution cabinet, with visual obstruction and complex background. The robot triggers the precise fire source alignment process, using the onboard visible light camera and infrared thermal imager to image the environment in front, obtaining a visible light image containing the power distribution cabinet and background and an infrared thermal image highlighting the high-temperature area. Data comparison shows that when using the visible light image alone, the flame features may not be significant due to smoke obstruction. When using the infrared thermal image alone, the lack of environmental texture information leads to inaccurate fire source edge positioning. The image after fusion processing can simultaneously retain texture and significant heat source information, improving recognition accuracy.

[0032] In practice, visible light images and infrared thermal images are fused. Gaussian pyramid decomposition is performed on both images to obtain their respective multi-scale image sequences. The number of decomposition layers is set according to the original image resolution and feature scale. At the same scale layer, texture features from the visible light image layer and temperature features from the infrared thermal image layer are extracted. Texture features are calculated using a gradient operator, and temperature features are mapped using pixel grayscale values. The texture and temperature features are then weighted and fused to generate a fused feature map for that scale layer. The fusion process follows the formula: in: Indicates the first fused feature maps at scale layers Indicates the first Visible light image features at scale layers Indicates the first Infrared thermal image features of scale layers and These represent the dynamic weights assigned to texture features and temperature features, respectively. The adjustment of these dynamic weights is determined by the saliency of the fire source region's features; if the fire source region has more saliency in temperature features in the infrared thermal image, then... Increased size makes contour and texture features more prominent in visible light images. The scale is increased. Laplacian pyramid reconstruction is performed on the highest-scale fused feature map, and detailed information from lower-scale layers is fused upwards step by step to obtain a fused image containing rich texture information and significant heat source information. In some embodiments, feature saliency is determined by comparing the feature contrast of the same region in different images; the dynamic adjustment strategy for the weights is shown in Table 1. Table 1: Feature Significance Weight Adjustment Strategy A feature-matching-based fire source identification algorithm is used to accurately identify the image pixel coordinates of the target fire source in the fused image. The algorithm searches for regions with high brightness, high temperature, and specific texture patterns in the fused image and matches them with the prior location provided by the overall fire situation map to determine the precise pixel coordinates of the target fire source. Based on the camera imaging model and calibration parameters, the image pixel coordinates of the target fire source are transformed into three-dimensional coordinates in the substation robot's own coordinate system. The deviation angle between the aiming axis of the fire extinguishing device and the direction vector pointing to the three-dimensional coordinates of the target fire source in the current pose of the substation robot is calculated, and a pose adjustment command is generated. The pose adjustment command includes the required rotation angle, and the substation robot is controlled to rotate until the deviation angle approaches zero, thus completing the precise alignment of the fire source.

[0033] Optionally, in the data comparison of the example scenario, without image fusion, the fire source feature identification in the visible light image may be incorrect due to the reflective interference of the metal casing of the power distribution cabinet, resulting in a large deviation in the calculated image pixel coordinates. However, after multi-scale feature weighted fusion, the high-temperature feature compensates for the texture interference, and the identified image pixel coordinates are closer to the actual fire source center. In the specific implementation, the number of Gaussian pyramid decomposition layers is set to 4. Starting from the original resolution image, each layer undergoes Gaussian smoothing and downsampling, while the Laplacian pyramid reconstruction restores details step by step through upsampling and interpolation. It can be understood that the calculation of the deviation angle involves a chain transformation between the robot's own coordinate system, the camera coordinate system, and the image pixel coordinate system, relying on the pre-calibrated camera intrinsic and extrinsic parameter matrices.

[0034] In specific implementations, the temperature feature extraction of the infrared thermal image layer directly uses the image grayscale values, because the pixel grayscale values ​​of the infrared thermal imager have a linear mapping relationship with temperature. The texture features of the visible light image layer are obtained by calculating the gradient magnitude of the Sobel operator. In some embodiments, the fire source identification algorithm based on feature matching first segments candidate high-temperature regions in the fused image, then extracts the shape and grayscale distribution features of the regions, matches them with the fire source template features, and the region with the highest score is identified as the target fire source. Optionally, the predetermined extinguishing distance is set according to the type and spray range of the fire extinguishing device, for example, 3 meters for a dry powder fire extinguishing device. It can be understood that the precise fire source alignment process is a preliminary step to trigger the fire extinguishing operation, ensuring that the extinguishing agent can accurately cover the target fire source area. In specific implementations, the pose adjustment command is sent to the rotation mechanism of the robot chassis, controlling the robot to rotate around its center until the spray axis of the fire extinguishing device is aligned with the calculated three-dimensional coordinate direction of the target fire source's own coordinate system.

[0035] In one embodiment of the present invention, the fire extinguishing device is activated to extinguish the target fire source. Based on the fire intensity level of the target fire source in the overall fire situation map, corresponding spray parameters are matched from a pre-set fire extinguishing strategy library. These spray parameters include the extinguishing agent spray flow rate, spray sweep mode, and spray duration. The pan-tilt unit of the fire extinguishing device is controlled to move according to the spray sweep mode, ensuring the extinguishing agent coverage area encompasses the target fire source area. During the spray duration, the extinguishing agent valve is opened at the extinguishing agent spray flow rate to execute the spray operation. During the spray operation, the temperature change of the target fire source area is continuously monitored using an infrared thermal imager, and the updated overall fire situation map is continuously received. After the spray duration ends, the latest overall fire situation map generated by the multi-sensor fusion detection device is acquired. From the latest generated overall fire situation map, it is checked whether the fire intensity level corresponding to the target fire source has been updated. If the fire intensity level of the target fire source has dropped below a predefined safety threshold, the target fire source is determined to have been extinguished, and a fire extinguishing success signal is generated. If the fire intensity level of the target fire source does not drop below the safety threshold, the global navigation path will be recalculated or the spray parameters will be adjusted based on the latest three-dimensional coordinates of the fire source and the fire intensity level, and a new round of fire extinguishing operations will be initiated.

[0036] In practical implementation, taking a fire in a transformer bushing at a substation as an example, the overall fire situation map indicates the fire intensity level of the target fire source is "medium." The robot has accurately aligned with the fire source and activated the fire extinguishing device. Data comparison shows that using fixed parameters for fire extinguishing without matching the spray parameters according to the fire intensity level may lead to waste of extinguishing agent or insufficient extinguishing intensity. Matching parameters according to the level can optimize the efficiency of fire extinguishing operations. In practical implementation, to activate the fire extinguishing device and perform fire extinguishing operations on the target fire source, firstly, based on the fire intensity level of the target fire source in the overall fire situation map, the corresponding spray parameters are matched from the pre-set fire extinguishing strategy library. The spray parameters include the extinguishing agent spray flow rate, spray sweeping mode, and spray duration. The fire extinguishing strategy library is indexed by the fire intensity level and stores parameter combinations determined in advance through experiments and simulations. For specific matching relationships, please refer to Table 2. Table 2: Matching Table of Fire Intensity Level and Fire Extinguishing Spray Parameters The pan-tilt unit of the fire extinguishing device moves according to the matched spray sweeping mode. For example, in the "fan-shaped sweeping" mode, the pan-tilt unit reciprocates horizontally at a specific angular velocity, ensuring that the extinguishing agent coverage area encompasses the target fire source area. During the spraying duration, the extinguishing agent valve is opened at the matched extinguishing agent spray flow rate to execute the spraying operation. With valve opening The relationship is determined by the calibration curve and follows ,in Represents the extinguishing agent spray flow rate. Representative describes valve opening degree With traffic A function relating the two. During the spraying operation, the temperature changes in the target fire source area are continuously monitored using an infrared thermal imager mounted on the robot to obtain real-time temperature distribution maps.

[0037] In some embodiments, during firefighting operations, a station-wide fire situation map, periodically updated by a multi-sensor fusion detection device, is continuously received to determine whether the fire intensity level of the target fire source has dropped below a safe threshold. After the spraying duration ends, the latest station-wide fire situation map generated by the multi-sensor fusion detection device is acquired. From the latest generated station-wide fire situation map, it is checked whether the fire intensity level corresponding to the target fire source has been updated. If the fire intensity level of the target fire source has dropped below a predefined safe threshold, it is determined that the target fire source has been extinguished, and a fire extinguishing success signal is generated. It can be understood that the safe threshold is a temperature or energy threshold below a "low" fire intensity level, used to confirm that the open flame has been extinguished and the residual temperature will not immediately reignite.

[0038] In practice, data comparison shows that without continuously updated global fire situation maps, relying solely on local infrared monitoring may lead to misjudgments of fire extinguishing due to partial smoke obscuring the view. Combining updated global situation maps allows for more accurate assessments. If the fire intensity level of the target fire source has not decreased below the safety threshold, the global navigation path is recalculated or the spray parameters are adjusted based on the fire source's 3D coordinates and fire intensity level provided in the latest global fire situation map, initiating a new round of firefighting operations. Optionally, adjusting spray parameters includes matching higher-level spray parameters from the firefighting strategy library based on the still-burning fire intensity level, such as changing from "medium" to "high" level parameters. Recalculating the global navigation path occurs when the target fire source location shifts due to the collapse or spread of burning materials; the robot's movement path is replanned based on the latest 3D coordinates of the fire source. It can be understood that obtaining the latest global fire situation map after the spraying duration ends is to evaluate the effectiveness of the previous round of firefighting operations and provide a basis for subsequent decisions. In some embodiments, the temperature change of the target fire source area is continuously monitored and compared with the temperature baseline before the spraying operation begins to form a thermal trend map of the fire extinguishing process. Optionally, the determination of the extinguishing agent spray flow rate and spraying duration needs to consider the extinguishing agent type, target fire source characteristics, and environmental factors, and multiple sets of parameters need to be preset in the fire extinguishing strategy library. In specific implementation, after a successful fire extinguishing signal is generated, the fire extinguishing device stops spraying, and the robot control system enters the subsequent process. If unsuccessful, parameter adjustment or path replanning is triggered, and the fire extinguishing operation is repeated until successful.

[0039] In one embodiment of the present invention, after determining that the current target fire source has been extinguished, an autonomous inspection and reignition monitoring process is executed. The system queries the overall fire situation map to check for other untreated fire sources. If other untreated fire sources exist, the nearest untreated fire source is set as the new target fire source, and the process returns to the step of calculating the global navigation path based on the real-time position of the substation robot. If no other untreated fire sources exist, the substation robot is controlled to move and inspect within the substation according to a predetermined inspection path. A pre-generated sequence of regular substation inspection path points is loaded from the substation control center or the robot's local storage. Each point in the inspection path point sequence is used as a temporary navigation target, and the substation robot is controlled to move to that point using local path planning and obstacle avoidance algorithms. Upon reaching each inspection path point, the substation robot pauses briefly and rotates its onboard visible light camera and infrared thermal imager to perform a 360-degree scan of the surrounding environment. The environmental data obtained from the scan is compared and analyzed with preset normal state baseline data to detect abnormal temperatures, smoke, or open flame characteristics. During the patrol, the data from the environmental perception sensor and the multi-sensor fusion detection device are continuously analyzed to monitor whether new fires have occurred or whether extinguished fires have shown signs of reignition.

[0040] In practical implementation, taking the case where a target fire source in the main transformer area of ​​a substation has been determined to be extinguished as an example, the substation robot, located near the main transformer, checks the current fire situation map of the entire substation and discovers another untreated fire source in the capacitor area. Data comparison shows that without an autonomous patrol process, the robot may remain in place or return to a standby point, unable to proactively detect and handle new fires or signs of reignition. However, after initiating the autonomous patrol process, the robot can systematically check the treated areas and respond to new fires. After determining that the current target fire source has been extinguished, the autonomous patrol and reignition monitoring process after fire extinguishing is executed. First, it checks the fire situation map of the entire substation to see if there are other untreated fire sources. If there are other untreated fire sources, the untreated fire source closest to the real-time location of the substation robot is set as the new target fire source, and the process returns to the step of calculating the global navigation path based on the real-time location of the substation robot. If no other unaddressed fire sources are found, the substation robot will move and inspect the substation according to a predetermined inspection path. A pre-generated sequence of routine substation inspection path points will be loaded from the substation control center or the robot's local storage. This sequence of inspection path points is represented as follows: in: Represents the sequence of patrol route points. Represents the first in the sequence The three-dimensional coordinates of each patrol route point This represents the total number of path points. The path point sequence will be inspected sequentially. Each point in As a temporary navigation target, the substation robot is controlled to move to the designated point using local path planning and obstacle avoidance algorithms. Upon arrival at each inspection path point, the robot pauses briefly and rotates its onboard visible light camera and infrared thermal imager to perform a 360-degree scan of the surrounding environment. The acquired environmental data is compared and analyzed with preset normal state baseline data to detect abnormal temperatures, smoke, or open flames. During the inspection, data from environmental perception sensors and multi-sensor fusion detection devices are continuously analyzed to monitor for new fires or signs of reignition of extinguished fires.

[0041] In some embodiments, the preset normal state baseline data includes background infrared temperature distribution images and visible light scene images at each path point under fire-free conditions. A comparative analysis is performed by calculating the difference between the real-time scan data and the baseline data. : in: This represents the mean square error between the real-time infrared temperature and the reference temperature field. This represents the degree of difference between a real-time visible light image and a reference image in a specific feature space. and This is the weighting coefficient. When the difference... When the set threshold is exceeded, an anomaly alarm is triggered. Optionally, in the data comparison of the example scenario, if only fixed-point monitoring is relied upon, a hidden reignition of a fire in the main transformer that has been extinguished may not be detected in time after the robot leaves. However, by moving along a predetermined patrol path and periodically revisiting key areas, the robot can detect temperature anomalies in advance through continuous scanning and comparative analysis. In specific implementation, the patrol path point sequence... The system is pre-planned and generated based on the substation equipment layout, fire risk level, and historical fault location information, covering key areas such as the main transformer, capacitor banks, distribution rooms, and cable trench entrances. It's understandable that setting the nearest untreated fire source as the new target is based on the principle of optimal distance, aiming to shorten the robot's overall movement time for a rapid response to a fire.

[0042] In practice, the switching of temporary navigation target points is automatic, when the robot arrives at the current patrol path point. After completing the environment scan, it will automatically set the next waypoint. The target is set as a temporary navigation target. In some embodiments, environmental perception sensors, including lidar and millimeter-wave radar, operate continuously during the patrol process to update local obstacle information to support local path planning and obstacle avoidance algorithms. Optionally, during panoramic rotational scanning, the rotational scanning speed and dwell time of the visible light camera and infrared thermal imager can be configured to ensure clear and complete circumferential environmental information is obtained. In specific implementations, if a new fire is detected during the patrol or if a previously extinguished fire shows signs of reignition, the routine patrol is immediately interrupted, the newly discovered fire source or reignition point is set as an emergency target fire source, and the global path planning and fire suppression process is triggered. It can be understood that the autonomous patrol and reignition monitoring process constitute a closed-loop safety monitoring cycle, continuing to perform preventive monitoring duties after completing the active fire suppression task.

[0043] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for autonomous path planning and fire extinguishing of a substation robot based on fire source localization, characterized in that, The method includes: The real-time positioning information of the substation robot and the three-dimensional point cloud map of the substation are obtained, and the real-time positioning information is mapped onto the three-dimensional point cloud map to obtain the real-time position of the substation robot in the substation coordinate system. Receive a fire situation map of the entire station generated by a multi-sensor fusion detection device. The fire situation map of the entire station includes the three-dimensional coordinates of at least one fire source, the fire intensity level, and the direction of flame spread. Based on the real-time location of the substation robot, the three-dimensional point cloud map, and the three-dimensional coordinates of the fire source in the overall fire situation map, a global navigation path from the real-time location to the target fire source is calculated using a global path planning algorithm based on the improved potential field method. The global navigation path bypasses obstacles in the three-dimensional point cloud map and avoids high-risk spread areas in the overall fire situation map. Based on the global navigation path, the substation robot is controlled to move along the global navigation path, and local obstacle information is generated in real time using the onboard environmental perception sensors. A local obstacle avoidance algorithm based on the dynamic window method is adopted, which integrates the guidance of the global navigation path, the local obstacle information and the current motion state of the substation robot to generate real-time speed control commands, enabling the substation robot to travel along the global navigation path in a dynamic environment.

2. The substation robot autonomous path planning and fire extinguishing method based on fire source localization according to claim 1, characterized in that, The global path planning algorithm based on the improved potential field method is used to calculate a global navigation path from the real-time location to the target fire source, including: In the three-dimensional point cloud map, the three-dimensional coordinates of the target fire source are set as the target point of the gravitational field to generate a global gravitational field; In the three-dimensional point cloud map, the positions of all static obstacles are set as the core of the repulsive field to generate a global repulsive field, and the static obstacles are identified from the three-dimensional point cloud map. In the fire situation map of the whole station, the direction of flame spread and the nearby fire source points are extracted. A repulsive gradient field is generated along the direction of flame spread and an additional repulsive field is generated for the nearby fire source points. The repulsive gradient field and the additional repulsive field together constitute the fire repulsive field. The global gravitational field, the global repulsive field, and the fire repulsive field are vector-superimposed to generate a comprehensive potential field. In the comprehensive potential field, starting from the real-time position of the substation robot, an iterative search is performed along the gradient descent direction of the potential field until the three-dimensional coordinates of the target fire source are reached, forming a series of path points; The global navigation path is generated by performing smoothing optimization on the series of path points.

3. The substation robot autonomous path planning and fire extinguishing method based on fire source localization according to claim 2, characterized in that, The process of extracting the flame spread direction and nearby fire sources from the overall fire situation map, generating a repulsive gradient field along the flame spread direction, and generating additional repulsive fields for nearby fire sources includes: The flame spread direction vector and fire intensity level are analyzed from the overall fire situation map. A fan-shaped virtual obstacle region is constructed in the opposite direction to the flame spread direction vector. The magnitude of the repulsive force of this fan-shaped region is proportional to the fire intensity level. Calculate the distance between the target fire source and the three-dimensional coordinates of other fire sources in the overall fire situation map, and mark other fire sources whose distance is less than the safe distance threshold as the neighboring fire source points; A repulsive field is generated for each of the adjacent fire sources, and the radius of action of the repulsive field is related to the fire intensity level of the corresponding fire source. The repulsive field corresponding to the fan-shaped virtual obstacle area is fused with the repulsive fields corresponding to all the adjacent fire sources to obtain the fire repulsive field.

4. The substation robot autonomous path planning and fire extinguishing method based on fire source localization according to claim 1, characterized in that, The local obstacle avoidance algorithm based on the dynamic window method generates real-time speed control commands, including: Based on the dynamic constraints of the substation robot, a series of candidate velocity pairs are generated in the velocity space, each candidate velocity pair containing a linear velocity value and an angular velocity value; For each candidate speed pair, predict the trajectory that the substation robot will generate when it moves according to that candidate speed pair in a future time interval. Each of the predicted trajectories is evaluated, and the evaluation criteria include: the proximity of the predicted trajectory to the global navigation path, the distance of the predicted trajectory's endpoint to the nearest local obstacle, and the substation robot's moving speed. Using the local obstacle information acquired in real time by the environmental perception sensor, it is determined whether the predicted trajectory will collide with the obstacle, and all candidate velocity pairs that may cause a collision are eliminated from the predicted trajectory. Among the remaining predicted trajectories, the candidate velocity pair corresponding to the predicted trajectory with the highest evaluation score is selected as the final adopted linear velocity and angular velocity. The final adopted linear velocity and angular velocity are converted into speed control commands and sent to the underlying motion controller of the substation robot.

5. The substation robot autonomous path planning and fire extinguishing method based on fire source localization according to claim 1, characterized in that, The method further includes: When the substation robot moves to a distance less than the predetermined fire extinguishing distance from the target fire source, the fire source precise alignment process is triggered, and the position of the substation robot is adjusted so that the fire extinguishing device it carries is aligned with the target fire source. The fire extinguishing device is activated to extinguish the target fire source, and during the fire extinguishing operation, the updated fire situation map of the entire station is continuously received to determine whether the fire intensity level of the target fire source has dropped below the safety threshold. The process for precisely aligning the ignition source includes: The substation robot uses a visible light camera and an infrared thermal imager to image the environment in front of it, obtaining visible light images and infrared thermal images. The visible light image and the infrared thermal image are fused to enhance the features of the target fire source in the image. Using a feature-matching-based fire source identification algorithm, the image pixel coordinates of the target fire source are accurately identified in the fused image; Based on the camera imaging model and calibration parameters, the image pixel coordinates of the target fire source are transformed into three-dimensional coordinates in the substation robot's own coordinate system; Calculate the deviation angle between the aiming axis of the fire extinguishing device of the substation robot in its current pose and the direction vector required to point to the three-dimensional coordinates of the target fire source in its own coordinate system. Generate a pose adjustment command to control the substation robot to rotate until the deviation angle approaches zero, thus completing the precise alignment of the fire source.

6. The substation robot autonomous path planning and fire extinguishing method based on fire source localization according to claim 5, characterized in that, The fusion processing of the visible light image and the infrared thermal image includes: Gaussian pyramid decomposition was performed on the visible light image and the infrared thermal image respectively to obtain their respective multi-scale image sequences; On the same scale layer, extract the texture features of the visible light image layer and the temperature features of the infrared thermal image layer; The texture features and temperature features are weighted and fused to generate a fused feature map for this scale layer, wherein the weights are dynamically adjusted according to the feature saliency of the fire source area; The highest-scale fused feature map is reconstructed using a Laplacian pyramid, and then fused upwards level by level to finally obtain a fused image containing rich texture information and significant heat source information.

7. The substation robot autonomous path planning and fire extinguishing method based on fire source localization according to claim 5, characterized in that, The step of activating the fire extinguishing device to perform fire extinguishing operations on the target fire source includes: Based on the fire intensity level of the target fire source in the fire situation map of the whole station, the corresponding spray parameters are matched from the preset fire extinguishing strategy library. The spray parameters include the fire extinguishing agent spray flow rate, spray sweeping mode and spray duration. The gimbal of the fire extinguishing device is controlled to move according to the spray sweeping mode so that the coverage of the fire extinguishing agent can cover the target fire source area. During the spraying duration, the extinguishing agent valve is opened at the extinguishing agent spray flow rate to perform the spraying operation; During the spraying operation, the temperature changes of the target fire source area are continuously monitored using an infrared thermal imager.

8. The substation robot autonomous path planning and fire extinguishing method based on fire source localization according to claim 7, characterized in that, The continuous receipt of updated fire situation maps of the entire station during firefighting operations to determine whether the fire intensity level of the target fire source has dropped below a safe threshold includes: After the spraying duration ends, acquire the latest fire situation map of the entire station generated by the multi-sensor fusion detection device; From the latest generated fire situation map of the entire site, check whether the fire intensity level corresponding to the target fire source has been updated; If the fire intensity level of the target fire source has dropped below the predefined safety threshold, it is determined that the target fire source has been extinguished and a fire extinguishing success signal is generated. If the fire intensity level of the target fire source does not drop below the safety threshold, the global navigation path will be recalculated or the spray parameters will be adjusted based on the latest three-dimensional coordinates of the fire source and the fire intensity level, and a new round of fire extinguishing operations will be initiated.

9. The substation robot autonomous path planning and fire extinguishing method based on fire source localization according to claim 1, characterized in that, The method also includes a post-fire extinguishing autonomous inspection and reignition monitoring process: After determining that the current target fire source has been extinguished, check whether there are any other untreated fire sources in the overall fire situation map of the station. If there are other untreated fire sources, the nearest untreated fire source is set as the new target fire source, and the process returns to the step of calculating the global navigation path based on the real-time position of the substation robot. If there are no other untreated fire sources, the substation robot will be controlled to move and patrol within the substation according to the predetermined patrol route. During the patrol, the data from the environmental perception sensor and the multi-sensor fusion detection device are continuously analyzed to monitor whether new fires have occurred or whether extinguished fires have shown signs of reignition.

10. The substation robot autonomous path planning and fire extinguishing method based on fire source localization according to claim 9, characterized in that, The substation robot is controlled to move and inspect within the substation according to a predetermined inspection path, including: Load the pre-generated routine substation inspection path point sequence from the substation control center or the robot's local storage; Each point in the inspection path point sequence is used as a temporary navigation target, and the substation robot is controlled to move to that point using local path planning and obstacle avoidance algorithms; Upon arriving at each inspection route point, the substation robot is controlled to make a brief stop and rotate its onboard visible light camera and infrared thermal imager to perform a 360-degree scan of the surrounding environment. The environmental data obtained by scanning is compared and analyzed with preset normal state baseline data to detect abnormal temperature, smoke or open flame characteristics.

Images