A method for cooperative detection of an underwater vehicle formation and related products

Abstract

The application discloses a kind of methods for the cooperative detection of submersible formation and related products.Therein, the method comprises: obtaining the obstacle distribution map and target candidate list according to the acoustic data, optical data, depth perception data and hydrological environment data collected by the sensor;According to the obstacle distribution map, update the local environment map of the target submersible;For any target, if the confidence corresponding to the target is greater than the first confidence threshold, send the first target discovery message to other submersibles;In response to the target submersible receiving the second target discovery message sent by other submersibles, update the global state table information according to the second target discovery message;According to the state of target submersible, local environment map and global state table information, cooperative detection is carried out with other submersibles.The cooperative detection method of submersible formation in the embodiment of the application can improve the cooperative efficiency and dynamic adaptability between each submersible in the submersible formation.

Description

Technical Field

[0001] This application relates to the fields of underwater unmanned systems and artificial intelligence technology, and in particular to a collaborative detection method for a fleet of submersibles and related products. Background Technology

[0002] Submarine vehicle swarm refers to an underwater intelligent cluster system in which two or more submarines, supported by an underwater communication network and a collaborative control system, conduct joint operations according to a preset spatial configuration, timing logic, and mission strategy.

[0003] In related technologies, before a mission is executed, operators pre-set the tracks and detection areas of each underwater vehicle based on known nautical chart information. The underwater vehicles then perform the mission according to a fixed procedure based on the pre-set tracks and detection areas, resulting in low coordination efficiency and dynamic adaptability of the underwater vehicle formation. Summary of the Invention

[0004] To address the aforementioned issues, this application provides a collaborative detection method and related products for underwater vehicle formations, thereby improving the collaborative efficiency and dynamic adaptability of underwater vehicle formations.

[0005] The embodiments of this application disclose the following technical solutions: In a first aspect, embodiments of this application provide a cooperative detection method for a formation of underwater vehicles, the method comprising: When the target underwater vehicle explores along the detection track, it obtains an obstacle distribution map and a target candidate list based on acoustic data, optical data, depth sensing data, and hydrological environment data collected by the sensors; each target in the target candidate list corresponds to coordinates and confidence level. Update the local environment map of the target submarine based on the obstacle distribution map; For any target, if the confidence level corresponding to the target is greater than the first confidence threshold, a first target discovery message is sent to other underwater vehicles; In response to the target submarine receiving a second target discovery message from another submarine, the global status table information is updated based on the second target discovery message. Based on the target underwater vehicle's status, local environment map, and global status table information, conduct collaborative detection with other underwater vehicles.

[0006] In one possible implementation, as the target underwater vehicle probes along its detection trajectory, it obtains an obstacle distribution map and a target candidate list based on acoustic data, optical data, depth sensing data, and hydrological environment data collected by sensors. These include: When the target underwater vehicle is exploring along the exploration track, the acoustic data, optical data, depth sensing data and hydrological environment data collected by the sensors are fused by the local perception fusion model to obtain fused features; among them, the acoustic data, optical data, depth sensing data and hydrological environment data are synchronized in time and aligned in spatial coordinates; Based on the fusion features, an obstacle distribution map and a list of target candidates are obtained.

[0007] In one possible implementation, based on the target underwater vehicle's state, local environment map, and global state table information, cooperative detection with other underwater vehicles is performed, including: When the target submersible's status indicates that the target submersible has malfunctioned or is low on power, determine the undetected area of ​​the target submersible; Based on the information from the unexplored area, the local environment map, and the global status table, the first cooperative underwater vehicle is identified and a cooperative exploration command is generated. Send cooperative detection commands to the cooperative underwater vehicle.

[0008] In one possible implementation, the method also includes: If the confidence level of the target is greater than the second confidence threshold and less than or equal to the first confidence threshold, the second cooperative underwater vehicle is identified and a cooperative confirmation command is sent to the second cooperative underwater vehicle. The joint confidence level of the target is obtained based on the confidence level of the target underwater vehicle with respect to the target and the confidence level of the second cooperative underwater vehicle with respect to the target. If the joint confidence level is greater than the first confidence level threshold, the first target discovery information is sent to other underwater vehicles.

[0009] In one possible implementation, the method also includes: When the target submarine has accumulated effective identification experience in a jammed environment, or when the target candidate list includes new targets, the incremental model parameters of the target submarine's local perception fusion model are compressed and encoded to obtain a compressed data packet. Compressed data packets are sent to other underwater vehicles via underwater acoustic communication packet segmentation, so that the other underwater vehicles can update their local perception fusion models based on the compressed data packets.

[0010] In one possible implementation, the method also includes: Receive compressed status summaries for each underwater vehicle; The mission execution status is assessed based on the summaries of each compressed state. If the increment of the detection coverage rate within a preset time period is less than a preset increment, a detection area reallocation command is issued. Upon detecting an external threat to the underwater vehicle formation, the request was modified, and an emergency evasion command was issued. If the priority of a probe mission changes, a mission redirection command will be issued.

[0011] Secondly, embodiments of this application provide a cooperative detection system for a formation of underwater vehicles, including: a data acquisition module, an environmental map update module, a transmission module, a reception module, and a track update module; The data acquisition module is configured to obtain an obstacle distribution map and a target candidate list based on acoustic data, optical data, depth sensing data, and hydrological environment data collected by the sensors when the target underwater vehicle is exploring along the detection track; each target in the target candidate list corresponds to coordinates and confidence level. The environment map update module is configured to update the target submarine's local environment map based on the obstacle distribution map; The sending module is configured to send a first target discovery message to other underwater vehicles for any target if the confidence level of the target is greater than a first confidence threshold. The receiving module is configured to update the global state table information based on the second target discovery message received by the target submarine from another submarine. The collaboration module is configured to conduct collaborative detection with other underwater vehicles based on the target underwater vehicle's status, local environment map, and global status table information.

[0012] In one possible implementation, the acquisition module is configured to perform feature fusion on acoustic data, optical data, depth sensing data, and hydrological environment data acquired by the sensors through a local perception fusion model when the target underwater vehicle is exploring along the detection track, thereby obtaining fused features; wherein the acoustic data, optical data, depth sensing data, and hydrological environment data are time-synchronized and spatially aligned; based on the fused features, an obstacle distribution map and a target candidate list are obtained.

[0013] Thirdly, embodiments of this application provide a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a cooperative detection method for a submarine formation as described in any embodiment of the first aspect.

[0014] Fourthly, embodiments of this application provide a computer storage medium storing instructions that, when executed on a terminal device, cause the terminal device to perform a cooperative detection method for a submarine formation as described in any embodiment of the first aspect.

[0015] To improve the coordination efficiency and dynamic adaptability of underwater vehicle (UV) formations, this application provides a cooperative detection method for UV formations. When a target UV vehicle is conducting detection along a detection track, an obstacle distribution map and a target candidate list are obtained based on acoustic data, optical data, depth sensing data, and hydrological environment data collected by sensors. Each target in the candidate list corresponds to coordinates and a confidence level. The local environment map of the target UV vehicle is updated based on the obstacle distribution map. For any target, if the confidence level corresponding to the target is greater than a first confidence threshold, a first target detection message is sent to other UV vehicles. In response to the target UV vehicle receiving a second target detection message from other UV vehicles, the global state table information is updated based on the second target detection message. Cooperative detection is then performed with other UV vehicles based on the target UV vehicle's state, local environment map, and global state table information. In this application embodiment, by sending a first target detection message to other UV vehicles and receiving a second target detection message from other UV vehicles, the coordination efficiency and dynamic adaptability among the UV vehicles in the formation are improved.

[0016] In addition, in this embodiment, acoustic data, optical data, depth sensing data and hydrological environmental data are used for target detection to improve target detection accuracy and reduce the false negative rate. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a cooperative detection method for a submarine formation provided in this application embodiment; Figure 2 This is a schematic diagram of multi-sensor sensing fusion data provided in an embodiment of this application; Figure 3 A flowchart of dynamic task allocation provided in an embodiment of this application; Figure 4 A schematic diagram of a message format and communication protocol for collaborative communication provided in an embodiment of this application; Figure 5 This is a schematic diagram of a cooperative detection system for a submarine formation provided in an embodiment of this application. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0020] The terms "first" and "second," etc., used in the specification and claims of this application are used to distinguish different objects, not to describe a specific order of objects. For example, "first operation instruction" and "second operation instruction," etc., are used to distinguish different operation instructions, not to describe a specific order of operation instructions.

[0021] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0022] In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more, for example, multiple processing units means two or more processing units, multiple elements means two or more elements, etc.

[0023] In related technologies, underwater vehicle fleets employ predefined path planning schemes, where operators pre-determine the tracks and detection areas of each underwater vehicle based on known nautical chart information before mission execution. The underwater vehicles then execute tasks according to fixed procedures, lacking online replanning capabilities. They also employ single-sensor perception schemes, meaning existing underwater collaborative detection systems primarily rely on single-type sensor data (such as single-beam sonar, side-scan sonar, or forward-looking sonar) for environmental perception. The data types collected by each underwater vehicle are limited, and different types of perception data (acoustic images, optical images, depth data, hydrological parameters) cannot be effectively fused. Furthermore, they employ rule-based task allocation schemes, where task allocation in multi-underwater collaborative systems is typically based on preset rules and fixed collaborative strategies, lacking the ability to dynamically adjust collaborative strategies according to real-time situational awareness. Finally, they employ centralized control schemes, where existing systems generally rely on surface mother ships or shore-based control centers for centralized task scheduling of multiple underwater vehicles. The autonomous decision-making capabilities of each underwater vehicle are weak, and a large amount of perception data needs to be uploaded to the control center for processing before issuing commands.

[0024] Therefore, the relevant technologies have the following technical problems: First, the system lacks adaptability to dynamic environments, resulting in a high mission failure rate. Existing predefined path planning schemes cannot cope with complex and dynamic changes in the underwater environment (e.g., sudden changes in ocean currents, the appearance of obstacles, sensor malfunctions, etc.). Under complex sea conditions, the mission completion rate of multi-submarine vehicle systems using predefined path schemes is only 55% to 70%, with an average mission failure rate of 30% to 45%. Many missions need to be terminated and replanned, leading to low detection efficiency.

[0025] Secondly, the lack of sensor data fusion capabilities results in low target identification accuracy. Existing single-sensor sensing solutions lead to incomplete and fragmented underwater environmental information acquired by each submersible. Acoustic images have low resolution but long detection range, optical images have high resolution but short detection range, depth data provides three-dimensional structural information, and hydrological parameters reflect environmental characteristics. When these four types of data are processed individually, the accuracy rate of underwater target detection is only 58%~65%, with a false negative rate as high as 35%~42%, which cannot meet the needs of refined detection.

[0026] Third, the rigid collaborative strategies result in low resource utilization among multiple underwater vehicles (UUVs). Existing rule-based engine-based rigid collaborative strategies cannot dynamically optimize task allocation based on real-time mission situation, UUV energy status, and detection progress. In actual test scenarios, rigid collaborative strategies lead to a large number of overlapping detection areas among UUVs (overlap rate of 25%~40%), with an overall detection coverage of only 72%~78%, and the multi-UUV collaborative gain is less than 1.8 times.

[0027] Fourth, real-time collaborative decision-making under underwater communication constraints suffers from significant delays. Existing centralized control schemes heavily rely on underwater acoustic communication (bandwidth of only 10-100 kilobits / second, latency of 100 milliseconds to 5 seconds), resulting in a large amount of sensing data failing to be uploaded in a timely manner. The average delay in issuing collaborative decision-making commands reaches 8-15 seconds, severely impacting the system's real-time tracking and collaborative response capabilities for dynamic targets. When the target's moving speed exceeds 0.5 knots, the system's tracking success rate drops below 40%.

[0028] Fifth, the lack of a cross-submarine knowledge-sharing mechanism results in low group learning efficiency. In the existing system, each submarine's perception model is trained and runs independently, lacking an effective cross-submarine online knowledge-sharing and model co-evolution mechanism. When a submarine detects a new type of target or encounters new environmental interference, its accumulated recognition experience cannot be transferred to other submarines in real time, leading to extremely low group learning efficiency. The overall perception performance improvement cycle in new scenarios can take several days to several weeks, failing to meet real-time task requirements.

[0029] In this embodiment, by sending a first target detection message to other underwater vehicles and receiving a second target detection message from other underwater vehicles, the underwater vehicles possess real-time obstacle avoidance and mission replanning capabilities, improving the coordination efficiency and dynamic adaptability among the underwater vehicles in the formation. In this embodiment, acoustic data, optical data, depth sensing data, and hydrological environmental data are used for target detection, improving target detection accuracy and reducing the false negative rate.

[0030] The technical solution of this application will now be described in conjunction with the accompanying drawings.

[0031] See Figure 1 The figure is a flowchart of a cooperative detection method for a submarine formation provided in an embodiment of this application.

[0032] like Figure 1 As shown, the cooperative detection method of underwater vehicle formation includes the following steps: S1000: When the target underwater vehicle is conducting exploration along the detection track, it obtains an obstacle distribution map and a target candidate list based on the acoustic data, optical data, depth sensing data and hydrological environment data collected by the sensors; each target in the target candidate list corresponds to coordinates and confidence level.

[0033] When a formation of underwater vehicles (UVs) performs a certain exploration mission, each UV is assigned a sub-exploration mission under that mission. Each sub-exploration mission corresponds to a exploration track, so that the corresponding UV is able to navigate according to the corresponding exploration track and complete the corresponding sub-exploration mission.

[0034] The target underwater vehicle can be any underwater vehicle in the underwater vehicle formation. For example, if the underwater vehicle formation includes underwater vehicle A, underwater vehicle B, underwater vehicle C, and underwater vehicle D, the target underwater vehicle can be underwater vehicle A, underwater vehicle B, underwater vehicle C, or underwater vehicle D. This application does not impose specific limitations on any of these underwater vehicles.

[0035] Sensors may include forward-looking sonar, underwater cameras, Doppler velocity log (DVL) sensors, and conductivity-temperature-dept (CTD) sensors. The forward-looking sonar acquires acoustic data, the underwater camera acquires optical data, the Doppler velocity log acquires depth sensing data, and the CTD sensor acquires hydrological environmental data.

[0036] The surface support vessel sends mission parameter packages to each submersible in the formation. Each submersible records its local perception fusion model and confirms that all sensors are functioning correctly. Subsequently, each submersible descends to the inspection depth according to its assigned track and begins its inspection mission. When a target submersible performs its corresponding inspection mission, it uses the local perception fusion model to fuse acoustic, optical, depth sensing, and hydrological environmental data collected by the sensors to obtain fused features. The acoustic, optical, depth sensing, and hydrological environmental data are time-synchronized and spatially aligned. Based on the fused features, an obstacle distribution map and a target candidate list are obtained.

[0037] like Figure 2 As shown, the above steps include a data alignment preprocessing stage, a feature extraction stage, a fusion processing stage, and an output result stage.

[0038] In the data alignment preprocessing stage, time synchronization (e.g., accuracy ≤10ms) and spatial coordinate alignment are performed on four types of sensor data (acoustic data, optical data, depth sensing data, and hydrological environment data), and they are uniformly projected onto the coordinate system of the unmanned underwater vehicle body to eliminate the timing inconsistencies caused by the difference in sampling frequency of different sensors.

[0039] In the feature extraction section, an acoustic encoder is used to process the acoustic image, outputting a 512-dimensional feature vector; an optical encoder is used to process the optical image, outputting a 768-dimensional feature vector; a point cloud encoder is used to process the depth-sensing data (point cloud data), outputting a 256-dimensional feature vector; and a hydrological encoder is used to process the hydrological environment data, obtaining a 128-dimensional feature vector.

[0040] In the fusion processing stage, an attention-weighted fusion mechanism is used to fuse the four types of features, and the weights of each sensor are dynamically adjusted according to the current underwater visibility.

[0041] For example, when visibility is less than 3m, the weight of acoustic data is increased to 0.6 and the weight of optical data is decreased to 0.1; when visibility is greater than 10m, the weight of optical data is increased to 0.4 and the weight of acoustic data is decreased to 0.4.

[0042] In the output stage, based on the fused 1024-dimensional features, the system outputs an obstacle distribution map (e.g., grid resolution 0.5m), a list of target candidates (coordinates + confidence level), and an environmental risk assessment score (0~1).

[0043] In this embodiment, acoustic data, optical data, depth sensing data, and hydrological environmental data are used for target detection to improve target detection accuracy and reduce the false negative rate.

[0044] S2000: Update the local environment map of the target submarine based on the obstacle distribution map.

[0045] In this embodiment of the application, the surface mother ship can incrementally update the local environment map corresponding to the target underwater vehicle based on the obstacle distribution map corresponding to the target underwater vehicle.

[0046] S3000: For any target, if the confidence level of the target is greater than the first confidence level threshold, send the first target discovery message to other underwater vehicles.

[0047] The first target discovery message can be 64 bytes of data, including the submarine's location, battery level, mission status, and the identifier of the covered area, broadcast every 5 seconds. Other submarines can learn the target submarine's location, battery level, mission status, and the identifier of the covered area through the first target discovery message.

[0048] When the confidence level of a target exceeds a first confidence level threshold (e.g., 0.75), target information compression encoding is triggered, compressing the target information into a compact 128-byte representation and broadcasting it to other underwater vehicles via underwater acoustic communication. For uncertain targets with confidence levels between a second confidence level threshold (e.g., 0.5) and a first confidence level threshold (e.g., 0.75), a collaborative confirmation request is triggered, sending instructions to a second collaborative underwater vehicle to coordinate secondary confirmation of the suspicious target from different perspectives. If the combined confidence level of the target underwater vehicle and the second collaborative underwater vehicle regarding the target exceeds the first confidence level threshold (e.g., 0.75), the target is confirmed as valid and recorded in the global status table information, and a first status broadcast message is sent to other underwater vehicles.

[0049] It should be noted that there can be multiple second cooperative underwater vehicles in the embodiments of this application.

[0050] S4000: In response to the target submarine receiving a second target discovery message from another submarine, update the global state table information based on the second target discovery message.

[0051] The second target discovery message can be 64 bytes of data, containing the submarine's location, battery level, mission status, and covered area identifier, broadcast every 5 seconds. The target submarine can obtain global status table information through the second target discovery message. The global status table message includes the location, battery level, mission status, and covered area identifier of other submarines.

[0052] S5000: Based on the target underwater vehicle's status, local environment map, and global status table information, it conducts cooperative detection with other underwater vehicles.

[0053] If the target submersible's status indicates a malfunction or low battery, determine the undetected area of ​​the target submersible; based on the undetected area, the local environment map, and the global status table information, determine the first cooperating submersible and generate a cooperative detection command; send the cooperative detection command to the cooperating submersible.

[0054] In the embodiments of this application, such as Figure 3 As shown, at the start of the detection mission, the total detection area is divided into N×M equal-area sub-grids (for example, the grid side length ranges from 50 to 200m, set according to the mission accuracy requirements), and a greedy allocation algorithm based on energy constraints is used to allocate the sub-grids to each underwater vehicle. During mission execution, whenever a status broadcast message from any underwater vehicle is received, the module reassesses the current detection progress. If the coverage is greater than or equal to the coverage threshold, or if the battery power of each underwater vehicle is lower than the battery power threshold, a return to base and data transmission are triggered. Whenever a status broadcast message from any underwater vehicle is received, if a redistribution trigger condition is detected, the remaining incomplete sub-grid list is obtained and the first cooperating underwater vehicle is determined; a cooperative detection command is sent to the first cooperating underwater vehicle and a redistribution command is broadcast.

[0055] It should be noted that if the target submarine malfunctions and returns to base or has insufficient power, its remaining undetected subgrids will only be redistributed to other submarines for detection if a redistribution trigger condition is detected.

[0056] In this application embodiment, the method of determining the first cooperative underwater vehicle is not specifically limited. For example, the redistribution priority is sorted according to the following rules: the underwater vehicle that is closest to the target has priority (weight 0.5), the underwater vehicle with the most remaining power has priority (weight 0.3), and the underwater vehicle with the lightest current mission load has priority (weight 0.2).

[0057] In this embodiment, by sending a first target detection message to other underwater vehicles and receiving a second target detection message from other underwater vehicles, the underwater vehicles acquire real-time obstacle avoidance and mission replanning capabilities, improving the coordination efficiency and dynamic adaptability among the underwater vehicles in the formation. Furthermore, this embodiment employs acoustic data, optical data, depth sensing data, and hydrological environmental data for target detection, improving target detection accuracy and reducing the false negative rate.

[0058] Based on the aforementioned embodiments, the following section will introduce the cooperative communication of underwater vehicle formations.

[0059] like Figure 4 As shown, the cooperative communication of underwater vehicles includes the following three methods: The first communication method involves periodically sending status broadcast messages (including position coordinates, remaining battery power, mission status, and hash values ​​of the covered area) during timed broadcasts (e.g., a timed broadcast period of 5 seconds). The status broadcast messages are compressed and sent to the surface mother ship, enabling the surface mother ship to maintain the global status table information of each underwater vehicle in the formation in real time based on the status broadcast messages.

[0060] The second communication method is event-triggered. When the confidence level of a detected target is greater than the first confidence threshold, a target discovery message is sent (the target discovery message may include the target category ID, confidence level, target coordinate bounding, and multi-dimensional feature embedding vector); a compressed status broadcast message is sent to the surface mother ship, so that the surface mother ship can maintain the global status table information of each underwater vehicle in the formation in real time according to the status broadcast message.

[0061] The third communication method is planning triggering. If any submarine in the formation malfunctions or has insufficient power, the first cooperating submarine is identified and a cooperative detection command is generated. The cooperative detection message is then sent to the first cooperating submarine (the cooperative detection command may include a task reassignment command, a subgrid ID list, and priority weights).

[0062] To facilitate understanding of the technical solution of this application, the entire process will be described below with reference to two specific embodiments.

[0063] Example 1: A detection formation consisting of four identical unmanned underwater vehicles (UUVs) (maximum speed 3 knots, maximum range 110 km, battery capacity 15 kWh) was used to conduct an inspection mission on a subsea natural gas pipeline in a certain sea area. The mission required completing full-coverage detection of a 100-meter-wide corridor along a 20 km pipeline within 8 hours. Target types included four categories: pipeline suspension, pipeline corrosion, foreign object attachment, and pipeline fracture. The specific implementation included the following five stages: In the first phase, the mission initialization phase, from 0 to 5 minutes, the surface mother ship sequentially sends mission parameter packets (e.g., each mission parameter packet is 256 bytes) to the four submersibles via underwater acoustic communication. Each submersible completes the loading of its local perception fusion model (e.g., loading time is about 45 seconds) and confirms that the sensors are working properly (e.g., the data frame rate of all four types of sensors meets the standard). Subsequently, each submersible descends to an inspection depth of 5 meters from the seabed (e.g., seabed depth is about 80 meters) according to its assigned initial trajectory and begins to perform the inspection mission.

[0064] In the second phase, the normal inspection and operation phase lasted 5 minutes to 6 hours. Each submersible advanced at a constant speed of 1.5 knots along the pipeline, fusing data from four types of sensors and updating the local environmental map in real time every 200 milliseconds. Every 5 seconds, each submersible broadcast a status broadcast message (e.g., 64 bytes), with a measured communication latency of approximately 0.8 seconds within the formation. Submersible 3's side-scan sonar detected a strong echo anomaly at subgrid number 128 (confidence level 0.81, exceeding the threshold of 0.75), triggering a target detection message broadcast. The other three submersibles received the message and updated their local global target list. Submersible 3 continued its detailed detection and switched to its optical camera for close-range confirmation (adjusting the depth to 2 meters from the target). It was ultimately confirmed to be an approximately 3-meter-long fishing net entanglement, with the target's final confidence level increasing to 0.94. The target's position coordinates (latitude, longitude, and depth: accuracy ±0.8 meters, determined by the cumulative error of the position sensor) and target feature images were recorded.

[0065] In the third phase, the dynamic replanning phase, approximately 6 hours later, Submarine 2 broadcasts a battery warning message (remaining battery power has dropped to 21%, approaching the 20% return threshold). Upon receiving this message, the group collaborative task planning module immediately reassesses the task allocation. Submarine 2 has 8 remaining unfinished subgrids (numbers 93-100). Based on the priority sorting algorithm: Submarine 1 is the closest (approximately 2.3 km from the unfinished area, weight 0.5, score 0.81), has 62% remaining battery power (weight 0.3, score 0.74), and has completed 60 tasks (weight 0.2, score 0.70), resulting in the highest overall score. Ultimately, the 8 subgrids are reassigned to Submarine 1. A reassignment coordination command message (256 bytes) is broadcast. Upon receiving the command, Submarine 1 updates its local task planning within 0.3 seconds, adjusting its course to head towards the takeover area. Submarine 2, upon receiving confirmation, triggers the return-to-base protocol and safely returns to the surface. The entire redistribution process has an end-to-end latency of 1.1 seconds (including a communication latency of 0.8 seconds and a local calculation latency of 0.3 seconds).

[0066] In the fourth phase, the cross-submarine indication synchronization phase, approximately 4 hours later, Submarine 3 detected a new type of target in subgrid 142—a large area of ​​soft coral attachment on the pipe surface (different from the four preset target types). After accumulating 60 new samples, the model incremental update broadcast was triggered. Submarine 3 calculated the model incremental parameters (original size 1.8 megabytes), compressed them to 168 kilobytes (compression ratio approximately 10.7:1), and sent them sequentially to the other three submarines via underwater acoustic communication in 12 data packets (14 kilobytes each). Due to the bandwidth limitation of the underwater acoustic channel (measured effective bandwidth approximately 28 kilobits / second), the transmission of all data packets took approximately 49 seconds. Each receiving submarine completed model merging (federated average weight: Submarine 3's sample size 60, accounting for 100%, this was a single-source update directly replacing the incremental update) and used 10 local historical validation samples for regression validation. The validation accuracy of all three receiving submarines did not decrease, and the model update took effect. Subsequently, submersible 1 independently detected similar coral attachment targets in subgrid 47 and submersible 4 in subgrid 178, with confidence levels of 0.79 and 0.82, respectively, both exceeding the threshold, indicating that knowledge dissemination was effective.

[0067] The fifth phase, mission completion phase, lasted approximately 7-8 hours. By the 7.8-hour mark (including the additional time for Submarine 1 to take over the area after Submarine 2's return), the overall detection coverage of the formation reached 96.5% (193 / 200 subgrids completed detailed inspection; 7 subgrids experienced slight timeouts due to obstacle avoidance path adjustments, but coverage was subsequently replenished). Each Submarine surfaced sequentially, establishing a broadband wireless connection (54 Mbps) with the surface mother ship and uploading complete detection data: approximately 18 gigabytes of raw acoustic image data and 6 gigabytes of optical video per Submarine, compressed locally and uploaded to approximately 4.2 gigabytes per Submarine, totaling approximately 16.8 gigabytes for all four Submarines, with an upload time of approximately 42 minutes. The surface monitoring module fused and stitched together the detection data from the four submersibles to generate a complete inspection report for the 20-kilometer pipeline. A total of abnormal targets were found: one entanglement with fishing nets, two sections of pipeline suspended in the air (4.2 meters and 6.8 meters in length, respectively), three areas of coral attachment, and one area of ​​slight external wall corrosion. The positioning accuracy of all targets was better than ±1.5 meters.

[0068] The embodiments of this application significantly improve the mission completion rate of multi-submarine vehicles and dynamically adapt to complex environments. Each submarine vehicle has real-time obstacle avoidance and mission replanning capabilities, increasing the mission completion rate from 55%~70% in the prior art to 88%~93%, an improvement of approximately 30%. In complex sea state tests simulating sudden changes in ocean currents and the appearance of random obstacles, the system of this invention achieved an average completion rate of 89% for 50 consecutive missions.

[0069] Significantly improves underwater target detection accuracy and reduces false negative rate. The multi-sensor perception fusion module integrates data from four types of sensors through an attention-weighted fusion mechanism, increasing the underwater target detection accuracy from 58%~65% of existing single-sensor technologies to 87%~92%, and reducing the false negative rate from 35%~42% to 8%~13%, an improvement of approximately 35%. In tests conducted in turbid waters with visibility less than 1 m, the system of this invention still maintains a target detection accuracy of 82%.

[0070] Optimize detection coverage and enhance multi-submarine vehicle (UV) collaborative gains. The swarm collaborative mission planning module dynamically allocates detection sub-regions based on real-time global status table information, reducing the overlap rate of repeated detection areas among UVA vehicles from 25%~40% in the existing fixed strategy to 5%~8%, increasing detection coverage from 72%~78% to 93%~96%, and enhancing multi-UV collaborative gains from less than 1.8 times to 2.6~3.1 times.

[0071] Significantly reducing collaborative decision-making latency and enhancing dynamic target tracking capabilities, the distributed collaborative communication module, through extremely low-bandwidth adaptive compression coding, compresses the total communication bandwidth of multi-submarine vehicle collaboration to less than 10 kilobits per second. The end-to-end latency for collaborative decision-making is reduced from 8-15 seconds in existing centralized solutions to 1-2 seconds, a reduction of 85%. In dynamic target tracking tests (target speed 0.5-2 knots), the tracking success rate of this invention's system is increased from below 40% in existing solutions to 79%-85%.

[0072] The system achieves continuous evolution in its group perception capabilities, significantly shortening the adaptation cycle to new scenarios. The cross-submarine knowledge sharing and model update module, through a federated knowledge sharing mechanism, reduces the propagation latency of new target identification experience within the formation from several days to weeks in existing schemes to less than 2 minutes. In a test where four submarine vehicles continuously performed a 72-hour detection mission, the system's overall accuracy in identifying new target types steadily improved from an initial 62% to 84%, demonstrating a measurable and continuous evolutionary trend in its group perception capabilities.

[0073] In summary, this invention effectively solves the core problems in existing multi-submarine vehicle cooperative detection technologies, such as poor dynamic adaptability, lack of multi-sensor fusion, low cooperative efficiency, large communication latency, and stagnation of group learning, and realizes true autonomous cooperative detection of multiple submarine vehicles in complex underwater environments.

[0074] Example 2: A search and rescue mission for a sunken ship using a formation of six underwater vehicles in strong ocean currents. This example describes a variation of the invention under more complex environmental conditions, building upon Example 1.

[0075] Mission Scenario: A shipwreck has occurred in a certain sea area. A full-coverage search must be completed as quickly as possible within a 2km x 2km search area to locate any possible survivor compartments or important items. Current Sea State: Seabed depth approximately 50 m, ocean current speed 0.8~1.5 knots (direction varies randomly), underwater visibility approximately 2 m (turbid water), sea state level 4.

[0076] Key parameter adjustments: Number of underwater vehicles: 6 (2 more to cope with the accelerated energy consumption caused by strong ocean currents); Multi-sensor attention weight settings: Due to the visibility of approximately 2m, the system automatically adjusts the acoustic data weight to 0.58, the optical data weight to 0.12, the depth perception weight to 0.18, and the hydrological weight to 0.12; Sub-grid side length: adjusted to 50m×50m (1600 sub-grids in total) to meet the needs of fine-grained searching in low visibility conditions; Obstacle avoidance trigger threshold of end-side decision module: reduce the forward monitoring distance from 50m to 30m (to adapt to the situation where the maneuverability of the submersible is reduced under strong ocean currents). The energy consumption threshold for returning to shore has been increased to 30% (the energy consumption for returning to shore under strong ocean currents is approximately 1.6 times that under calm waters).

[0077] The key operating process is as follows: Approximately 45 minutes into the mission, a strong ocean current caused Submarine 2 to deviate from its preset track by more than 15 meters (after detecting a position deviation exceeding the threshold of 10 meters, the track correction was automatically triggered, adjusting the thrusters to compensate for the ocean current's influence; the correction time was approximately 20 seconds). This deviation resulted in a coverage gap of approximately 200 square meters near subgrid number 312 for Submarine 2. After receiving the position correction message broadcast by Submarine 2, the group collaborative mission planning module automatically marked this blank area as an uncovered grid and reassigned it to the nearest Submarine 5 (approximately 180 meters away). Submarine 5 completed supplementary exploration of this area within 8 minutes, without any surface intervention.

[0078] Approximately 1.5 hours into the mission, Submersible 4 detected a large, regular metallic echo target (abnormally high echo intensity, confidence level 0.88) acoustically in subgrid number 521, triggering a target discovery broadcast. Due to the limited optical visibility of only 2 meters, the underwater target detection and classification module initiated a collaborative close-range confirmation process, dispatching Submersible 3 and Submersible 6 to approach the target from different directions (within 3 meters). After acquiring optical images, the multi-sensor fusion confidence level increased to 0.96, ultimately identifying it as the main hull of the sunken ship (approximately 12 meters × 4 meters × 3 meters). The system automatically performed a fine scan of the area (switching to fine-grained search mode, temporarily adjusting the subgrid side length to 10 meters × 10 meters, and the fine scan range to 50 meters × 50 meters). Three sealed compartment structures were discovered inside and around the main hull, with a coordinate positioning accuracy of ±0.6 meters (achieving high-precision positioning even in turbid water).

[0079] Summary of key performance data: Task completion time: 5 hours (originally planned for 6 hours); Detection coverage: 94.1% (due to the need for some submersibles to frequently correct their tracks in strong ocean currents, some areas were not covered). Detection rate of key targets (main hull and sealed compartments of the sunken ship): 100%; Target detection accuracy under acoustic modality dominance: 84% (slightly lower than calm water, but much higher than the 61% of existing single acoustic solutions); Autonomous recovery time after ocean current disturbance: average 22 seconds; End-to-end delay for collaborative decision-making: 1.4s on average (the quality of underwater acoustic communication channels is slightly reduced in strong ocean current environments). This embodiment verifies the robustness and adaptability of the system of the present invention in extreme underwater environments such as strong ocean currents and low visibility, as well as the effectiveness of the multimodal weight adaptive adjustment mechanism under different environmental conditions.

[0080] In addition, a cooperative detection system for a submarine formation in this application includes: a data acquisition module 1000, an environmental map update module 2000, a transmission module 3000, a receiving module 4000, and a cooperative module 5000. The data acquisition module 1000 is configured to obtain an obstacle distribution map and a target candidate list based on acoustic data, optical data, depth sensing data and hydrological environment data acquired by the sensors when the target underwater vehicle is conducting exploration along the detection track; wherein each target in the target candidate list corresponds to coordinates and confidence level; The environment map update module 2000 is configured to update the local environment map of the target submarine based on the obstacle distribution map; The sending module 3000 is configured to send a first target discovery message to other underwater vehicles for any target if the confidence level of the target is greater than a first confidence threshold. The receiving module 4000 is configured to update the global status table information based on the second target discovery message received by the target submarine from another submarine. The Cooperative Module 5000 is configured to conduct cooperative detection with other underwater vehicles based on the target underwater vehicle's status, local environment map, and global status table information.

[0081] In this embodiment, by sending a first target detection message to other underwater vehicles and receiving a second target detection message from other underwater vehicles, the underwater vehicles acquire real-time obstacle avoidance and mission replanning capabilities, improving the coordination efficiency and dynamic adaptability among the underwater vehicles in the formation. Furthermore, this embodiment employs acoustic data, optical data, depth sensing data, and hydrological environmental data for target detection, improving target detection accuracy and reducing the false negative rate.

[0082] In one possible implementation, the acquisition module is configured to perform feature fusion on acoustic data, optical data, depth sensing data, and hydrological environment data acquired by the sensors through a local perception fusion model when the target underwater vehicle is exploring along the detection track, thereby obtaining fused features; wherein the acoustic data, optical data, depth sensing data, and hydrological environment data are time-synchronized and spatially aligned; based on the fused features, an obstacle distribution map and a target candidate list are obtained.

[0083] In one possible implementation, the coordination module is configured to, when the target submarine's status indicates that the target submarine has malfunctioned or has insufficient power, determine the undetected area of ​​the target submarine; determine the first cooperating submarine and generate a cooperative detection command based on the undetected area, the local environment map, and the global status table information; and send the cooperative detection command to the cooperating submarine.

[0084] In one possible implementation, the sending module is configured to: determine the second cooperating underwater vehicle and send a cooperation confirmation command to the second cooperating underwater vehicle when the confidence level corresponding to the target is greater than a second confidence level threshold and less than or equal to a first confidence level threshold; obtain the joint confidence level of the target based on the confidence level of the target underwater vehicle and the confidence level of the second cooperating underwater vehicle regarding the target; and send first target discovery information to other underwater vehicles when the joint confidence level is greater than the first confidence level threshold.

[0085] In one possible implementation, the system also includes: a model update module; The model update module is configured to compress and encode the incremental model parameters of the target submarine's local perception fusion model to obtain a compressed data packet when the target submarine has accumulated effective identification experience in a jammed environment, or when a new target is included in the target candidate list; and send the compressed data packet to other submarines through underwater acoustic communication packetization so that the other submarines can update their respective local perception fusion models based on the compressed data packet.

[0086] In one possible implementation, the instruction issuing module: The command issuing module is configured to receive compressed status summaries of each submarine; evaluate the execution status of the detection mission based on each compressed status summary; issue a detection area reallocation command when the increment of the detection coverage indicated by the execution status within a preset time period is less than a preset increment; issue an emergency evasion command when a request for modification of an external threat to the submarine formation is detected; and issue a mission redirection command when the priority of the detection mission changes.

[0087] In addition, embodiments of this application also provide a computer device and a computer storage medium.

[0088] The computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the cooperative detection method for a submarine formation as described in any of the foregoing embodiments.

[0089] A computer storage medium, the computer-readable storage medium storing instructions, which, when executed on a terminal device, cause the terminal device to perform the cooperative detection method of the submarine formation as described in any of the foregoing embodiments.

[0090] It should be understood that the computer equipment and computer storage medium described in the embodiments of this application have the same beneficial effects as the cooperative detection method of submarine formation.

[0091] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the device and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments. The device and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components indicated as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the solution in this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0092] The above description is merely one specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A cooperative detection method for a formation of underwater vehicles, characterized in that, The method includes: When the target underwater vehicle explores along the detection track, it obtains an obstacle distribution map and a target candidate list based on acoustic data, optical data, depth sensing data, and hydrological environment data collected by the sensors; wherein, each target in the target candidate list corresponds to coordinates and confidence level; Update the local environment map of the target submarine based on the obstacle distribution map; For any of the aforementioned targets, if the confidence level corresponding to the target is greater than a first confidence threshold, a first target discovery message is sent to other underwater vehicles; In response to the target submersible receiving the second target discovery message sent by the other submersibles, the global status table information is updated according to the second target discovery message; Based on the status of the target underwater vehicle, the local environment map, and the global status table information, it conducts cooperative detection with other underwater vehicles.

2. The method according to claim 1, characterized in that, When the target underwater vehicle is conducting exploration along its detection trajectory, it obtains an obstacle distribution map and a target candidate list based on acoustic data, optical data, depth sensing data, and hydrological environment data collected by sensors, including: When the target submersible is conducting exploration according to the exploration trajectory, the acoustic data, optical data, depth sensing data, and hydrological environment data collected by the sensors are fused using a local perception fusion model to obtain fused features; wherein the acoustic data, optical data, depth sensing data, and hydrological environment data are time-synchronized and spatially aligned; Based on the fusion features, the obstacle distribution map and the target candidate list are obtained.

3. The method according to claim 1, characterized in that, The step of conducting cooperative detection with other underwater vehicles based on the status of the target underwater vehicle, the local environment map, and the global status table information includes: If the status of the target submersible indicates that the target submersible has malfunctioned or has insufficient power, determine the undetected area of ​​the target submersible; Based on the undetected area, the local environment map, and the global status table information, the first cooperative underwater vehicle is identified and a cooperative detection command is generated; Send the cooperative detection command to the first cooperative underwater vehicle.

4. The method according to any one of claims 1-3, characterized in that, The method further includes: If the confidence level corresponding to the target is greater than the second confidence level threshold and less than or equal to the first confidence level threshold, then the second cooperative underwater vehicle is identified and a cooperative confirmation command is sent to the second cooperative underwater vehicle. The joint confidence level of the target is obtained based on the confidence level of the target underwater vehicle with respect to the target and the confidence level of the second cooperative underwater vehicle with respect to the target. If the joint confidence level is greater than the first confidence threshold, the first target discovery information is sent to the other underwater vehicles.

5. The method according to claim 2, characterized in that, The method further includes: When the target submarine has accumulated effective identification experience in a jammed environment, or when the target candidate list includes new targets, the incremental model parameters of the target submarine related to the local perception fusion model are compressed and encoded to obtain a compressed data packet. The compressed data packet is sent to other underwater vehicles via underwater acoustic communication packet segmentation, so that the other underwater vehicles can update their respective local perception fusion models based on the compressed data packet.

6. The method according to any one of claims 1-3, characterized in that, The method further includes: Receive a compressed state summary of each of the aforementioned underwater vehicles; The execution status of the probe mission is assessed based on the compressed state summary described above; If the increment of the detection coverage within the preset time period is less than the preset increment, a detection area reallocation command is issued. Upon detecting an external threat to the underwater vehicle formation, an emergency evasion command is issued; If the priority of the detection task changes, a task redirection command is issued.

7. A cooperative detection system for a formation of underwater vehicles, characterized in that, include: The system includes a data acquisition module, an environmental map update module, a sending module, a receiving module, and a collaboration module. The acquisition module is configured to obtain an obstacle distribution map and a target candidate list based on acoustic data, optical data, depth sensing data, and hydrological environment data acquired by the sensors when the target submersible is conducting exploration along the detection track; wherein, each target in the target candidate list corresponds to coordinates and confidence level; The environment map update module is configured to update the local environment map of the target submarine according to the obstacle distribution map; The sending module is configured to send a first target discovery message to other underwater vehicles for any of the targets if the confidence level corresponding to the target is greater than a first confidence threshold. The receiving module is configured to update the global status table information according to the second target discovery message in response to the target submarine receiving the second target discovery message sent by other submarines; The collaboration module is configured to conduct collaborative detection with other underwater vehicles based on the status of the target underwater vehicle, the local environment map, and the global status table information.

8. The system according to claim 7, characterized in that, The acquisition module is configured to, when the target submersible is conducting exploration according to the exploration trajectory, perform feature fusion on the acoustic data, optical data, depth sensing data, and hydrological environment data acquired by the sensors through a local perception fusion model to obtain fused features; wherein the acoustic data, optical data, depth sensing data, and hydrological environment data are time-synchronized and spatially aligned; and based on the fused features, obtain the obstacle distribution map and the target candidate list.

9. A computer device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the cooperative detection method for a submarine formation as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a terminal device, cause the terminal device to perform the cooperative detection method for a submarine formation as described in any one of claims 1-6.

Images