Underwater robot inspection obstacle avoidance control method and system

Abstract

The invention discloses an underwater robot inspection obstacle avoidance control method and system, and relates to the technical field of robots, an event camera is introduced to construct a two-dimensional event response distribution diagram, local feature vectors acquired by multiple sensors are fused, a semantic point cloud is formed, and a geometric, semantic and dynamic residual triple optimization mechanism is constructed in combination with a historical environment map. According to the method, depth alignment of a current state and a historical map is realized, the dynamic obstacle identification precision and the adaptive ability of obstacle avoidance path planning are further improved, risk level division can be carried out on an updated environment map through calculation of a dynamic confidence coefficient C and a color coding mechanism, high-dynamic-risk, medium-risk and low-risk areas are distinguished, and the accuracy of obstacle avoidance is improved. According to the method, the intelligent decision-making capability of path selection is effectively enhanced, and on the basis, the data in the red area is preferentially uploaded by combining a return priority scheduling mechanism, so that the consistency matching of the task risk level and the data uploading priority is realized, and redundant data transmission is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of robots, and specifically to an underwater robot inspection and obstacle avoidance control method and system. Background Art

[0002] In underwater application scenarios, underwater robots are widely used in tasks such as aquaculture monitoring, seabed facility maintenance, and energy pipeline inspection. They need to face complex and changeable environmental conditions with limited vision and significant sensor interference. Under such technical requirements, the control strategy with the core of "identifying dynamic obstacles during the autonomous inspection process of underwater robots" has become the research focus. Especially in a dynamic environment, the ability to detect abnormal disturbances and adjust the real-time path directly relates to the safety of the robot and the efficiency of task completion.

[0003] Most of the current mainstream underwater robot obstacle avoidance control methods rely on sonar, structured light, or traditional vision sensors to obtain static environment information. When facing a highly dynamic disturbance environment, such as surging water flow, temporarily moving objects, or biological disturbances, they often lack high responsiveness and adaptive update capabilities and are difficult to accurately identify abnormal change areas. Such defects may cause a series of system anomalies during the task execution process. For example, the robot misjudges a highly disturbed area as a passable area, resulting in an increased collision risk. Even during the multi-region task feedback process, due to incorrect risk judgment, important information is delayed in uploading, affecting the overall task scheduling and the efficiency of the human-machine collaboration mechanism, and ultimately damaging the operation stability and precision reliability. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention provides an underwater robot inspection and obstacle avoidance control method and system, which solves the problems in the above background art.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: An underwater robot inspection and obstacle avoidance control method, including the following steps, S1: Use a plurality of groups of sensors and event cameras to continuously monitor the environmental conditions of the underwater scene to generate a two-dimensional event response distribution map. Based on the two-dimensional event response distribution map, determine the fused strong feature vectors , and identify abnormal dynamic change areas; S2: Pre-acquire the environmental map constructed at the previous moment, combine it with the two-dimensional event response distribution map, determine several groups of point pairs, and analyze the positional relationship, semantic relationship, and dynamic change situation between each point pair to construct a global objective function , and through optimization and solution, obtain the optimal state variables . Based on the optimal state variables , update the environmental map constructed at the previous moment to generate an updated environmental map; S3: Determine the dynamic situation of each position in the updated environmental map to generate the dynamic confidence C of each sub-region in the updated environmental map, so as to provide the underwater robot with the choice of obstacle avoidance path; S4: Trigger the backhaul priority scheduling mechanism based on the value of the dynamic confidence C of each sub-region.

[0006] Preferably, S11: Pre-capture the events that occur when the brightness of each pixel in the underwater scene changes by using an event camera to obtain the time stamps of the events and pixel coordinates; S12: Based on the parameters obtained in S11, simulate the real dynamic change information in the underwater scene and consider the noise interference to calculate and obtain the event signal strength at different pixel coordinates , specifically: ; In the formula, is the event signal strength at the pixel coordinates (x, y) at time t, t is the current time point, is the time offset index of multiple events that occur, is the time decay constant, is the position weight; S13: Traverse all pixel coordinates on the image plane involved in the event camera once to form a two-dimensional event response distribution map with the event signal strength value as the pixel intensity.

[0007] Preferably, S14: Use several groups of sensors to monitor the environmental conditions of the underwater scene in real time, and divide the two-dimensional event response distribution map into several groups of regions to generate the local feature vectors h collected by different sensors. By fusing the local feature vectors h collected by each sensor, the fused strong feature vector ; S15: Associate the fused strong feature vector with the monitoring positions of the corresponding sensors to form a representation corresponding to each local region in the underwater scene to generate a semantic point cloud ; S16: According to the event signal strength at different pixel coordinates obtained in S12, use the threshold segmentation method to identify the dynamic target area to form a binary dynamic obstacle mask , and based on the binary dynamic obstacle mask , identify the abnormal dynamic change area.

[0008] Preferably, S21: Pre-obtain the environmental map constructed at the previous moment, and for each semantic point cloud in each local region by using the nearest neighbor search algorithm Establish a preliminary correspondence with the environmental map constructed at the previous moment to obtain several sets of point pairs; S22: Preset the state variables to be optimized and, according to the position of each semantic point cloud in several sets of point pairs and the matching point corresponding to it in the environmental map constructed at the previous moment, obtain the geometric residual of each point pair Specifically: ; where is the geometric residual of the k-th semantic point cloud, is the position of the k-th semantic point cloud, is the position in the environmental map constructed at the previous moment that matches the position of the k-th semantic point cloud; is the rigid transformation matrix to be optimized; S23: According to the point pairs determined in S21, extract the corresponding semantic point clouds in each point pair and the fused strong feature vectors corresponding to the positions that match the positions of the corresponding semantic point clouds in the environmental map constructed at the previous moment, and obtain the semantic residual by analyzing the internal feature differences of each point pair Specifically: ; In the formula, is the semantic residual of the k-th semantic point cloud, is the fused strong feature vector within the k-th semantic point cloud, is the fused strong feature vector corresponding to the position that matches the position of the corresponding semantic point cloud in the environmental map constructed at the previous moment ; S24: According to the abnormally dynamic change area identified by S16, determine the number of times of abnormally dynamic changes in the abnormally dynamic change area in the past, and obtain the dynamic residual of each point pair by considering the interference of the abnormally dynamic change area on the update of the environmental map constructed at the previous moment Specifically: ; In the formula, is the dynamic residual of the k-th semantic point cloud, is the dynamic weight, where ; where is the dynamic probability calculated according to the binary dynamic obstacle mask

[0009] Preferably, S25: Construct a global objective function to minimize the geometric residual , semantic residual and dynamic residual by weighted sum of squares, so that the state variables to be optimized obtained by optimization and solution​ As the optimal state variable , through the optimal state variable , project the currently fused strong feature vector into the coordinate system defined by the environmental map constructed at the previous moment to achieve an alignment effect, specifically: ; where is the new position coordinate after converting the position of the k-th semantic point cloud in the two-dimensional event response distribution map to the coordinate system defined by the environmental map constructed at the previous moment.

[0010] Preferably, S31: Based on the environmental map constructed at the previous moment, combined with the new position coordinate after converting the position of the k-th semantic point cloud in the two-dimensional event response distribution map to the coordinate system defined by the environmental map constructed at the previous moment , obtain the updated environmental map.

[0011] Preferably, S32: Divide the updated environmental map into several groups of sub-regions; S33: In the underwater scenario, in order to consider the dynamic degree of each position in the underwater scenario, present a dynamic trajectory in the updated environmental map, where the updated environmental map contains the dynamic confidence C of each sub-region, specifically: ; In the formula, is the dynamic confidence of the A-th sub-region in the updated environmental map, is the number of static obstacle points in the A-th sub-region, is the corresponding semantic point cloud in the A-th sub-region is the total amount of positions, is the variance value of the change rate of the event signal intensity at the a-th position in the corresponding sub-region, and a is the position number in the corresponding sub-region.

[0012] Preferably, S34: Preset a confidence range. If the dynamic confidence C of each sub-region exceeds the confidence range, then present the corresponding sub-region in the updated environmental map in green; if the dynamic confidence C of each sub-region falls within the confidence range, then present the corresponding sub-region in the updated environmental map in orange; if the dynamic confidence C of each sub-region does not exceed the confidence threshold and does not fall within the confidence range, then present the corresponding sub-region in the updated environmental map in red; S35: According to the position of the current underwater robot in the updated environmental map, and combined with the sub-regions presented in red in S34, provide a choice of obstacle avoidance path for the underwater robot.

[0013] Preferably, S41: According to the color rendering of the corresponding sub-regions in the updated environmental map in S34, execute the feedback priority scheduling mechanism. The specific execution steps are as follows: S411: Extract the sub-regions presented in red in the updated environmental map and perform a priority upload operation on them; S412: Extract the sub-regions presented in orange and the sub-regions presented in green in the updated environmental map. First, perform data compression processing on the sub-regions presented in orange and the sub-regions presented in green, and then perform an upload operation. Among them, the sub-regions presented in orange are uploaded after the sub-regions presented in red, and the sub-regions presented in green are uploaded after the sub-regions presented in orange.

[0014] The underwater robot inspection and obstacle avoidance control system includes a map construction module, a map update module, a rendering module, and a feedback module; The map construction module is used to use a number of groups of sensors and event cameras to monitor the environmental conditions of the underwater scene in real time to generate a two-dimensional event response distribution map. Based on the two-dimensional event response distribution map, determine the fused strong feature vectors , and identify abnormal dynamic change areas; The map update module is used to pre-obtain the environmental map constructed at the previous moment, combine the two-dimensional event response distribution map, determine a number of groups of point pairs, and analyze the positional relationship, semantic relationship, and dynamic change situation between each point pair to construct a global objective function , through optimization and solution, obtain the optimal state variables , based on the optimal state variables , update the environmental map constructed at the previous moment to generate an updated environmental map; The rendering module is used to determine the dynamic situation of each position in the updated environmental map to generate the dynamic confidence C of each sub-region in the updated environmental map, so as to provide a choice of obstacle avoidance path for the underwater robot; The feedback module is used to trigger the feedback priority scheduling mechanism based on the numerical values of the dynamic confidence C of each sub-region.

[0015] The present invention provides an underwater robot inspection and obstacle avoidance control method and system, which has the following beneficial effects: By introducing an event camera, a two-dimensional event response distribution map is constructed, and local feature vectors obtained by multiple sensors are fused to form a semantic point cloud. Combining with the historical environment map, a triple optimization mechanism of geometry, semantics, and dynamic residuals is constructed to achieve deep alignment between the current state and the historical map, further improving the dynamic obstacle recognition accuracy and the adaptive ability of obstacle avoidance path planning. Through the calculation of dynamic confidence C and the color coding mechanism, the updated environment map can be divided into risk levels, distinguishing high-dynamic risk, medium-risk, and low-risk areas, effectively enhancing the intelligent decision-making ability of path selection. On this basis, combined with the backhaul priority scheduling mechanism, the data in the red area is uploaded first, while the orange and green areas are compressed and then uploaded in sequence, realizing the consistency matching between the task risk level and the data upload priority. This hierarchical upload mechanism further optimizes the data backhaul efficiency of the underwater robot under limited bandwidth, reduces redundant data transmission, and ensures that the data in high-risk areas is processed and responded in a timely manner, effectively improving the system's perception agility and operation safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of the inspection and obstacle avoidance control method for the underwater robot of the present invention; Figure 2 Overall logic diagram of the inspection and obstacle avoidance control method for the underwater robot of the present invention; Figure 3 Partial logic diagram of the inspection and obstacle avoidance control method for the underwater robot of the present invention; Figure 4 Block diagram of the inspection and obstacle avoidance control system for the underwater robot of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Embodiment 1 Please refer to Figures 1 to 3 , the present invention provides an inspection and obstacle avoidance control method for an underwater robot, including the following steps: S1: Using a plurality of groups of sensors and an event camera, the environmental conditions of the underwater scene are monitored in real time to generate a two-dimensional event response distribution map. Based on the two-dimensional event response distribution map, the fused strong feature vector is determined, and the abnormal dynamic change area is identified; S2: Obtain the environment map constructed at the previous moment in advance, combine it with the two-dimensional event response distribution map, determine several groups of point pairs, and analyze the position relationship, semantic relationship and dynamic change between each point pair to construct the global objective function , after optimization, the optimal state variables are obtained , based on the optimal state variables , update the environment map constructed at the previous moment and generate an updated environment map; S3: Determine the dynamic situation of each position in the updated environment map to generate a dynamic confidence C of each sub-area in the updated environment map to provide the underwater robot with an obstacle avoidance path selection; S4: Based on the value of the dynamic confidence C of each sub-area, trigger the backhaul priority scheduling mechanism.

[0019] In this embodiment, a dynamic perception mechanism based on event camera and multimodal sensor fusion is used to realize real-time recognition of disturbed areas in the underwater environment and dynamic map update. By introducing semantic point cloud, fusion residual optimization, dynamic confidence metric quantitative evaluation and color-coded feedback scheduling mechanism, the robot can autonomously judge the path passability and risk level, thereby realizing a closed-loop control link of "environment modeling-dynamic recognition-obstacle avoidance path-data scheduling".

[0020] In particular, the dynamic confidence C and color partitioning mechanism introduced in steps S34-S41 can realize hierarchical management and intelligent scheduling of inspection data: high-risk areas (red) are judged as dynamic active areas and uploaded first to ensure timely response to sudden risks; medium-risk areas (orange) are uploaded later after compression; low-risk areas (green) are also uploaded last after compression to save bandwidth and computing resources. For example: in a submarine pipeline inspection operation, when the underwater robot passes a certain place, it detects a sudden high-speed moving target (such as deep-sea creatures or drifting devices). The intensity of its event signal changes dramatically, and the corresponding area is marked as red. The system immediately transmits the data of this area back to the control platform for analysis and adjustment of the robot path, while the slight dynamic disturbance in the adjacent area is classified as orange or green, and is only compressed and transmitted slowly to avoid system congestion. It can be seen that this method effectively improves the task response speed and data utilization efficiency, and enhances the robot's operation autonomy and information perception reliability in complex underwater environments.

[0021] Example 2 Please refer to Figures 1 to 3 , specifically: S11: Use an event camera to capture the events that occur when the brightness of each pixel in the underwater scene changes in advance to obtain the timestamp of the event occurrence and pixel coordinates; An event camera is a sensor designed based on the biological vision mechanism. Different from traditional frame cameras, its working principle is "asynchronous event-driven". Traditional cameras capture whole-frame images at fixed time intervals, while event cameras respond immediately when a change in pixel brightness is detected, generating an "event", rather than capturing static images. That is, instead of capturing images frame by frame, it records events at the pixel level. When there is a brightness change in the environment, the sensor generates an event at the corresponding pixel position; S12: Obtain parameters (the timestamp of the event occurrence and pixel coordinates) based on S11, simulate the real dynamic change information in the underwater scene, and consider noise interference to calculate and obtain the event signal intensity at different pixel coordinates Specifically: ; In the formula, is the event signal intensity at pixel coordinates (x, y) at time t, t is the current time point, is the time offset index for multiple events that occur, is the time decay constant; is the position weight, which characterizes the importance of each pixel position in the event response. This item can be obtained through pre-collected underwater environment sample data and neural network training. The training data includes a large number of underwater scene images labeled with real obstacles and background noise, so that the model can automatically learn to have a low response to noise, which is used to suppress noise interference caused by water particles, light scattering, etc.

[0022] The time decay constant is used to regulate the decay rate over time and can be obtained through sensor calibration experiments; in the underwater environment, a reasonable value can be fitted according to experimental data to ensure the distinction between noise and real dynamic events.

[0023] is the Dirac function, which is activated at the moment of event occurrence and is directly output by the event camera. When the brightness of a certain pixel changes at moment, this item takes 1, otherwise it is 0. Using the Dirac function is only activated at the moment of event occurrence and is weighted by the subsequent exponential decay function in the time domain to reflect its influence over time. is the time exponential decay function, which is used to smooth and weight the event signal to ensure that the signal contribution is smaller the farther away from the event moment; Due to the introduction of exponential decay and position weight , It can condense information such as the frequency and intensity of recent events for pixel coordinates (x, y), reflecting whether the dynamic changes are active.

[0024] S13: Traverse all pixel coordinates on the image plane involved in the event camera once to form a two-dimensional event response distribution map with the event signal intensity value as the pixel intensity.

[0025] The pixel position refers to the specific pixel coordinates on the event camera sensor, which clearly indicates the specific area where the brightness change occurs; In this embodiment, the present invention uses an event camera to capture the brightness change of each pixel in real time in an underwater environment, obtaining the event timestamp and its coordinates of each pixel; subsequently, based on the parameters in S11, through exponential decay and spatial weighting processing, the event signal intensity of each pixel is calculated, which comprehensively reflects the frequency and intensity of events occurring in the pixel area in the near future, and effectively suppresses short-term interference introduced by noise; Finally, the system traverses all pixel coordinates on the image plane to form a two-dimensional event response distribution map with the event signal intensity as the pixel value. The beneficial effect of this method is that it can quickly and accurately condense the real dynamic change information in the underwater environment, providing high-quality data support for subsequent obstacle detection and obstacle avoidance control. For example, in practical applications, when a school of fish suddenly enters the monitoring area of the robot, the event signal intensity of the pixels in this area will rise sharply. After the above processing, in the event response distribution map, this area will show an obvious high-intensity area, which is thus recognized as an abnormal dynamic change area by the system, prompting the robot to adjust the obstacle avoidance strategy to ensure safe passage.

[0026] Embodiment 3 Please refer to Figures 1 to 3 , specifically: S14: Use several groups of sensors to monitor the environmental conditions of the underwater scene in real time, and divide the two-dimensional event response distribution map into several groups of regions to generate local feature vectors h collected by different sensors. By fusing the local feature vectors h collected by each sensor, a fused strong feature vector is obtained , where specifically: ; In the formula, is the fused strong feature vector in the i-th region, both i and j are the numbers of the regions, is the neighborhood set corresponding to the i-th region. The neighborhood set represents the regions adjacent to the corresponding region. By summarizing the information of these nodes, cross-modal and cross-data source feature alignment and enhancement are achieved; is the attention weight, reflecting the correlation importance between regions i and j during the cross-modal fusion process, which is calculated using the input features by setting a gating network or attention mechanism (such as dot product attention), and its value is normalized to ensure the sum is 1; and are the local feature vectors of the i-th and j-th regions, is to concatenate the local feature vectors of the i-th and j-th regions, directly cascading the two vectors to form a higher-dimensional feature representation, is the activation function (Sigmoid function), used to introduce non-linear mapping; is the feature fusion matrix, obtained by backpropagation algorithm for each local feature vector h, used to perform a linear transformation on the concatenated feature vector, enabling information from different modalities to be uniformly represented in the same feature space, thus ensuring that the features of each modality maintain effective semantic information after fusion; Specifically, in the underwater environment, multi-source sensor data such as sonar, event cameras, and IMUs are often integrated. Each sensor has different signal-to-noise ratios and spatial resolutions. The main goal of cross-modal fusion is to align the feature representations of each sensor in the same coordinate system to avoid the limitations of single-modal information.

[0027] The fused strong feature vector is used to fully integrate the local information from different sensors and has stronger robustness and discriminative ability.

[0028] S15: Associate the fused strong feature vector with the monitoring positions of the corresponding sensors to form a representation corresponding to each local area in the underwater scene, so as to generate a semantic point cloud ; The semantic point cloud contains the position information and the strong feature vector at the corresponding position ; S16: According to the event signal intensity at different pixel coordinates obtained in S12, use the threshold segmentation method to identify the dynamic target area to form a binary dynamic obstacle mask . Based on the binary dynamic obstacle mask , identify the abnormal dynamic change area. The binary dynamic obstacle mask means marking those pixels determined to be "dynamically changing" as 1 and the remaining pixels as 0, thus forming a binary map that can represent the distribution of dynamic targets.

[0029] In this embodiment, the present invention realizes the accurate perception and modeling of the underwater environment through steps S14 to S16, further improving the recognition ability of dynamic obstacles and the efficiency of obstacle avoidance decision-making. Specifically, in step S14, multiple groups of sensors are used to monitor the environmental conditions of the underwater scene in real time, and local feature vectors from different sensors are fused to generate high-dimensional strong feature vectors. By weighting and splicing the feature vectors of multiple regions and introducing an activation function and a feature fusion matrix, this process effectively expands the information expression ability, enabling data from different modalities to be uniformly represented in the same feature space, thereby realizing the accurate analysis and dynamic perception of the underwater scene.

[0030] Taking the recognition of dynamic obstacles as an example, when the underwater robot cruises in an underwater tunnel, the surrounding water flow and countercurrent may cause changes in the surrounding environment. Through the semantic point cloud generated in step S15, the system can accurately reflect the dynamic changes in different regions, helping the robot determine which regions are static and which regions have dynamic obstacles. At this time, according to step S16, the system identifies the dynamic target region through the threshold segmentation method and generates a binary dynamic obstacle mask. This mechanism effectively reduces false alarms and missed detections caused by dynamic obstacles. Through the processing in this section, the system not only enhances the sensitivity to the local dynamic obstacle area, but also can extract useful feature information from high-resolution images, while removing noise, improving the autonomous navigation and obstacle avoidance capabilities of the robot in complex underwater environments. For example, under the influence of strong water flow on the underwater tunnel structure, the system can dynamically distinguish the interference signal of the water flow from real obstacles, plan a safer path for the robot, and avoid collisions. This solution effectively improves the adaptability of the system to dynamic environments and ensures the efficiency and stability of task execution. For example, in an underwater pipeline inspection, the system divides the pipeline area into blocks. When a large number of event signals are detected by sensors in a certain area and the fused local features indicate that the edge of the area is a pipeline, the semantic point cloud generated by the fused features will accurately depict the pipeline contour; at the same time, the binary obstacle mask extracted through threshold segmentation identifies the dynamic interference area, ensuring that when the obstacle is severe, the system preferentially transmits complete image data to support fine decision-making, thereby greatly reducing data redundancy and communication load.

[0031] Embodiment 4 Please refer to Figures 1 to 3 , specifically: S21: Pre-acquire the environmental map constructed at the previous moment, and establish a preliminary correspondence between each semantic point cloud in each local region and the environmental map constructed at the previous moment by using the nearest neighbor search algorithm to obtain several groups of point pairs; S22: Preset the state variables to be optimized and, according to each semantic point cloud in several groups of point pairs ​The position and the corresponding matching points in the environmental map constructed at the previous moment, and obtain the geometric residuals of each point pair , specifically: ; where is the geometric residual of the k-th semantic point cloud, is the position of the k-th semantic point cloud, is the position in the environmental map constructed at the previous moment that matches the position of the k-th semantic point cloud; is the rigid transformation matrix to be optimized, which is the transformation matrix obtained from the state variables to be optimized , and the rigid transformation matrix to be optimized is the transformation matrix from the current coordinate system to the coordinate system in the environmental map constructed at the previous moment (belonging to the state variables to be optimized ); The geometric residual After passing through the current rigid transformation matrix T to be optimized, the degree of mismatch in the positions of the two sets of corresponding points in space reflects the geometric alignment error of the point pair; By using the ICP algorithm to find the nearest neighbor matching points in the environmental map constructed at the previous moment, and this nearest neighbor matching point is the position of the corresponding semantic point cloud ; where the number of point pairs is the same as the number of semantic point clouds , and the number of each semantic point cloud corresponds one-to-one with the number of the point pair; S23: According to the point pairs determined in S21, extract the corresponding semantic point clouds in each point pair and the position in the environmental map constructed at the previous moment that matches the position of the corresponding semantic point cloud, and the corresponding fused strong feature vectors , and obtain the semantic residuals by analyzing the internal feature differences of each point pair, specifically: ; in the formula, is the semantic residual of the k-th semantic point cloud, is the fused strong feature vector within the k-th semantic point cloud, is the fused strong feature vector corresponding to the position in the environmental map constructed at the previous moment that matches the position of the corresponding semantic point cloud ; The semantic residual reflects the matching degree in semantic features; S24: According to the abnormal dynamic change areas identified in S16, determine the number of times of abnormal dynamic changes in the abnormal dynamic change areas in the past, and obtain the dynamic residuals of each point pair by considering the interference of the abnormal dynamic change areas on the update of the environmental map constructed at the previous moment , specifically: ; where, is the dynamic residual of the k-th semantic point cloud, is the dynamic weight, and its value is calculated based on the probability of being identified as dynamic at that point. Among them, ; where is the dynamic probability calculated according to the binarized dynamic obstacle mask , reflecting the possibility of being identified as dynamic at that point. The larger is, the smaller is, indicating that the influence of the dynamic area on the overall residual is weakened, avoiding large interference of dynamic points on global optimization. The dynamic residual

[0032] S25: Construct the global objective function , by minimizing the geometric residual , semantic residual and dynamic residual through weighted sum of squares, so that the state variable to be optimized obtained by optimization solution is used as the optimal state variable . Through the optimal state variable , project the current fused strong feature vector into the coordinate system defined by the environmental map constructed at the previous moment to achieve alignment effect. Specifically: ; where is the new position coordinate after converting the position of the k-th semantic point cloud in the two-dimensional event response distribution map to the coordinate system defined by the environmental map constructed at the previous moment; The specific expression of the global objective function is: ; In the formula, k is the number of the semantic point cloud, is to find the optimal state variable that makes the global objective function reach the minimum value among all the values of the state variables to be optimized , that is: among all combinations of the state variables to be optimized , find a set of parameter combinations that make the global objective function reach the minimum value; and are both weights, represents controlling the influence of semantic information on overall optimization, is used to adjust the contribution of the constraint of this part ( ), and These two parameters obtain the optimal configuration through offline experiments. represents the Euclidean distance norm; state variables to be optimized refers to the key parameter used to describe the spatial relationship between the current local scan data (semantic point cloud ) collected by the underwater robot and the environmental map constructed at the previous moment. Its main function is to map the data in the current local coordinate system to the coordinate system of the environmental map constructed at the previous moment through rotation and translation, so as to achieve the alignment and fusion of the two. State variables to be optimized consists of a 3×3 rotation matrix and a 3×1 translation vector; By optimizing and solving the state variables to be optimized , it is possible to update and correct the previous map using the latest information in each data acquisition, thereby effectively reducing the cumulative error. If the direct mapping is not optimized, it is difficult to correct the positioning error caused by sensor noise and dynamic interference.

[0033] Specifically, aligning the current data with the previous map is to fuse the new data with the historical data and dynamically update the map to make it more in line with the actual situation of the current environment.

[0034] In this embodiment, the present invention realizes the precise alignment and fusion of the currently collected data and the previous environmental map through steps S21–S25, thereby dynamically updating the map to make it closer to the real underwater environment. For example, use the nearest neighbor search to establish a preliminary correspondence between each semantic point cloud in each local area and the previous map, and then calculate the geometric residual based on the matching point pairs to reflect the spatial position deviation, the semantic residual to reflect the consistency of feature description, and the dynamic residual to reflect the interference of abnormal dynamic areas; finally, construct a global objective function, and obtain the optimal state variable, that is, the optimal rigid transformation matrix T, by minimizing the weighted sum of squares of each residual, and project the currently fused strong feature vector into the coordinate system of the previous environmental map. The beneficial effect of this method is that it can effectively fuse the real-time collected data with the historical map, correct the cumulative error, and take into account the interference of dynamic information at the same time, so that the updated environmental map can more accurately reflect the current environmental state.

[0035] Embodiment 5 Please refer to Figure 1 and Figure 3 , specifically: S31: On the basis of the environmental map constructed at the previous moment, combined with the new position coordinates after converting the position of the k-th semantic point cloud in the two-dimensional event response distribution map to the coordinate system defined by the environmental map constructed at the previous moment , obtain the updated environmental map, which is used to provide a reference benchmark for the robot's global navigation and obstacle avoidance.

[0036] S32: Divide the updated environment map into a plurality of groups of sub-areas; S33: In the underwater scene, in order to consider the dynamic degree of each position of the underwater scene, a dynamic trajectory is presented in the updated environment map, wherein the updated environment map includes the dynamic confidence C of each sub-area, specifically: ; In the formula, is the dynamic confidence of the Ath sub-area in the updated environment map, is the number of static obstacle points in the Ath sub-area, is the semantic point cloud corresponding to the Ath sub-region The total number of positions, is the variance value of the event signal intensity change rate at the a-th position in the corresponding sub-region, a is the position number in the corresponding sub-region; exp(*) is an exponential function; is the proportion of static points, To quantify dynamic information, an exponential function is used as a decay term to suppress areas with obvious motion. The smaller the value, the more drastic the dynamic changes in the corresponding sub-region are and the higher the risk is; In this embodiment, first, step S31 uses the new position coordinates after converting the kth semantic point cloud in the two-dimensional event response distribution map to the previous map coordinate system to accurately update the global map so that it reflects the current environmental status in real time and provides accurate navigation and obstacle avoidance benchmarks for the underwater robot.

[0037] Then, step S32 divides the updated map into several sub-areas to achieve segmented management of local risks. Through step S33, the static proportion is calculated in each sub-area based on the number of static obstacle points and the total number of points, and combined with the variance of the event signal strength change rate at the corresponding position, it is adjusted by the exponential decay function to generate the dynamic confidence C, which is the quantitative index of dynamic information. Through zoning management and dynamic risk quantification, the system can intelligently allocate path selection and decision-making basis on a global scale, further improve obstacle avoidance efficiency and navigation safety, and reduce the risk of misjudgment caused by dynamic changes in the environment.

[0038] Example 6 Please refer to Figure 1, specifically: S34: Preset a confidence range. If the dynamic confidence C of each sub-region exceeds the confidence range, it indicates that most positions in the corresponding sub-region are judged to be static, with a small movement amplitude and low risk. At this time, the corresponding sub-region in the updated environmental map will be presented in green; if the dynamic confidence C of each sub-region falls within the confidence range, it indicates that some positions in the corresponding sub-region are judged to be static, with a relatively small movement amplitude and medium risk. At this time, the corresponding sub-region in the updated environmental map will be presented in orange; if the dynamic confidence C of each sub-region does not exceed the confidence threshold and does not fall within the confidence range, it indicates that there are many moving points in the corresponding sub-region and high risk. At this time, the corresponding sub-region in the updated environmental map will be presented in red; This map can be color-coded internally according to risk levels (such as green, orange, and red areas) to reflect the dynamic risks and feasible obstacle avoidance paths in each region.

[0039] S35: Based on the position of the current underwater robot in the updated environmental map and combined with the sub-regions presented in red in S34, provide options for the underwater robot's obstacle avoidance paths.

[0040] S41: According to the color presentation of the corresponding sub-regions in the updated environmental map in S34, execute the backhaul priority scheduling mechanism. The specific execution steps are as follows: S411: Extract the sub-regions presented in red in the updated environmental map and give them priority upload operations; S412: Extract the sub-regions presented in orange and green in the updated environmental map, first perform data compression processing on the sub-regions presented in orange and green, and then perform upload operations. Among them, the sub-regions presented in orange are uploaded after the sub-regions presented in red, and the sub-regions presented in green are uploaded after the sub-regions presented in orange. This priority scheduling enables the gradual upload of data in batches, regions, and priorities even if the entire map is large, rather than uploading all data at once.

[0041] Since the bandwidth of the underwater communication system is usually very limited, directly transmitting a large amount of high-resolution data is likely to occupy all the bandwidth resources, resulting in delays and packet losses. Therefore, some data needs to be appropriately compressed. At critical moments, such as when encountering high-risk areas, the system needs to quickly obtain complete information to support decision-making, and there is no need to transmit all details during other periods. Therefore, compressing data helps improve transmission efficiency and response speed.

[0042] The backhaul priority scheduling mechanism is a process of classifying and arranging the transmission order of the generated data according to priorities. This mechanism dynamically allocates upload tasks under limited bandwidth resources according to the importance and real-time requirements of different data.

[0043] In this embodiment, first, the updated environmental map is color-coded using a preset confidence range. This color coding intuitively reflects the dynamic risks of each sub-region, facilitating the underwater robot to preferentially select green and orange areas when planning an obstacle avoidance path and avoid red high-risk areas. For example, assuming that in the environmental map obtained by the robot during the seabed inspection, a sub-region is shown in red, it indicates that the dynamic activities of the obstacle are frequent and the risk is high. According to S35, the robot will avoid this area as a driving route and choose a green or orange area as a safe passage.

[0044] Meanwhile, in S41, the system executes a data transmission scheduling mechanism according to the map color information: First, the detailed sensing data of the red area is preferentially uploaded to ensure complete data support for obstacle avoidance decisions in high-risk areas; while the data of the orange and green areas is uploaded after data compression, and the orange area is uploaded before the green area, thus effectively saving communication bandwidth and processing resources while ensuring the timely transmission of key decision-making data. This hierarchical upload strategy significantly reduces data redundancy and system load, and improves the overall communication efficiency and obstacle avoidance safety.

[0045] Embodiment 7 Please refer to Figure 4 , specifically: The underwater robot inspection and obstacle avoidance control system includes a map construction module, a map update module, a presentation module, and a transmission module; The map construction module is used to use a number of groups of sensors and event cameras to monitor the environmental conditions of the underwater scene in real time to generate a two-dimensional event response distribution map, and based on the two-dimensional event response distribution map, determine the fused strong feature vector , and identify abnormal dynamic change areas; The map update module is used to pre-obtain the environmental map constructed at the previous moment, combine it with the two-dimensional event response distribution map, determine a number of groups of point pairs, and analyze the positional relationship, semantic relationship, and dynamic change situation between each point pair to construct a global objective function , through optimization and solution, obtain the optimal state variable , based on the optimal state variable , update the environmental map constructed at the previous moment to generate an updated environmental map; The presentation module is used to determine the dynamic situation of each position in the updated environmental map to generate the dynamic confidence C of each sub-region in the updated environmental map, so as to provide the underwater robot with the choice of an obstacle avoidance path; The transmission module is used to trigger a transmission priority scheduling mechanism based on the numerical value of the dynamic confidence C of each sub-region.

[0046] Although embodiments of the present invention have been shown and described, those of ordinary skill in the art will appreciate that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An underwater robot inspection and obstacle avoidance control method, characterized in that: The following steps are included: S1: Use several sets of sensors and event cameras to monitor the environmental conditions of underwater scenes in real time to generate a two-dimensional event response distribution map. Based on the two-dimensional event response distribution map, determine the fused strong feature vector , and identify areas of abnormal dynamic changes; S2: Obtain the environment map constructed at the previous moment in advance, combine it with the two-dimensional event response distribution map, determine several groups of point pairs, and analyze the position relationship, semantic relationship and dynamic change between each point pair to construct the global objective function , after optimization, the optimal state variables are obtained , based on the optimal state variables , update the environment map constructed at the previous moment and generate an updated environment map; S3: Determine the dynamic situation of each position in the updated environment map to generate a dynamic confidence C of each sub-area in the updated environment map to provide the underwater robot with an obstacle avoidance path selection; S4: Based on the value of the dynamic confidence C of each sub-area, trigger the backhaul priority scheduling mechanism.

2. The underwater robot inspection and obstacle avoidance control method according to claim 1, characterized in that: S11: Use an event camera to capture the events that occur when the brightness of each pixel in the underwater scene changes in advance to obtain the timestamp of the event occurrence and pixel coordinates; S12: Based on the parameters obtained in S11, simulate the real dynamic change information in the underwater scene and consider noise interference to calculate the event signal intensity at different pixel coordinates , specifically: ； In the formula, is the event signal intensity at the pixel coordinate (x, y) at time t, where t is the current time point, is a time offset index for multiple events that occurred, is the time decay constant, is the position weight; S13: Traverse all pixel coordinates on the image plane involved in the event camera to form a A two-dimensional event response distribution graph where the values ​​are pixel intensities.

3. The underwater robot inspection and obstacle avoidance control method according to claim 2, characterized in that: S14: Use several groups of sensors to monitor the environmental conditions of the underwater scene in real time, and divide the two-dimensional event response distribution map into several groups of regions to generate local feature vectors h acquired by different sensors, and fuse the local feature vectors h acquired by each sensor to obtain a fused strong feature vector ; S15: The fused strong feature vector Associated with the monitoring positions of the corresponding sensors to form a representation corresponding to each local area in the underwater scene to generate a semantic point cloud ; S16: according to the event signal intensity at different pixel coordinates obtained in S12 , the threshold segmentation method is used to identify the dynamic target area to form a binary dynamic obstacle mask , dynamic obstacle mask based on binarization , identifying the abnormal dynamic change area.

4. The underwater robot inspection and obstacle avoidance control method according to claim 3, characterized in that: S21: Obtain the environment map constructed at the previous moment in advance, and use the nearest neighbor search algorithm to find the semantic point cloud in each local area. Establish a preliminary correspondence with the environment map constructed at the previous moment to obtain several groups of point pairs; S22: Preset the state variables to be optimized , and according to each semantic point cloud in several groups of point pairs The position of and the corresponding matching point in the environment map constructed at the previous moment, and obtain the geometric residual of each point pair , specifically: ;in, is the geometric residual of the kth semantic point cloud, is the position of the kth semantic point cloud, is the position that matches the position of the kth semantic point cloud in the environment map constructed at the previous moment; is the rigid transformation matrix to be optimized; S23: According to the point pairs determined in S21, extract the corresponding semantic point cloud in each point pair and the fused strong feature vector corresponding to the position matching the position of the corresponding semantic point cloud in the environment map constructed at the previous moment , by analyzing the feature differences within each point pair to obtain the semantic residual , specifically: ; In the formula, is the semantic residual of the kth semantic point cloud, is the fused strong feature vector in the kth semantic point cloud, The fused strong feature vector corresponding to the position that matches the position of the corresponding semantic point cloud in the environment map constructed at the previous moment ; S24: According to the abnormal dynamic change area identified in S16, determine the number of abnormal dynamic changes that occurred in the abnormal dynamic change area in the past, and obtain the dynamic residual of each point pair by considering the interference of the abnormal dynamic change area on the update of the environment map constructed at the previous moment , specifically: ; In the formula, is the dynamic residual of the kth semantic point cloud, is the dynamic weight, where ;in, is a dynamic obstacle mask based on the binary Compute the dynamic probability.

5. The underwater robot inspection and obstacle avoidance control method according to claim 4, characterized in that: S25: Constructing the global objective function , in order to minimize the geometric residual by , semantic residual and dynamic residual Perform weighted square sum so that the state variables to be optimized obtained by optimization are As the optimal state variable , through the optimal state variable , the current fused strong feature vector Projected into the coordinate system defined by the environment map constructed at the previous moment to achieve alignment, specifically: ;in, The new position coordinates are obtained by converting the position of the kth semantic point cloud in the two-dimensional event response distribution map to the coordinate system defined by the environment map constructed at the previous moment.

6. The underwater robot inspection and obstacle avoidance control method according to claim 5, characterized in that: S31: Based on the environment map constructed at the previous moment, the new position coordinates after the position of the kth semantic point cloud in the two-dimensional event response distribution map is converted to the coordinate system defined by the environment map constructed at the previous moment , get the updated environment map.

7. The underwater robot inspection and obstacle avoidance control method according to claim 6, characterized in that: S32: Divide the updated environment map into a plurality of groups of sub-areas; S33: In the underwater scene, in order to consider the dynamic degree of each position of the underwater scene, a dynamic trajectory is presented in the updated environment map, wherein the updated environment map includes the dynamic confidence C of each sub-area, specifically: ； In the formula, is the dynamic confidence of the Ath sub-area in the updated environment map, is the number of static obstacle points in the Ath sub-area, is the semantic point cloud corresponding to the Ath sub-region The total number of positions, is the variance value of the event signal intensity change rate at the a-th position in the corresponding sub-region, and a is the position number in the corresponding sub-region.

8. The underwater robot inspection and obstacle avoidance control method according to claim 7, characterized in that: S34: presetting a confidence range, if the dynamic confidence C of each sub-region exceeds the confidence range, the corresponding sub-region in the updated environment map is displayed in green; if the dynamic confidence C of each sub-region falls within the confidence range, the corresponding sub-region in the updated environment map is displayed in orange; if the dynamic confidence C of each sub-region does not exceed the confidence threshold and does not fall within the confidence range, the corresponding sub-region in the updated environment map is displayed in red; S35: According to the current position of the underwater robot in the updated environment map and in combination with the sub-area presented in red in S34, an obstacle avoidance path selection is provided for the underwater robot.

9. The underwater robot inspection and obstacle avoidance control method according to claim 8, characterized in that: S41: According to S34, the corresponding sub-areas in the updated environment map are presented in color, and the return priority scheduling mechanism is executed. The specific execution steps are: S411: extracting the sub-areas in red in the updated environment map and uploading them first; S412: Extract the sub-areas in orange and green in the updated environmental map, perform data compression on the sub-areas in orange and green, and then upload them, wherein the sub-areas in orange are uploaded after the sub-areas in red, and the sub-areas in green are uploaded after the sub-areas in orange.

10. An underwater robot inspection and obstacle avoidance control system, used to implement the underwater robot inspection and obstacle avoidance control method according to any one of claims 1 to 9, characterized in that: It includes a graph construction module, a graph update module, a presentation module and a feedback module; The graph construction module is used to use several sets of sensors and event cameras to monitor the environmental conditions of the underwater scene in real time to generate a two-dimensional event response distribution map. Based on the two-dimensional event response distribution map, the fused strong feature vector is determined. , and identify areas of abnormal dynamic changes; The graph update module is used to pre-acquire the environment map constructed at the previous moment, combine it with the two-dimensional event response distribution map, determine several groups of point pairs, and analyze the position relationship, semantic relationship and dynamic change between each point pair to construct the global objective function. , after optimization, the optimal state variables are obtained , based on the optimal state variables , update the environment map constructed at the previous moment and generate an updated environment map; The presentation module is used to determine the dynamic situation of each position in the updated environment map to generate the dynamic confidence C of each sub-area in the updated environment map to provide the underwater robot with the choice of obstacle avoidance path; The feedback module is used to trigger the feedback priority scheduling mechanism based on the value of the dynamic confidence C of each sub-area.

Images